Print 128 comment(s) - last by vignyan.. on Sep 12 at 1:21 AM

AMD "Barcelona" die shot.  (Source: AMD)

AMD's Opteron and upcoming Phenom logos.  (Source: AMD)

AMD guidance suggests the breakout of AMD's three thermal envelops for Opteron.  (Source: AMD)

AMD's performance estimates compared to comparatively priced Intel processors. (Fluent 6.4.3 is the actual version used, AMD made a typo)  (Source: AMD)

AMD's explanation of the new Average CPU Power metric.  (Source: AMD)
AMD pulls the wraps off "Barcelona," partners are now receiving shipments

AMD is prepared to launch its next-generation Barcelona CPU architecture this Monday. Barcelona is the first K8-based product to feature a substantial amount of architectural changes since the original launch of AMD’s Opteron and Athlon 64 processors. Substantial architectural changes aside, Barcelona features evolutionary enhancements to the existing K8.

Barcelona is the company’s first quad-core CPU architecture and features a native quad-core design. Intel’s previously released Clovertown, Kentsfield and upcoming Harpertownand Yorkfieldquad-core processors feature two Core-architecture dies on a single package – effectively quad-core, however, not a native design like Barcelona.

AMD equips Barcelona with plenty of new tweaks and features to boost performance. New features of Barcelona include tweaked cache, memory controller, branch predictors, prefetch logic, power management and additional AMD-V extensions.

Barcelona’s cache configuration includes L3-cache – a feat AMD has not taken advantage of since its K6-III+ and K6-2+ processors. All CPU cores on Barcelona-based processors share 2MB of L3-cache. L1 and L2-cache remain unchanged with 128KB of L1-cache per core and 512KB of L2-cache per core. The cache configuration is unchanged with 2-way associative L1-cache and 16-way associative L2-cache. The shared L3-cache is 32-way associative.

Barcelona-based processors feature a total of 4.5MB of on-die cache. In comparison, Intel’s Clovertown and Kentsfield quad-core architectures feature 64KB of L1-cache, 4MB of shared L2-cache per pair of cores for a total of 8.25MB on-die cache.

AMD tweaked Barcelona’s memory controller for greater bandwidth efficiency and lower latency. This time around, AMD took a different approach for the memory controller. Instead of a single 128-bit wide memory controller, AMD split the memory controller into two 64-bit wide memory controllers. This allows the memory controllers to achieve greater efficiency by operating independently.

AMD designed the new memory controller with future memory technologies in mind. Barcelona will initially debut with support for DDR2 memory, but it’s first refresh, in the form of Shanghai, will support DDR3 memory.

New to the memory controller is a DRAM prefetcher. The DRAM prefetcher intelligently prefetches data it deems useful in the future. DRAM prefetching does not store data in the L1, L2 or L3-caches as it has access to its own buffer.

Barcelona features a new 512-entry indirect branch predictor – a feat Intel debuted on its Pentium M processor. The new indirect branch predictor reduces mispredicted branches for greater efficiency. Greater efficiency also translates into lower power consumption as well.

In addition to the new 512-entry indirect branch predictor, Barcelona has improved prefetcher logic too. The new prefetcher logic retains the same two prefetchers per core as the K8 architecture; however, AMD has tweaked it for greater performance. With the new improved prefetcher logic, Barcelona brings prefetched data directly into the L1-cache. AMD’s K8 architecture brought prefetched data into L2 cache.

AMD’s SSE implementation sees substantial upgrades as well. Barcelona increases the SSE execution width to 128-bits. K8 featured an SSE execution width of 64-bits that can execute two 64-bit SSE instructions at the same time.

Although K8 featured parallel 64-bit SSE instruction execution capabilities, 128-bit SSE instruction execution required extra time to divide the 128-bit instructions into two 64-bit operations. This allows Barcelona to execute SSE instructions quicker than K8.

SSE instruction fetch bandwidth is also improved. Instruction fetch bandwidth increases to 32-bytes per cycle over the previous 16-bytes per cycle of K8 with Barcelona. AMD increased the internal interconnect between the memory controller to L2-cache to 128-bits per cycle over K8’s 64-bits per cycle too.

AMD’s Barcelona has power management changes too. With Barcelona, the power planes are split, allowing the processor and memory controller to operate independently at different speeds and voltages. However, to take advantage of split power planes, a new motherboard is required, as current motherboards lack the required power circuitry.

Each processor core can dynamically adjust its clock speed depending on load too. The new power management features allow Barcelona quad-core processors to operate with the same thermal envelope as current dual-core Opteron processors.

Lastly, AMD has added new AMD-V instructions. The new instructions provide hardware acceleration of shadow paging – allowing guest operating systems to have independent memory management. AMD refers to the new feature as nested paging.

All the architectural improvements and quad-cores bring the Barcelona transistor count to 463-million transistors. Intel’s Kentsfield features 582-million transistors, though it has nearly twice as much cache. Nevertheless, Barcelona-based processors will be manufactured on a 65nm fabrication process.

Power consumption of quad-core Barcelona processors is identical to dual-core counterparts. AMD has three thermal bins for Barcelona, similar to dual-core models. Standard, HE low-power and SE high-performance thermal bins will be available. However, AMD will not launch SE models until Q4’07.

AMD’s ACP measures the entire CPU’s power draw, including cores, memory controller and HyperTransport links. The measurements are conducted using “commercially useful high utilization workloads,” according to AMD’s Barcelona presentation.

The workloads used to measure ACP include TPC-C, SPECcpu2006, SPECjbb2005 and STREAM. AMD ACP ratings result in lower power consumption numbers, which the company claims is more reflective of real world use, instead of the overestimation of the TDP rating system.

AMD Opteron 2300 Series


2.0 GHz 95W$372

23471.9 GHz 95W$312
2347 HE
1.9 GHz 68W$372
2346 HE
1.8 GHz 68W$251
2344 HE
1.7 GHz

AMD Opteron 8300 Series


2.0 GHz 95W$1,004

83471.9 GHz 95W$774
8347 HE
1.9 GHz 68W$861
8346 HE
1.8 GHz 68W$688

AMD has nine Barcelona-based Opteron 2300 and 8300 series models set for launch. Launch clock speeds range from 1.7 GHz-to-2.0 GHz, with higher speeds available Q4'07. The company also expects speeds to ramp up to 2.3 GHz and above in Q4’07 with SE-bin models, which typically have 120-Watt TDPs.

Expect AMD to debut Barcelona-based Opteron 2300 and 8300-series on September 10. Socket AM2 users looking for a quad-core processor will have to wait until later this year for Budapest-based single-socket Opteron or Agena-based Phenom X4 and FX processors.

Comments     Threshold

This article is over a month old, voting and posting comments is disabled

the first wait is almost over
By nerdye on 9/7/2007 6:37:41 PM , Rating: 3
This article leaves me starving for some benchmarks, seeing how inexpensive some of these chips truly are has me concerned. Whether or not these chips beat their intel counterparts we all are dying to see the advantages of this new architecture. I'm assuming that AMD will not be able to fight intel for the heavy weight performance crown until ramping the clock speed on Barcelona a bit. I hope 4th quarter this year (the second wait)is the time if it is not this upcoming monday.

RE: the first wait is almost over
By zpdixon on 9/7/2007 8:46:33 PM , Rating: 5
Look at this comparison of the Opteron 2350 (2.0 GHz) against the Xeon 5345 (2.33 GHz):

Despite a 14% slower clock frequency, the Opteron is:
- faster : between 7% and 189% faster than the Xeon [1]
- cheaper : $372 vs. $455

I told you so ! [2]

Interestingly enough, it means that the only quad-core Xeon (2P) processors faster than the Opteron 2350 are the Xeon X5355 and X5365 at, respectively, $744 and ~$1100. IOW, since the high-end processors ($700+) represent only 10-20% of the 2P market, Barcelona seems the fastest/cheapest solution for 80-90% of the 2P market !

Can't wait for the 2.3 GHz Barcelona announced for Q4 :)

[1] These AMD's benchmarks should give reasonably accurate numbers, at least as accurate as the Intel's benchmarks for Core2 in Q1/Q2 2006.
[2] "K10 will assuredly regain the lead over the Core2 microarchitecture:"

RE: the first wait is almost over
By TomZ on 9/7/2007 8:57:48 PM , Rating: 2
Don't get so aroused yet - remember you're looking at AMD marketing benchmark numbers. Let's wait for real, third-party benchmarks from AnandTech and the like before we get all luvvy-duvvy.

RE: the first wait is almost over
By Spuke on 9/7/2007 9:47:22 PM , Rating: 3
Oh so it's ok to use Intel's benchmarks and get all giddy over those but AMD's benchmarks are crap? Right! Quit bogarting Tom. Puff puff pass...puff puff pass.

RE: the first wait is almost over
By TomZ on 9/7/2007 9:59:40 PM , Rating: 4
No, I distrust both AMD and Intel-supplied benchmarks equally. Obviously it is each company's best interest to cherry-pick benchmarks to show the superiority of their product relative to the other. That happens all the time.

The only first-party benchmarks that I would trust even a little would be those where all the details are provided so they can be independently verified. Better yet is when a third-party can "witness" the benchmark running on real hardware. Even then you don't see a complete performance picture because the hardware vendor is guiding you towards only seeing results that flatter their product, again.

AMD's benchmark slide offers none of the above assuances of authenticity, hence I would take them with a grain of salt.

RE: the first wait is almost over
By vignyan on 9/12/2007 12:06:22 AM , Rating: 2
Hey... completely in agreement with you... The STREAM benchmarks are made by AMD themselves! I would like to see the day where a company posts benchmarks that had poor performance wrt competition.

Also, the SPECfp benchmark really projects the bandwidth of the processor... which AMD has advantage with its NUMA... IT DOES NOT GIVE THE ACTUAL FLOATING POINT PERFORMANCE... SPECfp is not even a realistic usage model... even in the MP segments!

RE: the first wait is almost over
By nerdye on 9/7/2007 9:00:21 PM , Rating: 2
I do also hope and forsee amd's barcelona as being a great bang for the buck, but we honestly cannot judge its performance based on 8 benchmark tests, obviously barcelona will have advantages in some benchmarks over core 2, but not until a complete set of tests ran by anandtech/dailytech is completed can we fully speculate on true performance of these new chips. The wait continues.

RE: the first wait is almost over
By shabby on 9/7/2007 9:03:51 PM , Rating: 1
You actually believe those "benchmarks"? You seem very gullible... your the perfect amd customer.

RE: the first wait is almost over
By zaki on 9/7/2007 10:53:56 PM , Rating: 2
im with shabby, until 3rd party people test out these new amd cpus, i wont be convinced of their performance,

also, arent there new (higher clocked) intel cpus coming out, regardless, of course amd wont mention the tests in which the probably still cant beat intel, its anybody's guess, despite this single benchmark pic.

BUT i really hope these benchmarks are correct, that would be awesome for the consumers.

RE: the first wait is almost over
By wetlegs6 on 9/8/07, Rating: 0
RE: the first wait is almost over
By Continuation on 9/8/07, Rating: 0
By omnicronx on 9/8/2007 6:30:45 PM , Rating: 2
And you sound like the sort of person that likes to compare apples to oranges.. Lets wait for the real benchmarks people before the flaming commences. Especially since Barcelona is the server model, and phenom is barely on the horizon..

And lets face it people, there is not going to be a clear cut winner here, AMD is going to be much better at floating point, and Intel is going to be much better at integer calcs. There is no way in my mind that either of these cpus are going to kick ass in every area with the discrepancies between the two designs.

RE: the first wait is almost over
By slickr on 9/8/2007 12:18:35 AM , Rating: 2
Come on there. We all know internal benchmarks no mather if Intel's or AMD's they always increase the numbers in favor of their own products by at least 20%
Although i have no doubt AMD new quad core processors will be faster in some benchmarks i still think Intel will manage to win in some benchmarks too.

So overall we should wait for 3rd party benchmarks in order to get a clearer picture.

But lets not get our hopes too high for AMD beacose Intel are just 5-6 months from releasing new cores, while AMD will depend on Barcelona until Q2 2008 before they ship revised core!

RE: the first wait is almost over
By defter on 9/8/2007 2:19:51 AM , Rating: 2
Are you kidding? Do you really think that those benchmarks tells a lot of Barcelona's real-world performance? It's quite easy to cherry pick obscure benchmarks to show that you are in the lead (especially when comparing your fastest product to third fastest product from the competition). Who has heard before of "Fluent 6.3.4" and "Neon refined revised"? And what software "3 vehicle collision" uses??

RE: the first wait is almost over
By fungmak on 9/8/2007 2:57:12 AM , Rating: 2
Both are engineering software and both are very computationally intensive. However, i can't remember which two architectures it has favoured in the past.

FYI Fluent is computatioanl fluid dynamics and the 3 graphs at the bottom, of which "3 vehicle collision" is LS-DYNA: a dynamics finite element analysis sofware package.

By Amiga500 on 9/8/2007 1:58:50 PM , Rating: 2
I use Fluent 6.3

You wanna program that will ask serious questions of your CPU, look no further than any CFD code.

[Can't wait till they lever GPUs onto it though!]

RE: the first wait is almost over
By maroon1 on 9/8/2007 5:17:02 AM , Rating: 2
The benchmarks that were tested were synthetic benchmarks

Synthetic benchmarks don't tell you which processor is faster.

For example, 6400+ beats E6750 in synthetic benchmarks, but in real world applications and gaming, E6750 is around 5-10% faster than 6400+

By Justin Case on 9/9/2007 8:16:55 PM , Rating: 3
The SPEC benchmarks use actual code from real world software (ex., povray, gcc, bzip, h.264, etc.).

FLUENT and the other programs tested are also very much "real-world applications".

There is more to computing than Doom 3, you know?

RE: the first wait is almost over
By bradley on 9/7/2007 9:27:09 PM , Rating: 3
I hope we're not seeing quad core simply for its own sake. Just like we got DDR2 because the issue and hands were forced. Yet, I actually do see a few really significant things in the architecture... including Average CPU Power, with split power planes allowing each processor and memory controller to operate at significant power savings based on usage. This is far more exciting to me than the equivalent Intel offering, probably based on practicality alone. I've always appreciated AMD's elegant approach to the brute-force method. Otherwise, I don't see real-world quad core fully realized for a long time to come.

TDP - AMD Changes Position
By TomZ on 9/7/2007 9:02:53 PM , Rating: 2
You'll also notice that AMD is now adopting the same measure of power that Intel has been using for many years. AMD is calling it ACP, but it is effectively the same definition that Intel has been using for TDP.

Therefore, I am hoping we don't have to have any more debates about whose definition of TDP is right or wrong! Both companies are measuring power the same way.

RE: TDP - AMD Changes Position
By Captain Orgazmo on 9/7/07, Rating: 0
RE: TDP - AMD Changes Position
By KernD on 9/7/2007 11:16:57 PM , Rating: 2
"because it is impossible to force true 100% cpu usage with normal programs" !!! no it's not, I don't know if you remember the TDP on the P4 dual core extreme edition with HT? it was 115W, and that was stated as the what the cooler had to dissipate, yet at my work we had to add sleep in our code to prevent the application from overheating these processor, on which we were developing on, to the point of shutting down... and it did that in about 15 minute of running our software. It would seem they measured there TDP on single threaded applications...

RE: TDP - AMD Changes Position
By TomZ on 9/7/2007 11:27:41 PM , Rating: 2
Well that's just sillyness. If your processor was overheating, then you should have fixed the hardware, not handicapped the software. Computer OEMs should be validating their cooling solutions with the Intel tool - that's its purpose.

I do agree, however, that it is not impossible to drive all the CPU cores to 100%. I could write a program in 10 minutes that would accomplish that.

By Captain Orgazmo on 9/8/2007 12:09:36 AM , Rating: 2
Of course you can write a program to force 100% CPU usage (like I said, TAT, benchy/stress tool), but regular programs using normal instructions only typically utilize up 90% of the maximum possible operations per second of the processor, just because it is damn near impossible to optimize code well enough for a program doing multiple things to always utilize 100% of the processor (and of course programs aren't written for just 1 processor, but with a generic instruction set designed to work on all processors in a class).

As far as the P4 EE, I figure Intel may have fudged the numbers a tad on the TDP, and I definitely agree that cooling was totally inadequate (having a fan running at 5400 RPM, sounding like a lawnmower is hardly acceptable for a modern CPU; the first thing I did with my old Prescott was swap in a Zalman).

RE: TDP - AMD Changes Position
By Justin Case on 9/9/2007 8:47:14 PM , Rating: 2
Did you completely miss his point, or are you just being obtuse?

He develops software, and that software has to run on millions of systems worldwide, equipped with the manufacturer-provided cooling solution, which in turn is based on Intel's specifications (which, in the case of some P4 models, were extremely optimistic, to put it mildly).

"Fixing" his system (by adding better-than-"required" cooling) would not magically prevent the software from overheating all the other systems (and freezing them or causing the CPUs to throttle down, making execution slower in the end).

In fact, the same thing happened to several people (I mean the ones not living in Bizarro world):

Oh, and congratulations on your l337 c0d1ng 5ki11z, but "driving all cores to 100%" is meaningless. You can have a CPU with no idle cycles which runs perfectly cool. It depends on what instructions it's executing (and, sometimes, on what specific combinations of instructions, and how those translate into superscalar microcode; the 80486 was a long time ago).

And it's precisely because of all those variables that a real (absolute maximum) TDP value is important.

RE: TDP - AMD Changes Position
By TomZ on 9/10/2007 9:03:53 AM , Rating: 2
1. How did you figure out that the OP's software is going to millions of users? Way to read between the lines!

2. Sure, I realize that certain types of instructions generate more heat than others (e.g., floating-point compared to integer), but it's not exactly rocket science, now is it?

3. I'm not debating that maximum TDP specification is not a good thing. If I were designing PCs, I would think this is exactly the kind of information that would be needed to design a system that would work well under all conditions.

RE: TDP - AMD Changes Position
By Justin Case on 9/10/2007 11:48:56 AM , Rating: 2
1. Where did I say "his software is going to millions of users"? I said it has to be able to run on millions of systems (those things called PCs, you know?). How would the number of PCs affect the CPU temperature in each one? It could end up running just on two, but if one of those two was a P4, it would risk freezing / throttling down due to the "optimistic" Intel-supplied TDP values. What's the alternative? Print "doesn't run on Intel CPUs" on the box? Shot in the foot, no?

2. Actually, unless you are extremely familiar with the CPU's internal architecture (not the ISA), how x86 instructions translate into microcode, and how branch prediction and cache management work in each particular CPU model, yes, it is (a lot more complicated than) rocket science. Just look at the errata for any modern CPU and multiply the complexity by all the instructions that don't have errors. Which is why AMD's ACP value (and Intel's pseudo-TDP value) is based on a few well-known benchmarks, and not any particular combination of instructions.

3. Yes, you are. You've repeatedly posted that Intel's "maximum likely power" approach to TDP is fine, and that it never causes any problems, despite the fact that several people have told you otherwise.

RE: TDP - AMD Changes Position
By TomZ on 9/10/2007 3:19:19 PM , Rating: 2
1. On the other hand, it could be an app that they only use internally. As I'm sure you're aware, most developers are writing apps used within their own companies, as opposed to only the very small minority of devs that work on apps sold in the millions of copies.

2. FUD. You can make is sound complicated if you want, but I don't buy it. Making the processor hot is not that complicated. If you don't believe me, then answer me this: If it were such a function of everything you said, then why would the various CPU burn-in apps be written so that a single app runs on a whole much of processors? If what you're saying were true, then these apps would be very expensive and have to be highly specialzed based on the processor they are using, which is not the case.

3. Any measure of TDP is fine, as long as AMD or Intel explain what it means, so the designer can determine whether it is suitable for their use. For example, if I am designing a cheap home computer, then relying on maximum likely power or maybe even AMD's average power might be okay, and let the processor throttle above that. If I'm making an enterprise workstation, then I might be more interested in worst-case TDP. Saying one definition is better than another is like saying measuring distance in km is better than measuring in miles.

RE: TDP - AMD Changes Position
By Justin Case on 9/11/2007 11:32:07 AM , Rating: 2
Given your obviously overwhelming understaing of CPU architecture, I feel it is pointless to go on...

RE: TDP - AMD Changes Position
By KernD on 9/7/2007 11:04:59 PM , Rating: 2
There just showing, if they were to use a method who you say is similar to Intel, they would have lower number. Where do they say that there going to sell them with ACP numbers? all I see in the price list is the same old TDP. Maybe there just showing these numbers so that people don't forget that there not using the same method for TDP which explains why they might have higher number than Intel in TDP.

RE: TDP - AMD Changes Position
By Treckin on 9/7/2007 11:26:00 PM , Rating: 2
ACP is a method for calculating TDP...

RE: TDP - AMD Changes Position
By rdeegvainl on 9/8/2007 2:44:21 AM , Rating: 1
weird coincidence

RE: TDP - AMD Changes Position
By Justin Case on 9/9/2007 9:01:20 PM , Rating: 3
No, it isn't. AMD continues to supply TDP numbers (maximum possible CPU power dissipation). ACP stands for "Average CPU Power (while running CPU-intensive benchmarks)". In other words "maximum likely power dissipation".

On average, ACP is about 20% lower than TDP (varies with each model). If ACP were "a method for calculating TDP", as you claim, the numbers would be identical.

ACP is useful when estimating system power consumption (and it's somewhat useful when comparing to Intel's "optimistic TDP" numbers) but I expect any sane OEMs to continue basing their cooling solutions on the (real) TDP figure.

RE: TDP - AMD Changes Position
By TomZ on 9/10/2007 3:22:44 PM , Rating: 2
You are assuming there is only one "correct" way to measure TDP. Remember, each chip manufacturer is able to define TDP however they please. If AMD tomorrow decided to market all their parts based on TDP = average power, that would not be an unreasonable thing to do.

Whether or not you (or I) like a particular manufacturer definition of TDP doesn't make it right or wrong.

RE: TDP - AMD Changes Position
By Justin Case on 9/11/2007 11:25:27 AM , Rating: 2
Of course it does.

You might as well say that "it's not wrong for Intel / AMD / whoever to simply claim a TDP of 5 watts for every CPU". It is wrong (both "morally" and in objective engineering terms).

Unless you want systems based on that CPU to malfunction, then any "right" way to measure TDP has to result in a value that is as high as the power the CPU can possibly dissipate. That can be calculated in a number of ways, some simple, some terribly complex, but the resulting value can never be lower than the real power dissipation of the CPU, under any circumstances.

Using "average" or "maximum likely" values for TDP is irresponsible and, in the long run, a shot in the foot. The only reason why Intel uses "estimated maximum likely blah blah blah" instead of real maximum thermal power (based on the chip's electrical characteristics) is that the real TDP for some P4 models was so embarrassingly high.

And the result of that was that several people got "burned" (ha-ha), and had to stop trusting Intel's numbers.

AMD has already stated that they will continue to supply the (real) TDP value, and that people should use that value when designing the system's cooling. The ACP value is just to give an idea of typical power consumption under load. My guess is they decided to add that due to all the retarded "web journalists" who kept comparing TDP values to conclude which CPU had better "performance per watt".

TDP values won't tell you that, of course. But neither will ACP. Different systems use different kinds of RAM, different chipsets, etc., so you always need to measure the whole thing, not just the CPU.

RE: TDP - AMD Changes Position
By MartinT on 9/8/2007 1:51:17 AM , Rating: 1
You'll also notice that AMD is now adopting the same measure of power that Intel has been using for many years. AMD is calling it ACP, but it is effectively the same definition that Intel has been using for TDP.

That's BS, both Intel and AMD have been using roughly the same TDP definition for years now, and neither's TDP equals the maximum possible thermal dissipation of their products.

ACP, on the other hand, is AMD's desperate try to seem more power efficient than they really are. What's the use for a spec that's apparently well short of the needed cooling power? Marketing BS, pure and simple.

RE: TDP - AMD Changes Position
By fungmak on 9/8/2007 3:04:23 AM , Rating: 2
Yes, they might have been using the same definition but if you look back at the X2s and A64s compared to the P-Ds and P4s, AMD's measured power consumption was short of their estimation under extreme load. Intel, on the other hand, were at the same level as their TDP if not slightly higher.

RE: TDP - AMD Changes Position
By TomZ on 9/8/2007 9:36:59 AM , Rating: 2
You sure about that?

If we are to believe AMD's claims that Intel uses dubious TDP numbers and has more power hungry chipsets on the motherboard, then we should be seeing a difference far lower than 50 watts and perhaps even a wash. But looking at the highlighted figures, AMD's overall power consumption was actually 66 watts higher than Intel!

RE: TDP - AMD Changes Position
By gescom on 9/8/2007 1:04:52 PM , Rating: 1
"AMD Beats Intel in 36 of 57 Power Efficiency Tests"

The end of the story :)

RE: TDP - AMD Changes Position
By TomZ on 9/8/2007 2:25:54 PM , Rating: 1
Yeah, that's nice, but it's also off-topic. We are talking about TDP definitions in this thread.

RE: TDP - AMD Changes Position
By omnicronx on 9/8/2007 6:32:56 PM , Rating: 2
To tell you the truth tomz, i have always been weary with the way Intel claims to calculate TDP. What do they consider high or max load anyways.

RE: TDP - AMD Changes Position
By TomZ on 9/10/2007 3:26:40 PM , Rating: 2
From that link above, it would seem that AMD's TDP numbers are the ones that are more suspect, not Intel's.

Compaired with C2D?
By RandomStyuff on 9/7/2007 6:41:03 PM , Rating: 2
Does anyone know how this will compare to Core 2 "arch"? It just seems like AMD took all the features found in C2D, added some more, and mixed it with K8... (I'm not saying thats a bad thing)

Any guesses on performance?

RE: Compaired with C2D?
By TheDoc9 on 9/7/2007 7:21:57 PM , Rating: 2
I've seen a few rumors suggesting it's about 20-30% faster per clock in floating point vs. Intels Kentfield. One said 30% slower in integer. They are just rumors...

RE: Compaired with C2D?
By rsasp on 9/7/2007 7:26:32 PM , Rating: 2
let's hope they're true

RE: Compaired with C2D?
By michael2k on 9/7/2007 7:55:03 PM , Rating: 2
Too bad the Barcelona is 33% slower clocked than a Kentfield...

RE: Compaired with C2D?
By DannyH246 on 9/9/2007 2:26:02 PM , Rating: 2
Yes the B0 revision is. Not the B2 revision.

RE: Compaired with C2D?
By MartinT on 9/8/2007 1:56:21 AM , Rating: 2
Those numbers weren't rumors, they were direct quotes from Michael Dell:
If you look at floating point instructions, Barcelona is about 30 percent faster than Clovertown. However, if you look at integer instructions, Clovertown is about 30 percent faster than Barcelona.

My personal guess is that he refers to SPECfp_rate2006 (known Barcelona strength) and SPECint_rate2006 (known Core strength), respectively.

RE: Compaired with C2D?
By TomZ on 9/10/2007 3:25:07 PM , Rating: 2
Who cares what Michael Dell says? Why do you assume he is fully informed and completely truthful?

I'd trust benchmarks by someone like AnandTech 10X over what Michael Dell says, any day.

RE: Compaired with C2D?
By vignyan on 9/12/2007 12:56:17 AM , Rating: 2
As some one rightly put, its the SPECfp benchmark that might be faster for AMD because of NUMA but thats not the actual floating point performance.

the new logos
By Gul Westfale on 9/7/2007 11:38:32 PM , Rating: 2
i like the phenom one, it looks dynamic. the opteron one reminds me of the recycling thingie on plastic bottles though... ah well. let's just hope we get to see some real benchmarks soon!

RE: the new logos
By Locutus465 on 9/8/2007 1:17:02 AM , Rating: 2
I'm pretty loyal to AMD, but honestly.... I think the phenom logo looks like a paper airplane :P

RE: the new logos
By TomZ on 9/8/2007 11:07:32 AM , Rating: 2
Actually, I think the Phenom logo is supposed to be a shuttlecock, like when you play badminton.

RE: the new logos
By MartinT on 9/8/2007 1:59:37 AM , Rating: 3
They'd be great, if they were logos for the AMD Donut® and AMD Party-Hat® processor series.

Unfortunately, they're not.

RE: the new logos
By rdeegvainl on 9/8/2007 2:49:07 AM , Rating: 2
I think that would really stir up the market if those were the names of the processors.
AMD Donut® and AMD Party-Hat®

RE: the new logos
By zsdersw on 9/8/2007 9:58:21 AM , Rating: 2
AMD's marketing sucks.. and the logos for Barcelona and Phenom are just another demonstration of that.

RE: the new logos
By mrdelldude on 9/9/2007 12:00:12 PM , Rating: 2
The Opteron one reminds me of the wait cursor in Vista.

Potential AMD Strategy?
By jtemplin on 9/7/2007 8:04:21 PM , Rating: 3
I was just thinking. If the Barcelona architecture is expected to deliver performance on par, clock to clock, to the Kentsfield archtecture. And if the processors are expected to be less costly, to directly comparable intel parts. Given the truth of these ifs, is it possible that AMD may try to push the "4x4" (now 8x8?) platform into the upper mainstream perhaps?

I would be crazy to assert that there would be a price/performance ratio twice as good as competing intel processors (which would then make the dual socket VERY mainstream...).More realistically; if the competing intel part (2.4 ghz I know..bear with me ; )) was only 30% more expensive, some of the upper mainstream power users may lick their lips at the thought of 8 cores under the hood.

I haven't done in depth research so constructive criticism is welcomed! Let me know if anyone has heard anything about the future of AMD consumer dual-socket for the future.


RE: Potential AMD Strategy?
By TomZ on 9/7/2007 8:51:01 PM , Rating: 1
I think that on the desktop there is only a small niche market for 8 cores, and I don't think that will change within the next 3 years (assumed lifecycle of this family of parts). We're only now beginning to start using the second core in some kinds of applications (although Windows kind of makes that happen automatically anyway), not to mention cores 3 and 4.

Therefore AMD is right to focus more on higher-margin server and HPC applications with these processors, especially since it may be the case that AMD has some performance advantage there relative to Intel. But since AMD has not allowed any real benchmarks yet, it's hard to tell.

Personally, I have no desire to run dual-socket on my workstation again. Considering the performance of today's dual- and quad-core processors, I don't feel there is for me enough of a performance benefit to justify the extra space/heat/cost of a separate processor. I don't run highly-specialized applications that needs tons of CPU, however; I'm more just a power user who runs lots of different apps, oftentimes at the same time.

RE: Potential AMD Strategy?
By JackPack on 9/7/2007 9:06:50 PM , Rating: 2
Intel is planning to introduce its Skulltrail dual-socket platform soon. If AMD wants to push 4x4, it'll probably be a short lived effort. Intel's MCM quad-core products should be less expensive to manufacture. With 45nm imminent, their advantage is even greater.

RE: Potential AMD Strategy?
By Targon on 9/8/2007 8:53:12 AM , Rating: 1
45nm for AMD is set for 2008, so AMD isn't all that far behind in that transition.

RE: Potential AMD Strategy?
By zsdersw on 9/8/2007 9:55:47 AM , Rating: 2
And where's the 45nm tape-out?

By Master Kenobi (blog) on 9/10/2007 7:29:06 AM , Rating: 3
Yes, typically the tape-out happens 1 year before product release. I haven't heard of a tape-out. Given the delays on the 65nm front, I would say 45nm in 2009.

Event Details
By crystal clear on 9/8/2007 8:14:42 PM , Rating: 3

Quad-Core AMD Opteron™ processor launch celebration
Monday, September 10, 2007 6:30 p.m. PT
Webcast Presentation Help
Click here for webcast

Title Quad-Core AMD Opteron™ processor launch celebration
Date and Time Monday, September 10, 2007 6:30 p.m. PT
Duration 1 Hour 30 Minutes

RE: Event Details
By KristopherKubicki (blog) on 9/8/2007 8:57:44 PM , Rating: 2
I've already seen the documentation for the event. It's exactly the same as what we posted already :)

RE: Event Details
By mrdelldude on 9/8/2007 10:55:17 PM , Rating: 2
So after the ND expires on Monday, we'll be getting some benchmarks. Right?

RE: Event Details
By KristopherKubicki (blog) on 9/8/2007 11:06:06 PM , Rating: 2
That's what I hear. I've gone through a dozen chips and a dozen boards and never got a retail sample to work correctly.

RE: Event Details
By JumpingJack on 9/9/2007 11:32:02 PM , Rating: 2
When you say 'fail to work correctly', is this failing to POST, or boot or locks up minutes after boot?

What exactly is fails to work correctly?


By ilkhan on 9/7/2007 7:19:34 PM , Rating: 2
What I want to see are benches of Agena vs Yorkfield, both at 2.0Ghz (for clock vs clock results) and each at the highest launch speed (2.0 vs 3.16 it seems) for max performance.

Actually, looking at those launch speeds, it seems like agena is pretty much dead in the water regardless if it doesn't have a 3:2 performance advantage per clock...

RE: benches
By ttnuagadam on 9/7/2007 10:03:40 PM , Rating: 2
this is just the server market. Price vs performance and performance vs wattage are far more important than raw speed. I want to think i remember reading that we should be seeing 3ghz phenoms by the end of the year.

RE: benches
By TomZ on 9/7/2007 10:08:37 PM , Rating: 3
What year? (Kidding)

RE: benches
By ttnuagadam on 9/7/2007 10:32:46 PM , Rating: 2
hehe. i know that a month or 2 ago AMD was demoing stranglehold on a 3ghz agena that wasnt using any fancy cooling. so im hoping we'll start hearing about them soon. im still waiting for some benchies so i can decide who to go with when i build a new pc in the spring.

Where's The Article Text?
By TomZ on 9/8/2007 9:42:25 AM , Rating: 3
It's missing as of this moment.

RE: Where's The Article Text?
By AmbroseAthan on 9/8/2007 10:52:35 AM , Rating: 2
Also missing to me, or was there just no article text?

RE: Where's The Article Text?
By TomZ on 9/8/2007 11:04:37 AM , Rating: 2
There was quite a bit of text last night. Maybe AMD paratroopers stormed the DT headquarters and demanded that Kris remove the article!

By KristopherKubicki (blog) on 9/8/2007 3:34:20 PM , Rating: 2
Yay interns! It's fixed.

By smilingcrow on 9/8/2007 4:35:48 AM , Rating: 2
Whether they name it TDP, ACP or ABC, for many people the more significant metric is how much their whole PC consumes at idle and ‘typical’ load. With more and more review sites including power consumption data we’ll soon get a picture of how the platforms compare. AMD should do well in the server market due to their native quad design and because Intel are handicapped by FB-DIMMs for now. They have a chipset that supports registered DDR2 due Q4 for low end 2P I believe.

At the desktop level the CPU power consumption is only really an issue at the top end and of course for over-clockers. Penryn is rated at 80W for 3GHz quad core which is remarkable when you consider that certain single core Prescotts had a higher TDP at a lower frequency not so long ago.

By Combatcolin on 9/8/2007 10:43:03 AM , Rating: 2
A cheaper running computer would be very welcome in my household.

By Falloutboy on 9/7/2007 8:34:21 PM , Rating: 3
reminds me alot of the orginal hammer release (I beleave hammer was first released for opteron first aswell). the chip it self was superior to intels Netburst architecture but lacked the speed at first to be a serious competetor. hopefully there able to up the clock as fast as they did with hammer in this situtaion

By Locutus465 on 9/8/2007 1:15:36 AM , Rating: 1
This is the first time I've been truly excited about a new CPU in a long time... So I just need to get this out of the way....


Thanks... I think it's great that AMD has some kind of answer for Core already, time will tell if this is thee answer, I suspect not. But this proves they are far from being out of the game!

RE: AMD!!!
By 3kliksphilip on 9/8/2007 12:28:34 PM , Rating: 2
Already? It's been like, a year, plus we don't know the performance yet. But yes, I hope it's worth it too. So many people I know have gone to Intel (Including me).

AMD x2 4000, 1 GB ram, motherboard, ati x1950 pro and 320 gb hard drive for £300 is such a good deal- if we all bought this I'm sure that AMD would be happy, and everybody would have a great little gaming machine.

By spock81 on 9/7/2007 11:38:33 PM , Rating: 2
Like most of you, I'm excited about the performance of these new chips. And I anticipate that they will shine - but only in the high-end, where multiple CPUs (and effective load-balancing) are already in use.

Simply put, Quad-core is major overkill for a single-user machine (yes, even running Windows :). The software just isn't out there. But for parallel systems employing many CPUs, more cores per chip has to beat the socks off the equivalent configuration using the same (total) number of cores on more chips. To say nothing of the L3 cache and reduced electricity draw..

Of course, for optimal performance, the software will have to be tailored to the exact number of cores, placing related threads (which communicate frequently, or share memory pages) on the same chip.

The other case where quad-core will shine is with virtualization, an obvious cycle-hog..


By GuSel on 9/8/2007 12:34:42 PM , Rating: 2
So, which cpu is going to cause less lag to the client while playing CS:Source?

Waiting until it's relevant
By wordsworm on 9/9/2007 9:08:26 PM , Rating: 2
I think AMD is quite acute to wait awhile longer for quad core on the desktop. After all, as everyone keeps saying, Q6600 doesn't have much software that knows how to use 4 cores. By the time Phenom is out, who knows? perhaps we'll start seeing software that knows how to use quad core cpus and 4 cores will therefore be relevant.

By jay401 on 9/10/2007 10:33:19 AM , Rating: 2
Engadget has the news scoop on Intel's plans:

New chips, end-of-line dates announced for all Solo and Duo Core2s, etc.


By tjr508 on 9/7/07, Rating: -1
By TomZ on 9/7/2007 10:53:37 PM , Rating: 2
SPECint focuses on integer performance, and Barcelona's strength is rumored to be in floating-point processing and memory bandwidth. How important would you say the latter two items are for real-world server applications?

By TomZ on 9/7/2007 11:05:28 PM , Rating: 2
I should add that my intuition would be that typical server applications would not use much floating-point, and so integer performance would be important there. But I recognize that FPU performance is probably most important for HPC applications. Anybody have any ideas on this?

By BitJunkie on 9/8/2007 5:31:40 AM , Rating: 2
..Only that their general purpose CPU is optimised for FP which means they aren't going direclty head to head with Intel. Or to put it another way, they are focusing on a segment of the market with the aim of providing an improved solution compared to the main player in the market and hoping to use this as a foundation to maintain market share (or even grow?!).

"Today our efforts are focused around business solutions that:

Enable resource consolidation in the data center
More effectively drive database-oriented business applications
Grow the emerging world of cluster and grid computing
Shorten time to solutions with innovative client computing
At the center of every successful enterprise is the ability to develop, manage, and analyze the business-critical information that drives day-to-day operations."

From here:

You have to take that at face value and assume that kind of focus has been driving the engineering process to some degree. Whether that's likely to work for them or not is something only time will show.

By BitJunkie on 9/8/2007 6:00:44 AM , Rating: 4
...It's also quite natural for people in the IT world to assume the world revolves around server farms moving general business related data around - emails, streamed video, audio etc. Basically the IT world is naturally internet and general computing focused.

In reality, those systems are the thread that holds together the wider world. At the end of those threads are people doing intersting things with numbers e.g. designing structures, interpretting / processing geophysical data, computational fluid dynmaics, finite element analyses - badging that as HPC and writing it off as relatively insignificant is probably an easy thing to do for most IT folk. AMD appears to be acting on that insight and giving people in the wider world who are modelling real world systems in various ways a much more efficient solution.

Bearing in mind that many of the larger corporations in the world today rely on crunching large volumes of FP data as part of their essential business processes, it's not such a bad insight to base your business on. This is especially so if the main player in the market is churning out CPUs to keep the IT-infrastructure-generalist crowd happy anyway.

By GeorgeOu on 9/8/2007 6:29:29 AM , Rating: 1
You make some great points, but no one is saying HPC doesn't matter; just that it matters less from a pure market share perspective and what it means to AMD’s bottom line. If you're in the HPC market, Barcelona will give you an absolute steal on a $300 chip that beats the top of the line Opteron dual-core processor. The problem for AMD is that they can't price their chips for the HPC market; they have to price it for the general market which is great news for HPC customers but bad news for AMD's financials.

By BitJunkie on 9/8/2007 6:59:55 AM , Rating: 3
I'd need to understand their business model in a lot more detail to be able to reply on that one. Also don't forget that they have to product lines to play a tune on, Phenom is going to cover the general market - and there's the whole issue of total cost of owernship, performance per watt and upgrade paths to throw in to the mix.

Maybe I wasn't clear enough about HPC. Businesses may see major benefits in using a CPU with strong FP performance - without getting into high performance computing clusters etc. Stand alone workstations can have a major boost from that - example is a complex soil-structure interaction model set up in an FE package that may currently take 1 day for the mesh to converge. If you can bring that down to 0.75 days then there are massive savings to be had.

In the business space strong FP performance and lesser integer performance is not a weakness.

By GeorgeOu on 9/8/2007 8:26:33 AM , Rating: 1
"example is a complex soil-structure interaction model set up in an FE package"

Ok stop right there. I have a little bit of an Engineering background and I can appreciate what you're saying here. But what I can tell for sure is that businesses for the most part don't care about this; they care about the mundane boring stuff.

Strong FP and strong INT performance is great, but strong FP and weak INT performance is a killer and it dictates prices. This is why AMD is being forced to set their prices so low on Barcelona. If FP was important or if it had a significant market share, AMD would love to sell Barcelona at higher prices.

By Targon on 9/8/2007 8:50:04 AM , Rating: 2
The initial Barcelona chips running at a low clock-rate is why they need to keep the price down. When AMD gets Barcelona to 3GHz, we might be seeing the new high end Opteron chips with higher price points.

It is also possible that AMD already has a number of performance enhancements for integer in the works, but they just were not ready for these first chips. It could be a silent new rev that suddenly makes integer performance jump by 30 percent for example. I am not saying it will happen, but it may happen.

By GeorgeOu on 9/8/2007 6:01:55 PM , Rating: 3
Getting to 3 GHz or anywhere close is the problem. Getting that massive single die perfect is extremely difficult. If it were easy, AMD and Intel would be doing it.

AMD won't make any major single-threaded enhancements till Bulldozer in 2009/2010 which is their second 45nm chip.

By DannyH246 on 9/9/2007 2:28:15 PM , Rating: 2
Cant be that much of a problem, seen as revision B2 of the silicon is already running at that speed.

By GeorgeOu on 9/9/2007 6:45:40 PM , Rating: 2
Running a hand picked Phenom desktop chip at 3 GHz with overclocked settings with TDP out the window doesn't count. Server chips are a little different to begin with and you have to be able to produce a chip within TDP within good yields. I don't care who you are; scaling the production clock by 50% isn't ever easy.

By BitJunkie on 9/8/2007 10:41:57 AM , Rating: 2
That makes sense George, as I said, I'm not au fait with the key drivers behind chip prices.

My intention was to poke around the possible logic behind AMD having such a strong FP bias in their general purpose CPU design as TomZ asked the question.

It is possible they were hoping to get the balance a little better to fall inline with the "competative integer perf. and outstanding fp perf." situation. And I still stand behind that example, in that you don't need to get in to HPC clusters to see the benefit of improved fp performance.

I'd be interested to see the effect of this "weakness" in integer arithmatic in a real world situation rather than a benchmark though, and how the CPU performs on average - given the 80/20 split you identified.

By GeorgeOu on 9/8/2007 6:08:46 PM , Rating: 3
Actually, Intel Core Micro-architecture smashes AMD K8 at the same 3 GHz clock on SPEC CPU 2006 FP for single and dual thread by 30.8% and 21.22%. The fortunes turn when you get to 4 threads and beyond because of memory bandwidth and AMD is a memory bandwidth monster compared to Intel. But that was a strategic tradeoff for Intel to spend their transistor budget on a faster execution engine and more cache to dominate more than 80% of the market.

When you get to late 2008, Intel will be using CSI and Integrated Memory controllers and they’ll make another huge leap on performance and scalability.

By DannyH246 on 9/9/2007 2:24:37 PM , Rating: 2

"When you get to late 2008, Intel will be using CSI and Integrated Memory controllers and they’ll make another huge leap on performance and scalability."

Just where do Intel fanboys like you get off? What possible evidence do you have to suggest that this would be true?

And dont give me "Well Core 2 beat everything so..." rubbish.

Prior to Core 2 what was Intel cpu's like? ....Crap!!

What makes you think that Intels first attempt at IMC and Hypertransport copy will be better than AMD 3rd attempt. (Because by the time Intels new architecture is out it will be competing against Bulldozer)

By zsdersw on 9/9/2007 7:30:01 PM , Rating: 2
Just where do AMD fanatics like *you* get off? What makes you think AMD's future architecture will be ready at the time necessary to compete with Nehalem/CSI given the delays of Barcelona?

By BitJunkie on 9/9/2007 3:24:43 PM , Rating: 2
"...Intel Core Micro-architecture smashes AMD K8 at the same 3 GHz clock on SPEC CPU 2006 FP for single and dual thread by 30.8% and 21.22%. The fortunes turn when you get to 4 threads and beyond because of memory bandwidth ...."

I think you just proved how forward thinking AMD are. After all, we know that multi-threaded apps are the way forward, so isn't having an advantage with memory bandwidth TODAY a pretty good place to be?

If you can drop the interger obsession and look at the bigger picture these CPU performance figures are actually very good news for the wider business community. The people I know who specify and procure computing systems for engineering, explorationist and scientific applications are very much interested in how long a calculation process will take to execute - that includes both integer and floating point performance. So perhaps you should qualify your comment regarding the mundane somehow.

At the end of the day, the market will decide what price point these CPUs will sit at - the reality of what that means for AMDs business is probably something only they will be able to answer with any confidence right now. At the very least AMD have the potential to acquire a lot of friends in the business and scientific communities.

By mrdelldude on 9/8/2007 1:06:48 PM , Rating: 2
Ok stop right there. I have a little bit of an Engineering background and I can appreciate what you're saying here. But what I can tell for sure is that businesses for the most part don't care about this; they care about the mundane boring stuff.

I would venture to say that a vast majority of server chips on the market do things like OLTP and database warehousing, web serving, virtual server hosting, email/exchange, file-serving/SAN, custom app servers, JSP, etc ...

These are the type of benchmarks I'm interested in seeing how the 7300+, Penryn Xeons compare to Barcelona.

By Amiga500 on 9/8/2007 1:50:59 PM , Rating: 2

Their 2.0GHz model comes in at $372.

Intel's 2.0GHz Penryn is going to be priced at $209.

By their pricing AMD are taking on the Penryn 2.5 GHz with the 2.0 GHz Barcelona.

With all the speculation of an aggressive ramp, you'll see higher prices on quicker components quite quickly. [Also indicated by the linear pricing scale - 1.8GHz is $61 less than the 1.9 which is $61 less than the 2.0. -> no massive pricing for the upper level component... because it isn't there yet]

By Amiga500 on 9/8/2007 1:53:30 PM , Rating: 1
Crap - the 1.7 is a HE model, which artificially increases its price a bit.

Oh, and 372-312 ain't 61 - D'OH!

By mrdelldude on 9/9/2007 11:26:45 AM , Rating: 2
Their 2.0GHz model comes in at $372.

Intel's 2.0GHz Penryn is going to be priced at $209.

By their pricing AMD are taking on the Penryn 2.5 GHz with the 2.0 GHz Barcelona.

Or for $320 you could also be getting DP 2.33GHz Penryn with only 50W (vs 95W for the 2GHz Barcey)

By mrdelldude on 9/9/2007 11:42:35 AM , Rating: 2
Or for $320 you could also be getting DP 2.33GHz Penryn with only 50W (vs 95W for the 2GHz Barcey)

Or a more accurate comparison would be (I wish there was an edit button) a 1.9GHz @65 Barcey vs 2.33GHz @50w Penryn.

By GeorgeOu on 9/8/2007 4:42:59 AM , Rating: 2
SPECint_rate2006 probably represents 80% or more of the market. SPECfp_rate2006 probably represents 20% of the market.

Also, I've made some update to include the most recent September numbers from Intel though it's possible AMD has something newer than 8/20/2007. I also added the SPECfp_rate2006 breakdowns.

It's on the same blog.

By vignyan on 9/12/2007 1:21:27 AM , Rating: 2
Eventually everyone seems to be deceived. The FP performace of both the processors is the same. The bandwidth of the processor makes the difference in the synthetic benchmarks like SPECfp_rate2006 but not in common applications. SPECfp_rate2006 is highly dependent on how quickly you can get data from the main memory into the processor (memory latency), and it requires so much data (memory bandwidth) that it overflows the large caches of even the Intel Core 2 processors.

That said, the SPECfp_rate2006 sure does project the performance in the HPC applications, speech recognition, 3D ray tracing, computational fluid dynamics, quantum physics and quantum chemical computations and weather forecasting. Some of the other applications like 3D rendering, video encoding using floating point.

Lets get somethings back to how they were defined actually. SPECfp_rate2006 != Floating point performance

By RjBass on 9/7/2007 11:01:01 PM , Rating: 2
Ya im not buying it either.

I'll wait a few more days for some better breakdowns of the CPU's before I make any decisions.

I need to build a new small business server in December and i have been waiting for the release of Barcelona before I make any decisions as to what to do with it.

I've waited this long, whats a few more days?

By Proteusza on 9/8/2007 6:52:52 AM , Rating: 2
Do you have proof of that?

Thought not.

Lets wait till the thing actually comes out. I happen to think a 2.0 Ghz Barcelona will be about equivalent to a 2.4 Ghz C2Q, only with lower energy requirements.

By GeorgeOu on 9/8/2007 7:59:35 PM , Rating: 1
Yes I do. Intel's advantage increases with fewer sockets because AMD doesn't have a multiple of memory controllers. Even at 8-threads and 16-threads, Intel Core Microarchitecture at 2 GHz beats Barcelona 2.0 GHz. The only exception is memory bandwidth intensive HPC applications like SPECfp_rate2006 at 4 threads and above where AMD dominates.

"Thought not". The only one not thinking is you my friend.

By Amiga500 on 9/8/2007 11:27:42 PM , Rating: 2
That doesn't quite tally with the numbers you have presented though.

On 4 sockets the x7350 is 1.3375 times faster than the Barcelona.

On 2 sockets this drops to 1.318.

By GeorgeOu on 9/9/2007 3:13:52 AM , Rating: 1
OK here's your proof confirmed by IBM.
"IBM also gives us the single threaded results for SPEC CPU 2006 which allows us to see how fast each core can perform on its own. This is very important for desktop applications (like games) and even server applications because there are some tasks that will always be single-threaded by nature. From IBM’s latest data, we can see that Intel still holds a 45% advantage on SPECint_2006 performance and a 32% advantage on SPECfp_2006. This pretty much confirms what I’ve been saying that Barcelona will not significantly improve single threaded IPC (Instructions Per Cycle) over AMD’s K8 architecture."

By TomZ on 9/7/2007 11:09:58 PM , Rating: 2
Also George, did you receive any SPECfp benchmarks? Don't those usually get peformed along with SPECint?

By GeorgeOu on 9/8/2007 4:45:02 AM , Rating: 2
It's been added as well as updated scores from Intel for September.

No benchmarks for SPECweb2005, SPECjbb2005, SAP, and TPC-C were included and that's too bad for AMD because Intel broke all those records. AMD only had SPECint, SPECfp, and a bunch of other HPC benchmarks that's rarely used in the market.

By biostud on 9/8/2007 9:39:47 AM , Rating: 2
it has been updated with the SPECfp, where it beats Intel.

By TomZ on 9/8/2007 9:41:20 AM , Rating: 2
...and by a pretty wide margin, it would appear.

By 16nm on 9/8/2007 10:07:22 AM , Rating: 2
Doesn't look good for AMD.

Considering how new the 65nm process is to AMD, these numbers are respectable. Remember, these are the first chips and over time AMD will be able to squeeze more performance out of the process. I'm sure they will shoot for 3+ Ghz.

By TomZ on 9/8/2007 10:57:07 AM , Rating: 2
But over time, Intel will be improving performance, too, so you can't think about comparing possible AMD future performance against current Intel performance.

But I think George's point is that 80% of the market looks at integer performance, and that 80% is not likely to see any appeal with Barcelona. This means that AMD will have to keep prices low to compensate, which will continue to hurt AMD's profitability.

It seems like Barcelona will dominate HPC applications. It makes me wonder why they wouldn't target Barcelona specifically for that application, price it high, and forget about the mainstream server market until they can shore up integer performance. I suppose AMD thought about that and decided on the current strategy instead.

By DannyH246 on 9/9/2007 2:17:55 PM , Rating: 2

You obviously haven't been paying much attention to the AMD news lately. The revision of Barcelona that will be launched on Mon is B0. Internally they are already at B2, and are working on B3. The demos that have been spotted around are apparently the B2 revision, which is supposededly already good for 3ghz. It wont be a 6-8 month wait (like it was with the K8) before the new revisions are out. It will be much sooner.

By GeorgeOu on 9/9/2007 7:03:40 PM , Rating: 2
You keep citing the 3 GHz clock when it's a handpicked overclocked part. Core 2 dual-cores have been scaling past 4 GHz since Q3 2006 and quad-cores have clocked to 3.8 GHz in fall IDF. That does not mean Intel gets to sell their parts at those clock speeds because they're not within TDP and they yields aren't there.

Don't confuse overclocked performance with production performance. Until you see Barcelonas overclock to at least 3.4 GHz, don't expect production clock speeds of 3 GHz. We’re talking about servers here where you can’t use overclocking.

By DannyH246 on 9/9/2007 7:26:00 PM , Rating: 2
I agree with some of the points you make, but the main point i was making is, there wont be a huge delay between the initial speeds that everyone is moaning about, and the introduction of faster clocked and lower powered parts.

By TomZ on 9/10/2007 8:53:36 AM , Rating: 1
Considering AMD's recent lack of ability to bring anything to market when they say they will, I wouldn't be too overly optimistic about AMD bring out radically improved parts.

By jazkat on 9/10/2007 11:12:19 AM , Rating: 2
well isnt there a LES B* 3.2 or 3.4 ghz phenom x2 and LES B* 3.2 ghz phemon x4 so.....
i find it hard to beleive anything you post, didnt you write the : amd can potentially smash intel?? so when does this smashing come to light??

"It seems as though my state-funded math degree has failed me. Let the lashings commence." -- DailyTech Editor-in-Chief Kristopher Kubicki

Related Articles
Intel Readies Desktop Penryn Models
August 30, 2007, 7:44 AM
Intel Prices "Penryn" Xeons
July 17, 2007, 4:06 PM
AMD Prepares Phenom FX Lineup
July 2, 2007, 5:06 PM
AMD Quad-core Opteron Models Unveiled
February 5, 2007, 7:02 PM

Latest Headlines
4/21/2014 Hardware Reviews
April 21, 2014, 12:46 PM
4/16/2014 Hardware Reviews
April 16, 2014, 9:01 AM
4/15/2014 Hardware Reviews
April 15, 2014, 11:30 AM
4/11/2014 Hardware Reviews
April 11, 2014, 11:03 AM

Copyright 2014 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki