backtop


Print 73 comment(s) - last by P4blo.. on Aug 7 at 10:50 AM


AMD "Barcelona" die shot  (Source: AMD)
AMD's next-generation launch date is still set for "summer 2007"

AMD is set to debut its third-generation Opteron server processor architecture later this quarter. The company publically announced its upcoming launch next month, but the company's latest processor guidance, released just this week, states otherwise. 

According to partners close to AMD, sampling for Barcelona processors will hit top tier system builders this week. One engineer speaking on conditions of anonymity claims, "We're finally getting close to a launch here, but [AMD] still hasn't finalized the date."

The August 27 launch date originally featured on the last Barcelona roadmap, issued less than a month ago, has since dissapeared.

AMD plans to offer initial quad-core Opterons at core frequencies below 2.0 GHz. The company will launch eight quad-core processors under the Opteron 8000 and 2000 series umbrellas. Standard and high-efficiency (HE) models are ready for launch later this quarter.

Quad-core Opteron 8000 Series Launch
Model
Core
Frequency
L2 Cache
L3 Cache
TDP
83502.0 GHz 4x512KB2MB
95W
83481.9 GHz4x512KB2MB
95W
8348 HE
1.9 GHz 4x512KB2MB
68W

AMD has three quad-core Opteron 8000 series models in two speed grades for launch – the 8350, 8348 and 8348 HE. The processors will ship with 2.0 GHz and 1.9 GHz clock frequencies, respectively.

Quad-core Opteron 2000 Series Launch
Model
Core
Frequency
L2 Cache
L3 Cache
TDP
23502.0 GHz 4x512KB2MB
95W
23481.9 GHz4x512KB2MB
95W
2348 HE
1.9 GHz 4x512KB2MB
68W
2346 HE
1.8 GHz
4x512KB
2MB
68W
2344 HE
1.7 GHz
4x512KB
2MB
68W

Quad-core Opteron 2000-series will have a similar launch schedule with the 2348, 2350 and 2348 HE. These models are identical to the Opteron 8000 series, except the chips can only operate in two-socket systems. The quad-core Opteron 2000 series also gains two lower speed variants – the 2346 HE and 2344 HE. The quad-core Opteron 2346 HE and 2344 HE feature 1.8 GHz and 1.7 GHz clock speeds, respectively. This concludes the quad-core launch lineup for the quarter.

AMD plans to launch the initial batch of quad-core Opterons this quarter. However, customers will not receive initial DVT samples until next month and actual production for customer order fulfillment does not begin until September. AMD still does not have an exact launch date set, but the AMD roadmap states a summer 2007 launch and autumn begins on September 23.

Quad-core Opteron 8000 Series Q4'2007
Model
Core
Frequency
L2 Cache
L3 Cache
TDP
8360 SE
2.5 GHz 4x512KB2MB
120W
8358 SE
2.4 GHz4x512KB2MB
120W
8356
2.3 GHz 4x512KB2MB
95W
8354
2.2 GHz
4x512KB
2MB
95W
8350 HE
2.0 GHz
4x512KB
2MB
68W

Quad-core Opteron 2000 Series Q4'2007
Model
Core
Frequency
L2 Cache
L3 Cache
TDP
2360 SE
2.5 GHz 4x512KB2MB
120W
2358 SE
2.4 GHz4x512KB2MB
120W
2356
2.3 GHz 4x512KB2MB
95W
2354
2.2 GHz
4x512KB
2MB
95W
2352
2.1 GHz
4x512KB
2MB
95W
2350 HE
2.0 GHz
4x512KB
2MB
68W

AMD will ramp up clock speeds of quad-core Opteron 8000 and 2000 series in Q4’2007, with the first quad-core SE models entering production. Eleven quad-core models with clock speeds excess of 2.0 GHz will enter production in Q4’2007 with an undetermined launch date. The new models will top out with a 2.5 GHz clock-speed. AMD promises DVT samples of the new models in early Q4’2007.

Quad-core Opteron 8000 Series Q2'2008
Model
Core
Frequency
L2 Cache
L3 Cache
TDP
8362 SE
2.6 GHz 4x512KB2MB
120W
8358
2.4 GHz4x512KB2MB
95W

Quad-core Opteron 2000 Series Q2'2008
Model
Core
Frequency
L2 Cache
L3 Cache
TDP
2362 SE
2.6 GHz 4x512KB2MB
120W
2358
2.4 GHz4x512KB2MB
95W

Quad-core Opterons will have a 2.5 GHz clock speed cap until Q2’2008 when the first 2.6 GHz 8362 SE and 2362 SE models enter production. AMD has not set DVT availability or launch dates of the 2.6 GHz models yet.

Quad-core Opteron 1000 Series
Model
Core
Frequency
L2 Cache
L3 Cache
TDP
1360 SE
2.5 GHz 4x512KB2MB
120W
1358 SE
2.4 GHz4x512KB2MB
120W
1356
2.3 GHz 4x512KB2MB
95W
1354
2.2 GHz
4x512KB
2MB
95W
1352
2.1 GHz
4x512KB
2MB
95W

AMD also has five quad-core Opteron 1000 series models in the pipeline. These models are only available in speed grades above 2.0 GHz. AMD has three models ready for production in November – the 1356, 1354 and 1352. DVT samples of theses processors are available next month. Higher-clocked 1360 SE and 1358 SE models will enter production in December, with DVT samples available in October.

Expect AMD to launch the first set of Barcelona-based quad-core Opteron 8000 and 2000 series processors later this summer.


Comments     Threshold


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

News out of AMD
By Hippiekiller on 7/25/2007 8:37:39 PM , Rating: 5
Good to see news coming out of the AMD camp finally.

I hope that this whole time they have been giving us the silent treatment, only to unleash Barcelona upon an unsuspecting marketplace.

Might be wishful thinking but it will be great to have competition in the performance area and not just the price area. Heck, who knows, prices may yet be slashed by both sides.

I wonder how accurate all the speculation about "has to be 2.6 ghz to be competitive" talk really is. Guess we shall merely wait a couple months to find out.

Good to see product line ups from AMD finally something solid to look at even if it isn't much. Would be nice to see prices too.




RE: News out of AMD
By ChronoReverse on 7/25/2007 8:41:17 PM , Rating: 5
The delays have really hurt AMD. Right now the Q6600's are selling like hotcakes and even I jumped on that bandwagon (stress testing a 3.3GHz overclock right now).

Hopefully the desktop version of K10 will be competitive.


RE: News out of AMD
By Hippiekiller on 7/25/07, Rating: -1
RE: News out of AMD
By Martimus on 7/25/2007 9:11:40 PM , Rating: 2
Overclocking killed one of my CPU's too. Well, it didn't kill it, but I had to drop the voltage and clock speeds down to below the original speed and voltage after about 1 1/2 years or so. I haven't overclocked a computer since. Of course, now I understand why it did that, after getting a job stress testing circuit components for automobiles. The silicon may run fine at a certain voltage and temperature, but it will slowly change over time. It is rated at that voltage and temperature for a reason, because it can sustain that state for its life expectancy.


RE: News out of AMD
By Operandi on 7/25/2007 11:32:13 PM , Rating: 5
Not likely.

What's probably happening here is a motherboard component slowly degrading over time and thus not supply the clean stable power to your CPU it once did.


RE: News out of AMD
By michael2k on 7/26/2007 12:22:57 AM , Rating: 2
I think he's talking about electromigration, and it is possible.

If you think of electrons as water and silicon atoms like rocks, overclocking pushes up the voltage and current, essentially making the electrons move faster and eroding the silicon it's traveling in.


RE: News out of AMD
By InsaneGain on 7/26/2007 3:28:58 PM , Rating: 2
Yeah , I think the water pressure carrying bits of sand that are wearing down rocks in a river is a pretty good analogy for electromigration. Uhhh except assume that when the rocks in a section completely break down then the water can't flow anymore. Maybe water pressure in a pipe carrying abrasive sand that eat away at the pipe is a better comparison. Momentum transfer from the energized electrons damages the conductors, and higher voltage is like higher water pressure. The smaller the interconnects are, the less matter there is to be eroded until failure.


RE: News out of AMD
By Martimus on 7/31/2007 3:23:55 PM , Rating: 2
Which is another example of exactly what I was talking about, just with a different component.


RE: News out of AMD
By P4blo on 8/7/2007 10:50:12 AM , Rating: 2
I've clocked every chip I've ever had and I've found they *do* tail off a bit, generally after the first few months.


RE: News out of AMD
By TomZ on 7/26/2007 12:06:40 AM , Rating: 2
I agree with Operandi, your conclusion is probably wrong.

More likely cause: your thermal transfer of CPU-to-ambient became less efficient. Aging thermal compound, dust on the HSF, etc. Take that apart, clean it thoroughly, re-apply thermal compound, and you'll probably be back in business. I've seen it myself many times.

Second most likely cause: capacitor aging - electrolytic capacitors do lose some some capacitance with time, which could affect your power supply stability, but I doubt that's the case.


RE: News out of AMD
By verndewd on 7/26/2007 12:56:43 AM , Rating: 2
Electromigration happens to anything that carries current and does so by means of a metal path for said current, and it is a constant. It happens weter you loke it or not. Overclocking increases the current density which is the major accelerant to electromigration, Heat simply magnifies what current density does.

Now if he oc'd well on one mobo and not another it could be the mobo or it could be EM, or it could be EOL damage or many other issues that are derived from high current density.

I would want to suspect the mobo as you guys do, But there is always the chance he could have EM or a weakened dielectric layer(EOL) which would cause alot more heat at stock Voltage.

Em is a constant, putting cold to it doesnt even stop it, it just slows it down.After a year and a half its good to think about a new PSU if you OC as well, that would have been my first test for the system in question.


RE: News out of AMD
By Viditor on 7/25/07, Rating: -1
RE: News out of AMD
By Questar on 7/25/2007 9:40:57 PM , Rating: 4
Funny, I see nothing about performace in that quote.

Perhaps you were referring to some technical advantage described by a blogger as being a demonstarted performance advantage?


RE: News out of AMD
By Viditor on 7/25/07, Rating: 0
RE: News out of AMD
By verndewd on 7/26/2007 1:00:36 AM , Rating: 2
There are no perf numbers to date that shine a kind light on it; The expected benches were removed at the AMD site at the behest of George Ou.


RE: News out of AMD
By trivik12 on 7/26/2007 7:48:46 AM , Rating: 2
He is talking about "inferior technology" of tigerton with FSB and no IMC.He also mentioned than tigerton was "speedy". So I dunno how u can spin that it wont match barcelona without any benchmarks.

The differences in CPU design are significant. Where Intel basically bolts two dual-cores together to make a quad-core, AMD is placing all four cores on a single chip, in all its HyperTransport and NUMA glory. I've found the Operton to be the better choice for lots of workloads, especially RAM-intensive applications, and found Intel's new Xeons to be speedy, but challenged in key areas, such as bus performance and memory access.


RE: News out of AMD
By ScythedBlade on 7/26/2007 8:22:56 AM , Rating: 2
As with the other poster, he spins on the technology of bus speed and memory ... but with Anandtech's reviews, we saw that a 1333 Mhz compared to a 1066 Mhz Bus doesn't affect stuff for Intel too much, as Intel's gone more efficient. The blogger should actually be less biased and know what he's talking about.


RE: News out of AMD
By bim27142 on 7/25/2007 9:18:56 PM , Rating: 2
... just upgraded to an Opteron 165 from my good old A64 3200+ Venice... well, yeah both are "old tech" but just can't afford to throw away my 939 rig now for a total overhaul (AM2 or C2D)...

at least i'm on the dual core bandwagon now?! :)


RE: News out of AMD
By Spoelie on 7/26/2007 8:11:39 AM , Rating: 2
With a good stepping you can get it to 2.6-2.8ghz without much trouble at default voltage. With the 1mb per core that should be able to compete quite well with core² cpus in the 1.86-2.33ghz range.


RE: News out of AMD
By jak3676 on 7/25/2007 9:19:40 PM , Rating: 3
quote:
I wonder how accurate all the speculation about "has to be 2.6 ghz to be competitive" talk really is. Guess we shall merely wait a couple months to find out.


AMD can also compete on price alone too. I'm not saying I want them to do that - I'd much rather see a race for the crown again, but the race for the cheapest chip is equally alluring for most consumers. Down with the days of >$1000 consumer grade CPU's!

I'd love to throw my hat in the ring and start speculating, but its really better to wait and see. I will guess that AMD won't be able to capture both the performance crown as well as be the price leader - but we'll have to wait.


RE: News out of AMD
By Khato on 7/26/2007 2:39:46 AM , Rating: 2
Pfeh, back with the days of $100 budget CPU's! =P

Seriously though, while AMD can be price competitive, it's not a sustainable business model. I love how investors were all happy with AMD in Q2 just because they upped their margins by, what was it, 6%? When the fact remains that they need to raise their ASP by about 1.5x to get that $1.4 billion income up to the point where it matches their expenses...


RE: News out of AMD
By BUL on 7/26/2007 1:31:48 PM , Rating: 2
Totally agree. Looking at AMD's July price cut, the range between the most expensive consumer-level desktop CPU and the least-expensive was ~$120. (I'm excluding notebook CPUs & high-end FXs). Seriously, what kind of business model is this?--You can get an AMD processor that's 2-3x more capable for $120 more... Sure, it moves the higher-end products but the lower-end ones languish. And still, the best Athlon 64 X2 can't beat a low-end C2D...


RE: News out of AMD
By defter on 7/26/2007 2:24:20 AM , Rating: 4
AMD definitely has not giving us the silent treatment. They put out a press release a month ago where they told us that 2GHz Barcelona servers will be available in Q3: http://www.amd.com/us-en/Corporate/VirtualPressRoo...


RE: News out of AMD
By RW on 7/26/07, Rating: 0
RE: News out of AMD
By Amiga500 on 7/26/2007 7:54:02 AM , Rating: 4
AMD is around 1/5th the size of Intel, so don't expect them to be able to match them everywhere.

They have trouble matching Intel in manufacturing, which snowballs a bit. But they are still doing extremely well.

For alot of HPC stuff, core-for-core, the "8-year old Athlon architecture" is still a match for the brand new Conroe Intel stuff... so obviously they must have done something right.


RE: News out of AMD
By 91TTZ on 7/26/2007 10:45:01 AM , Rating: 2
quote:
For alot of HPC stuff, core-for-core, the "8-year old Athlon architecture" is still a match for the brand new Conroe Intel stuff... so obviously they must have done something right


The Athlon64 and Barcelona are related to the original Athlon much the same way that the Core2Duo is related to the original P3.

They're not all-new designs, but they're heavily tweaked.


RE: News out of AMD
By Amiga500 on 7/26/2007 5:24:21 PM , Rating: 2
Hence my use of quotes for the "8-year old Athlon architecture" ;-)


RE: News out of AMD
By kamel5547 on 7/31/2007 3:57:14 PM , Rating: 2
Too little to late for me at this point, quite frankly I held out as long as possible in hopes of getting some realistic benchmark data... as it stands our Opteron cluster is getting some new Xeon nodes.

Frankly I don't think AMD has a fighting chance in 1P and 2P boxes at this point, I think their best shot is 4+P (which isn't bad considering margins are higher). Hopefully things change sometime next year to balance things out on both the CPU and GPU fronts.


Is AMD trying to catch up
By FS on 7/25/2007 9:14:29 PM , Rating: 1
or just trying to stay 6-9 months behind Intel forever.

Both are 65nm

AMD 2.4 GHz 95W - Q2 2008
Intel 2.4GHz 95W - Q3 2007




RE: Is AMD trying to catch up
By Viditor on 7/25/07, Rating: -1
RE: Is AMD trying to catch up
By Viditor on 7/25/2007 10:32:19 PM , Rating: 2
My bad...I see where you are going here.
What threw me was that the Opterons in the article are MP, and the Intel Quad cores you spec are only DP (not Tigerton). I don't know if that will change the spec, we will have to see.

I think what's really interesting here is that the AMD chips drop from a TDP of 120w in Q4 to 95w in Q2 08 for the same clockspeed...


RE: Is AMD trying to catch up
By coldpower27 on 7/26/2007 12:41:29 AM , Rating: 2
95W doesn't even exist on Intel Server SKU's besides the Xeon UP lineup.

If were doing a Q4 2007 comparison it is as follows:

2356 2.3GHZ 95W vs E5450 3.0GHZ 80W.

8356 2.3GHZ 95W vs E7340 2.4GHZ 80W.


RE: Is AMD trying to catch up
By Spoelie on 7/26/2007 8:16:43 AM , Rating: 2
The Q6600 is rated at 95W is it not? Not a server chip, true.


RE: Is AMD trying to catch up
By Khato on 7/26/2007 2:53:23 AM , Rating: 2
So, they're counting on process improvements/design changes that may or may not pan out. But, that's nothing different than the rest of the roadmap on display here. Granted, they really -should- have managed to work the kinks out of their 65nm process by now, I mean, they've supposedly been shipping products based upon it for just shy of 8 months now. But you wouldn't notice that if you wanted AMD's 'high end', seeing as how they're still built on the 90nm process.


By coldpower27 on 7/26/2007 12:31:29 AM , Rating: 2
The old B3 Stepping Q6600 is 105W the new G0 drops it down to 95W, not to mention in the new incoming Q6700 at 95W as well.


RE: Is AMD trying to catch up
By defter on 7/26/2007 2:22:00 AM , Rating: 2
quote:
The AMD quad core Phenoms coming out in November are to be 65w at 2.4-2.6 GHz, the 2.8 GHz will be 89w.


What are you smoking? Quad core Phenom is nothing but a Barcelona in a different socket.

Since Barcelona can reach 2.5GHz@120W in Q4, how on the earth can quad core Phenoms reach 2.6GHz@65W at the same time??


RE: Is AMD trying to catch up
By Viditor on 7/26/2007 4:02:59 AM , Rating: 3
quote:
Quad core Phenom is nothing but a Barcelona in a different socket

Um, no...
1. Phenom is HT 3.0 and Barcelona is HT 2.0
2. Phenom has split power planes, Barcelona may not
3. Barcelona must be qualified for servers, Phenom doesn't
4. Barcelona has more HT connections

There are more differences if you look


RE: Is AMD trying to catch up
By Viditor on 7/26/2007 4:17:34 AM , Rating: 2
I forgot to add probably the most important difference...
The memory controllers are quite different between the desktops and the Opterons. The servers require buffered ram that's shared to any other cHT source through the cache and the desktops do not.
That could very well account for a diference in clockspeed...


RE: Is AMD trying to catch up
By defter on 7/26/2007 5:05:55 AM , Rating: 1
Memory controllers are exactly the same. Support for ECC/buffered RAM is just disabled for desktop products.

If you disagree with me, then please tell me the difference at chip level between following chips:

Athlon64 (FX) (Windsor): 2x1MB L2, 227M transistors, 235mm^2@90nm
Opteron (Santa Rosa): 2x1MB L2, 227M transistors, 235mm^2@90nm


RE: Is AMD trying to catch up
By Proteusza on 7/26/2007 4:53:30 AM , Rating: 3
What you describe is the difference between sockets AM2 and AM2+.

AM2 is HT2, and does not have split power planes, while AM2+ has split power planes and HT3. That being said, Barcelona chips can run in AM2 and AM2+ sockets, so, if run in an AM2 sockets, they will function as you described - without HT3 and split power planes.

The new Opetrons and Phenom chips both use Barcelona core technology, which is mostly exactly the same, except for the memory controller to handle buffered and registered memory, and the fact that Opterons (depending on the model) use different sockets and are suitable for MP configurations.

Look it up on Wiki and Anandtech's articles on Barcelona if you dont believe me.


RE: Is AMD trying to catch up
By Viditor on 7/26/2007 6:16:42 AM , Rating: 2
quote:
What you describe is the difference between sockets AM2 and AM2+.

You're thinking desktop (Barcelona is 1207, not AM2)


RE: Is AMD trying to catch up
By Amiga500 on 7/26/2007 7:27:49 AM , Rating: 2
Doesn't matter the exact same difference applies.

The server end has 1207 and 1207+, only the + motherboards can use the split power planes.

Its a chipset thing, not the CPU.

I can't believe they'll have 65W 2.6 GHz components and 120W 2.5 GHz components out at the same time... 95 and 120 maybe (just due to the rating system), but not 65/120.

Power consumption is much more critical for servers & workstations that run 24/7 than desktops. Hence why AMD are producing low powered variants of Barcelona now and going for higher speed later (the low power chips will still sell in big numbers).


RE: Is AMD trying to catch up
By Viditor on 7/26/2007 9:32:00 AM , Rating: 2
quote:
Its a chipset thing, not the CPU

It absolutely has to be both.
Think about it...without a change to the structure of the chip, how could a chipset effect voltages on each individual core?
You couldn't (for example) control the power of individual cores if you used an AM2 chip in an AM2+ mobo, even though the chip will still work within it's limitations.

BTW, individual control in realtime of individual core voltages should drastically reduce both power and heat...


RE: Is AMD trying to catch up
By Proteusza on 7/26/2007 10:11:55 AM , Rating: 2
Yes, it is a CPU and chipset thing, sorry I mistated that in my last post.

But what I meant, is that all K10 class CPUs offer split power planes and HT3. All. Everyone.

Barcelona, Kuma and Agena all share those features.

AM2 and socket 1207 do not, thus any K10 in those sockets will not use them. But, as soon as you put said chip into an AM2+/1207+ slot, it will use them.


RE: Is AMD trying to catch up
By Viditor on 7/27/2007 12:33:18 AM , Rating: 2
quote:
Barcelona, Kuma and Agena all share those features

The Barcelona server chip does not...it's not 1207+ capable.


RE: Is AMD trying to catch up
By Proteusza on 7/26/2007 7:32:27 AM , Rating: 2
Same thing for 1207 and 1207+.

Please, look it up. Honestly, its a chipset thing, its NOT a CPU thing.


RE: Is AMD trying to catch up
By defter on 7/26/2007 4:59:48 AM , Rating: 2
1. Is this a limitation of a core or a socket?
2. Split power planes are function of a new socket, Phenom in AM2 socket will not support split power planes
3. Doesn't make any difference from the hardware point of view
4. Doesn't make any difference from the hardware point of view, current Athlon64 FXs are just Opterons with some HT links disabled

As a result, I still claim that quad core Phenoms will be using exactly the same Barcelona core as Barcelona based Opterons. Only socket will be different and some things maybe disabled in Phenom.

In any case, these things aren't significant enough for account for HUGE difference in power consumption: 2.5GHz@120W vs. 2.6GHz@65W claimed by you.


RE: Is AMD trying to catch up
By Viditor on 7/26/2007 6:28:03 AM , Rating: 2
quote:
1. Is this a limitation of a core or a socket?

Both are required...
quote:
2. Split power planes are function of a new socket, Phenom in AM2 socket will not support split power planes

Again, both are required...but the change is built into the core. It also requires a modified socket to access it (AM2+ or 1207+).
quote:
3. Doesn't make any difference from the hardware point of view

It is a limitation on bin splits and 100% guarantee of backwards compatability.

I think where you guys are confused is that the tech is called Barcelona, and the new server chip is called the same.

The first fully K10 server chip will be Agena (due in Q1), not Barcelona (which is a hybrid that is fully compatable).
They are entirely different steppings and designs...


By coldpower27 on 7/26/2007 8:57:05 AM , Rating: 2
Barcelona is what lots of people are referring when they are talking about K10's as a whole in addition to the Server variants coming in for Q3 2007.

What the heck are you referring to when you say "first fully K10 server chip" you mean the fact that it has HT3.0? Agena won't be able to access ECC/Registered Memory, so it's only going to be used in Opteron UP and Desktops.

Entirely different is a bit of stretch, they are so similar as they are both K10 technology.


RE: Is AMD trying to catch up
By Zurtex on 7/26/2007 7:30:13 AM , Rating: 2
Split power planes have to be built in to the architecture, but they have to be supported by the motherboard as well. You can't just magically add split power planes in to an old CPU by putting it in to a new socket.

I think people are making very poor speculative analysis here, we won't know the real difference until it comes out. But if they are exactly the same with a few things disabled, then they at least won't have any trouble coming out with the Phenom CPUs before Christmas, at least for the lower clock frequencies.


By coldpower27 on 7/26/2007 8:44:36 AM , Rating: 3
That's because thr SKU's Viditor is refering to are Dual Core's not Quad, so 2.8GHZ is projected at 89W.

Anyway these are the TDP of the 2 models as follows Phenom X2 GS-6550 (2.2GHZ-2.6GHZ) and Phenom X2 GP-6800 (2.4GHZ-2.8GHZ).

Both Agena variants will come in at 89W which is the desktop equivalent of 95W.


RE: Is AMD trying to catch up
By MrEMan on 7/26/2007 9:21:47 AM , Rating: 1
I read at the Inquirer ( http://theinquirer.net/default.aspx?article=41202 ) that the delay with getting higher speed 65nm K8s and potentially Barcelonas out has to do with leakage problems with the SOI process.

As I have questioned in the past, it appears that AMD's problems indeed have to do with their process technology collaboration with IBM.


RE: Is AMD trying to catch up
By Khato on 7/26/2007 1:09:14 PM , Rating: 2
Hehe, AMD's 65nm process has me looking forward to their transition to 45nm all the more. If they keep up their current 'progress', then 45nm will max out at 2.2GHz for a year after release =P


RE: Is AMD trying to catch up
By omnicronx on 7/26/07, Rating: -1
RE: Is AMD trying to catch up
By Khato on 7/26/2007 1:57:01 PM , Rating: 2
... May I laugh now? If SOI didn't "mean something for amd" until 45nm, then congrats to them for adding what was it, 10-15% to their manufacturing costs for no reason? No, SOI does give them some marginal benefit at current, and I believe that actual research shows that it'll become economical at around 32nm isn't it?


100 Mhz Increaments?!!
By ralith on 7/25/2007 10:34:53 PM , Rating: 1
Geez, come on 100 MHz increaments between models? This is really disappointing. I thought we stopped having to endure that crap a few years ago. IMO there should be at least a 10% frequency difference between models of like architectures or there really isn't a good reason for the model's existance. I know I can't tell the difference between processors of the same architecture that are clocked less than 10% apart when I'm gaming, compiling, etc.




RE: 100 Mhz Increaments?!!
By CrystalBay on 7/25/2007 10:43:05 PM , Rating: 2
pardon me for ignorance but where is the L3 cache located on the die shot...


RE: 100 Mhz Increaments?!!
By Accord99 on 7/25/2007 11:07:05 PM , Rating: 2
Bottom, the L shaped structures.


RE: 100 Mhz Increaments?!!
By Assimilator87 on 7/25/2007 11:41:27 PM , Rating: 2
Yeah, that's really annoying to me too. AMD's way oversaturating the lineup. Either the price difference will be too minute to make the lower end parts even worth looking at or the price differences will be remain similar to how it's set up now and no one will buy the higher end parts because the extra money isn't justifiable.


RE: 100 Mhz Increaments?!!
By CrystalBay on 7/25/2007 11:53:24 PM , Rating: 2
Or may be the performance is really 40% per clock difference.


RE: 100 Mhz Increaments?!!
By Some1ne on 7/26/2007 2:17:41 PM , Rating: 2
That would be the one saving grace, though it seems very unlikely that the processor would scale like that.

If not, then I agree with the sentiments that the lineup is screwed up and generally haphazard feeling. It's like AMD went "well, performance is just to poor at speeds of 1.8 GHz and less, and we can't get acceptable yeilds at speeds above 2.0 GHz, so in order to keep up the ruse that we're not in dire, dire straits let's release a 1.9 GHz chip and a 2.0 GHz chip and pretend that we're doing it on purpose and not because the whole Barcelona release process has been screwed". I like AMD, but they *sorely* need to get their act together.


AMD shows off 3.0Ghz Barcelona ¿?¿?
By hardwaremister on 7/26/2007 12:02:05 PM , Rating: 2
No benchies AGAIN... but definetely looks nice if it's what it's supossed to be... it'd be very rude if they're playing dirty games.

http://uk.theinquirer.net/?article=41256

let's hope with that little information, they are not really showing a k8 6000+ with a dirty "ES CPUID string"




RE: AMD shows off 3.0Ghz Barcelona ¿?¿?
By hardwaremister on 7/26/2007 12:19:36 PM , Rating: 2
Seems they're telling the truth...

The windows experience index for the athlon 6000+ (formerly 3ghz k8) is 5.4, not 5.9.

So... in order to get this new rating, this must not be a a64!

mmm.... mmm... How the hell have they overcomed such a wide frequency gap? Maybe, latest revision (b2?) was still TOO buggy...

Dunno... <sceptical>


RE: AMD shows off 3.0Ghz Barcelona ¿?¿?
By omnicronx on 7/26/2007 1:02:41 PM , Rating: 1
no its not, its 5.9, my friend has one.. although his score is lower because it takes the worst score you have and thats your experience.. stupid system...

I bet that pic is a fake for sure.. good ol inquirer.. goto love em because they have everything first.. too bad half of the stuff is wrong


RE: AMD shows off 3.0Ghz Barcelona ¿?¿?
By omnicronx on 7/26/2007 1:04:39 PM , Rating: 2
really weird.. i looked it up elsewhere and it says its a 5.4.. does it depend on the motherboard you are using at all i wonder? because i am 100% sure my friends says 5.9 for cpu


By Some1ne on 7/26/2007 6:53:16 PM , Rating: 2
A 3.4 GHz Core2 Duo is sufficient to score a 5.9 on the Windows Vista performance chart for the CPU.


By hardwaremister on 7/26/2007 1:06:07 PM , Rating: 2
http://www.hardocp.com/news.html?news=MjcxMjcsLCxo...

Hardocp is backing the information... maybe it's true at the end!


Sweeeet
By Treckin on 7/25/2007 9:20:41 PM , Rating: 3
I cant wait to see how the native quad-core competes against the 4 core, 2 die design by the boys in blue.

I seriously hope that the 65nm switch was as good for AMD as it was for intel




RE: Sweeeet
By maroon1 on 7/26/2007 8:46:42 AM , Rating: 2
Those processors are supposed to compete with Intel 45nm Xeon because both will be out at the same time


By crystal clear on 7/26/2007 3:04:46 AM , Rating: 2
AMD to Ship Industry’s First Native x86 Quad-Core Processors In August

—AMD’s Stable Infrastructure Strategy Enables Availability of Systems from Platform Partners in September—

Sunnyvale, Calif. -- 29th June 2007 --Continuing to lead the shift of mainstream enterprise computing to energy-efficient processors, AMD (NYSE:AMD) announced today that Quad-Core AMD Opteron™ processors, code-named “Barcelona,” are planned for shipment in both standard and low power versions at launch later this summer. This would be the first time AMD has made both standard and low power parts immediately available as part of a new processor launch.

Additionally, AMD today updated its projected timing on “Barcelona” availability and provided additional product details. AMD expects that the processors will begin shipping for revenue in August 2007, with systems from AMD platform partners beginning to ship in September 2007. Due to its enhanced architecture – it is the world’s first x86 CPU to integrate four processing cores on a single die of silicon – Quad-Core AMD Opteron™ processors can deliver significant performance and performance-per-watt enhancements over existing processor architectures yet are designed to be backwards compatible with existing AMD Opteron platforms.

With planned availability at launch in a range of frequencies up to 2.0 Ghz, AMD expects its native quad-core processors to scale to higher frequencies in Q407 in both standard and SE (Special Edition) versions. Designed to operate within the same thermal envelopes as current generation AMD Opteron processors, AMD estimates that the new processors can provide a performance increase up to 70 percent on certain database applications and up to 40 percent on certain floating point applications, with subsequent higher frequency processors expected to significantly add to this performance advantage.

“More than ever before, customers are expecting energy-efficiency and performance-per-watt leadership as much as absolute performance. With this new reality of computing, greater performance at the expense of greater power consumption is no longer an option,” said Randy Allen, corporate vice president, Server and Workstation Division at AMD. “AMD has prioritized production of our low power and standard power products because our customers and ecosystem demand it, and we firmly believe that the introduction of our native Quad-Core AMD Opteron processor will deliver on the promise of the highest levels of performance-per-watt the industry has ever seen.”


http://www.amd.com/gb-uk/Corporate/VirtualPressRoo...

also-

Dean McCarron, an analyst with Mercury Research, said AMD's use of the same 1,207-pin socket that it uses with its dual-core Opterons will give the company a built-in base of platforms that will enable customers to easily upgrade to Barcelona. That aspect of Barcelona, he said, was overlooked in some of the initial media coverage of the chip, when more emphasis was placed on clock speeds and performance benchmarks.

"They [AMD officials] have a fairly well-developed market of their own with tier-one OEMs, many of which use the dual-core products and the current sockets in their various [product] form factors," McCarron said. "With Barcelona, AMD has made it so users can just drop it into the platforms."

Even with the release of Barcelona, McCarron said the marketplace that AMD is about to enter is entirely different from the one it stormed into in 2003, taking market share from Intel and forcing the chip giant to rethink its companywide strategy. For one thing, Intel's Core microarchitecture is much improved, and the company's intention to move aggressively to its Penryn family of chips later this year shows that its manufacturing plans remain on schedule, McCarron said.

In the coming weeks, AMD customers can expect a familiar rundown of Barcelona innovations—from the chip's Direct Connect Architecture, which allows for improved memory and bandwidth by directly connecting memory to the CPU, to its PowerNow technology, which can increase or reduce the amount of power to the chip depending on the demand.

AMD will also begin talking about its processor's ability to handle 128-bit computing on each core. That will offer better performance for high-intensity scientific applications, Allen said. The chip has also been designed to allow each of the four cores to have its own Level 1 and Level 2 cache.

Allen said AMD also has been working aggressively with companies such as VMware, Microsoft and XenSource to improve the chip's ability to virtualize hardware and help the performance of virtualization software.



http://www.eweek.com/article2/0,1895,2161022,00.as...




By Amiga500 on 7/26/2007 6:20:11 AM , Rating: 2
When can we expect someone to get hold of one to produce some benchmarks?

I need a big HPC workstation, badly, and have been holding off until I see what way Barce shapes up.

If the sample chips hit makers this week, is there any chance of sneaky benchmarks by next week?

:-)




What are they prep'ing exactly?
By Regs on 7/25/2007 10:02:51 PM , Rating: 1
This sounds like the same stuff we heard weeks ago. I don't mean no disrespect, but what do you mean by prep? Taking orders, allocating fab space, finalizing the pricing?




well
By sprockkets on 7/25/2007 11:37:36 PM , Rating: 1
even if it is not worth the money in some respects compared to c2d or just plain athlons, the core is indeed pretty :), i would buy it just for that




"I modded down, down, down, and the flames went higher." -- Sven Olsen

Related Articles













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