backtop


Print 102 comment(s) - last by Tyler 86.. on Jan 30 at 8:34 AM

AMD has high hopes for its upcoming native quad-core design

When it comes to quad-core processors for the desktop and server arenas, Intel has pretty much had the market to itself since November 2006 (if you don't count AMD's Quad FX platform). Intel's quad-core processors were officially announced on November 14 in the form of the desktop-oriented Core 2 Extreme QX6700 and the server-oriented Clovertown Xeon 5300 series.

Intel has been using its quad-core Clovertown processors to grab marketshare back from AMD in the server markets and has been putting significant pricing pressure on AMD. "Early indications are that Clovertown is contributing a meaningful amount of business to Intel in a surprisingly short period of time. It's not marketing fluff," said Mercury Research analyst Dean McCarron.

Technology Business Research's John Spooner added that Clovertown "has allowed Intel to put some pricing pressure on AMD. Intel can tout a lower price per core, given that it's pricing much of the quad-core Xeon 5300 line the same as its dual-core Xeon 5100 chips."

AMD isn't taking this news lightly and is prepared to fight back with its native quad-core Barcelona processors in mid 2007. Intel's quad-core chips put two dual-core chips onto a single package while AMD's approach has one quad-core chip on a single package.

"We expect across a wide variety of workloads for Barcelona to outperform Clovertown by 40 percent," said AMD's corporate vice president for server and workstation products, Randy Allen.

AMD's Barcelona quad-core processors will be built on a 65nm manufacturing process and the company also claims that they will have the same thermal and electrical envelope as existing dual-core Opterons. The processors will also feature 2MB of shared L3 cache as well as AMD Virtualization (AMD-V) technology for x86 virtualization.

Barcelona processors will also fit nicely into existing Socket F systems and will only require a BIOS update for system compatibility.


Comments     Threshold


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

start the hype...
By livelouddiefast on 1/25/2007 4:32:56 PM , Rating: 1
everyone thought core 2 duo and quad was exciting... imagine how it's going to keep rolling in. 40% sounds ridiculous, but so did the preliminary benchmarks for c2d. I've got high hopes for Barcelona, and i can't wait to see if intel will ride out their c2d press (ahem p4) until people get that amd is beating them again. All that IF Barcelona is anywhere near the hype (I will undoubtedly be disappointed, i know)

I wonder if the day will come that we stop being so easily impressed with the amount of cores... 2,4,8...




RE: start the hype...
By Targon on 1/25/07, Rating: -1
RE: start the hype...
By Ringold on 1/25/2007 4:56:30 PM , Rating: 3
Which is in stark contrast to the full fledged Linux distro's and the ever increasing abilities of MacOS, right?

In terms of background services, yes, I remember when I could count on my fingers the number of processes being run by Win95 under full usage. I've currently got 41 buzzing away at the moment under XP. But if you compare the capabilities the two offer it is obviously merely natural evolution.

Linux could use a little more of that 'bloat', in my opinion (and their own opinion, those Linux users wise enough to recognize Linux is not yet a Windows killer for the avg joe).

More OT, though.. Eventually, the gains from adding a new core or two will be marginal.. 8 cores to 10 cores would be noticeable, slightly, but not like 1->2, 2->4, etc. Unless, that is, they keep being able to double it. Then it'll come down really quickly to software not being able to use more than x number of threads no matter how much you work on it.


RE: start the hype...
By Crank the Planet on 1/25/07, Rating: -1
RE: start the hype...
By SoCalBoomer on 1/25/07, Rating: 0
RE: start the hype...
By Pythias on 1/25/2007 10:25:41 PM , Rating: 3
I'm a pythias fanboi. So long as pythias has the best hardware, I don't care who makes it.

Now pardon me for talking about myself in 3rd person like a jackass.


RE: start the hype...
By Araemo on 1/26/2007 9:21:05 AM , Rating: 2
I like the way you think, and that's about the truth of it, though I don't buy the 'absolute best', I buy the absolute best that I can afford. ;)

That is often rather different than the best out there, since I'm not paying $800 for a CPU, that's about the price of my full computer(Minus case, monitor, hard drives, keyboard, mouse). Whoever can sell me the highest end computer for that amount of money is who will get the money from me.


RE: start the hype...
By stromgald on 1/25/2007 8:12:48 PM , Rating: 2
You sound like an AMD fanboy. Especially with all the 'wait and see' crap and trashing Intel for how long they worked on C2D. Actually, I don't think they worked on C2D for that long since they stuck it out with netburst so long that they had to come out with BTX.

It really doesn't matter how long they worked on it, all that matters is whether they put out better chips than AMD. That's the ultimate motivation for buying from either company.

Comparing a 3Ghz P4 to a AMD X2 4400 is like comparing a Core 2 Duo E6600 with an AMD FX-57. It's a stupid comparison because of the cost difference and the age difference of the chips.

The AMD/ATI merger may help AMD in the integrated graphics chipset market, which they've never had a presense in. But, Intel has been there for quite a while. The main problem with the merger is that ATI's discrete graphics cards are dropping like crazy. One of their major income sources (building integrated graphics chips for Intel motherboards) basically got eliminated in the merger.

Also, your statement that 'Intel' is scrambling for GPU experience because of Fusion is stupid. Intel is more likely thinking of entering the discrete GPU market because of the void created by ATI dropping off. Right now nVidia almost stands alone at the top of the graphics card market. Although the R600 series coming from AMD/ATI will help, I expect ATI to become a much smaller player in the video card market or bow out in the next few generations of graphics cards. Intel is looking to take advantage of that and possibly to protect themselves against AMD's Fusion.


RE: start the hype...
By edge929 on 1/26/2007 3:00:31 PM , Rating: 2
"Comparing a 3Ghz P4 to a AMD X2 4400 is like comparing a Core 2 Duo E6600 with an AMD FX-57. It's a stupid comparison because of the cost difference and the age difference of the chips."

In my experience of building hundreds of computers over the last 10 years, the average Joe looks at gigahertz speed over price and translates that into performance. So they automatically think that 3Ghz P4 will outperform everything from AMD 2.9Ghz and lower. You and I both know that's far from the truth but thanks to Intel's BS marketing, spamming that speed is king, everyone thinks the P4s are faster. You can thank Intel for AMD's confusing CPU naming schemes.

The AMD/ATi merger will do nothing but boost ATi's already impressive GPUs history. I'm loyal to nethier ATi or nVidia (own SLi mobo) but if you're calling for the demise of ATi, you'd be crazy. With no competition, do you really expect nVidia to "push the bar" on GPUs? Same goes for ATi. I hope a third party (Intel?) comes in to make mainstream and high-end GPUs from scratch. We all benefit from it.

From what I know, ATi was never a huge player in integrated graphics, that's Intel's arena all the way. Sure they had some, as did everybody, but nothing compared to Intel's sheer numbers.

I get a hard-on everytime I think about AMD/ATi's next generation of CPUs and GPUs. It's only logical to assume the next CPU from AMD will trump C2D (leap frog game) and will have ATi-only GPU speed enhancements to boot. Fusion, physics processors, whatever, it's all good.


RE: start the hype...
By StevoLincolnite on 1/26/07, Rating: 0
RE: start the hype...
By encryptkeeper on 1/26/07, Rating: -1
RE: start the hype...
By Adonlude on 1/26/07, Rating: -1
RE: start the hype...
By RedBeard on 1/26/2007 8:36:31 PM , Rating: 2
Nice parody post, your humor was not lost on me.

To the rest of you people who bought this flamebait need to slap yourselves, gimme a break.


RE: start the hype...
By piroroadkill on 1/26/2007 9:34:15 PM , Rating: 2
With 2mb of RAM it would perform worse than any 286 I can imagine.

That's 2 millibits, which is 2 thousands of a bit, or 0.25 thousands of a byte!

0.00025 bytes of RAM would be pretty lame, dude.


RE: start the hype...
By SexyK on 1/25/2007 4:56:28 PM , Rating: 1
quote:
everyone thought core 2 duo and quad was exciting... imagine how it's going to keep rolling in. 40% sounds ridiculous, but so did the preliminary benchmarks for c2d. I've got high hopes for Barcelona, and i can't wait to see if intel will ride out their c2d press (ahem p4) until people get that amd is beating them again. All that IF Barcelona is anywhere near the hype (I will undoubtedly be disappointed, i know)


Ah ha - but you've stated the difference between the C2D hype and Barcelona hype yourself in your post. We had C2D benchmarks to get excited about way before launch, while at the moment this "40% better" number is simply pulled out of the air by AMD. Sounds great and all, but I'm not going to bite until there are some cold hard numbers out there.


RE: start the hype...
By ScythedBlade on 1/25/2007 5:41:31 PM , Rating: 1
It's weird too. Remember when AMD fanboys STILL didn't believe any numbers when systerms were out? In fact, according to anand ... i think, intel was even underboasting its processor.


RE: start the hype...
By Ringold on 1/25/2007 5:48:53 PM , Rating: 3
I admitted AMD defeat when they ran a benchmark, I think at IDC.. I can't remember the details, but it was a common task. The AMD system they were using was balls-out OC'd, and the C2D system destroyed it. No point in debating it then. The margins were so astounding that even if they were doing something to introduce a little bias it wouldn't matter. Half the margin of victory still would've been crushing.

I think I'll take their 40% claim and divide by 2 for some benchmarks, and divide by 4 for more common tasks. That way maybe I'll even be surprised! :)


RE: start the hype...
By SacredFist on 1/26/2007 9:01:21 AM , Rating: 1
wow ... downgraded for a correct post?

Anyway, yea, many people cried "LIES" when benchmarks came out.


RE: start the hype...
By djcameron on 1/25/2007 6:08:05 PM , Rating: 1
Just think...someday we'll be measuring our chips in GigaCores instead of GigaHertz


RE: start the hype...
By masher2 (blog) on 1/25/07, Rating: -1
RE: start the hype...
By Kilim on 1/25/2007 11:15:09 PM , Rating: 2
Um...? No disrespect meant, but why would that have been a bold prediction even ten years ago?


RE: start the hype...
By KristopherKubicki (blog) on 1/26/2007 12:31:30 AM , Rating: 4
Judging by how much he got flamed, I would say it was pretty bold.


RE: start the hype...
By Tyler 86 on 1/30/2007 8:34:36 AM , Rating: 2
I still think it's loony toons. :)


RE: start the hype...
By IntelUser2000 on 1/25/2007 6:13:21 PM , Rating: 3
quote:
imagine how it's going to keep rolling in. 40% sounds ridiculous, but so did the preliminary benchmarks for c2d.


40% OVER NETBURST. Take a look at IDF presentations from Anandtech. Most of you have a 24 hour memory.


RE: start the hype...
By goz314 on 1/25/2007 6:35:57 PM , Rating: 2
Well, I guess it's about time that Intel starts pitching the 45nm Wolfdale DP and Harpertown processors to compete with Barcelona.


RE: start the hype...
By shabby on 1/25/2007 7:46:47 PM , Rating: 3
At least the c2d hype started from benchmarks, with amd its from hopes.


RE: start the hype...
By primer on 1/25/2007 8:57:40 PM , Rating: 2
quote:
At least the c2d hype started from benchmarks, with amd its from hopes.


AMD is making some major changes to the IPC with Rev H Barcelona and Agena/Kuma desktop parts. I believe this has been touched on before either on Anandtech or one of the other sites.

Changes to the core are: 32B instruction fetch, enhanced branch prediction, out-of-order load execution, 4 double precision FLOPS/cycle, dual 128-bit SSE dataflow, dual 128-bit loads per cycle, bit manipulation extensions, additional SSE extensions and enhanced thermal monitoring management. The inclusion of an L3 cache will help feed the cores and help decrease the load on the memory controller. There will also be a platform enhancement in that will further increase performance that is most likely included in the "40%" reference (not refering to HT 3.0).

So I think that AMD isn't talking so much hype. It will be interested to see in Q2 and Q3 as to how these chips (server and desktop) will perform.


RE: start the hype...
By shabby on 1/25/2007 10:31:21 PM , Rating: 1
Will those changes really increase the performance of the chip that much?
Remember amd's chip is slower then the c2d, first it has to match c2d's performance and then be 40% faster on top of that. Thats like a 60% increase in performance, its not beliveable.


RE: start the hype...
By Viditor on 1/25/2007 11:04:16 PM , Rating: 3
quote:
Thats like a 60% increase in performance

Actually, they are saying that FP will be an 80% increase in performance over current Opterons...
Looking at the new design, it really is fairly believable.

Going forward, I imagine that AMD will regain the crown across the board for awhile until Penryn allows Intel to increase clockspeeds enough to catch up.
After that, AMD will be going to 45nm in mid 08, followed by Nehalem (with CSI) at the end of 08.
It's going to be a regular ping-pong match for the next few years, and the performances will get closer and closer to each other...


RE: start the hype...
By Khato on 1/26/2007 2:40:51 AM , Rating: 2
Meh, AMD is also saying that they're going to get their gross margin up to at least 50% for the year. It's surprisingly easy for them to say such things and give no actual information about how they intend to execute upon what's said.

AMD just wants to make unfounded claims that they're certain they can meet in some obscure benchmark when the time comes. Doing such gives the investors and their server customers hope that they'll turn things around soon. Heh, so long as they actually have Barcelona running more than task manager by now, they still wouldn't dare show actual benchmarks for fear of what feather Intel would then pull out of the hat.


RE: start the hype...
By Viditor on 1/26/2007 2:46:31 AM , Rating: 2
quote:
AMD is also saying that they're going to get their gross margin up to at least 50% for the year

Not really...they say that they are going to target getting their Gross Margin to 50% in 2007...that's very different.
It's not a 50% total, it's a 50% target for a quarter (which should be quite easy for them...).

The Devil is in the details...


RE: start the hype...
By Shintai on 1/26/2007 5:02:44 AM , Rating: 2
The numbers AMD posted are pretty messed up. A C2D is about 20% faster than K8. On the serverside alittle less. But for AMDs numbers to match with K8 and Clovertown. It needs to be over 30% faster.

And looking on the improvement. This looks like its some SSE case only. It would be like saying Core 2 is 600% faster than Core in FP.

So unless AMD tells us what benchmark, what numbers (So we can test outside and validate) then this information is 100% useless and look like Apple/Sun marketing with bending of the truth and reality.


RE: start the hype...
By Shintai on 1/26/2007 5:06:06 AM , Rating: 2
Forgot to add:

If AMD got anything to brag about. They would be posting real numbers by now. And not some simulated expectation numbers like they base everything on. AMD is very very quiet about it, and it sounds like its a big ballon of hot air. Hell, we cant even get any Clovertown model numbers. So the newssites assumed it was a QX6700 Kentsfield. Sounds more like the 2eggs that became 5 chickens under its travel.


RE: start the hype...
By Viditor on 1/26/2007 9:27:54 AM , Rating: 3
quote:
A C2D is about 20% faster than K8. On the serverside alittle less. But for AMDs numbers to match with K8 and Clovertown. It needs to be over 30% faster


I think you are forgetting that even current Opterons scale much better than Woodcrest...this is mainly due to FSB vs HT.
The reason that this is important is that as we increase cores and sockets (like for Barcelona and Cloverton), AMD gains in relative performance because of the platform.
Therefore, even with current Opterons the comparitive performance to Woodcrest is far closer than with C2D vs X2.

Compared to Opteron, Barcelona (K10):
Has doubled the internal bandwidth for both loads and SSE...
Instruction fetch increased from 16B to 32B
Added out-of-order loads (VERY important!)
Added more HT connections and made them faster
etc...

As to what benchmark, obviously (as usual) the ones that are more FP heavy will be heavily in AMD's favour...


RE: start the hype...
By Phynaz on 1/26/2007 9:36:52 AM , Rating: 1
quote:
I think you are forgetting that even current Opterons scale much better than Woodcrest.


Proof?



RE: start the hype...
By Viditor on 1/26/2007 9:29:43 PM , Rating: 2
quote:
Proof?


"Here is a first indication that quad core Xeon does not scale as well as the other systems. Two 2.4GHz Opteron 880 processors are as fast as one Xeon 5345, but four Opterons outperform the dual quad core Xeon by 16%. In other words, the quad Opteron system scales 31% better than the Xeon system"
http://tinyurl.com/2cgnj8


RE: start the hype...
By Khato on 1/26/2007 12:21:22 PM , Rating: 2
Mmmm, maybe I'll have to listen to their earnings call again, as I'm quite certain they wished to claim 50% cumulative for the year. Heh, when even getting 50% for one quarter is going to be quite difficult indeed for them.


RE: start the hype...
By Viditor on 1/26/2007 9:46:02 PM , Rating: 2
quote:
even getting 50% for one quarter is going to be quite difficult indeed for them

I won't disagree...
It's going to be tough for both AMD and Intel, depending on how long the price war goes on.
However, Barcelona and the Stars CPUs give them a helluva lot more ammunition for pricing in H2. I suspect that Q4 (and possibly Q3) will be either very close or even exceed 50% GM. Keep in mind also that their costs are dropping all through this year (conversion to 300mm and 65nm is expected to be complete by mid-year).


RE: start the hype...
By Targon on 1/26/2007 5:22:16 AM , Rating: 3
A key issue here is the performance of a K8L core when compared to a "Core" core. That's the thing that you and many other people doubt. From reading the various reports on K8L, there is a huge 80% difference in the design compared to the current K8 design. So really, just about the entire CPU will be changed. It's not just two or four K8 cores tied together with some L3 cache, but some massive re-designs of the cores themselves.

If you compare what Intel did when they went from the Pentium-M to Core, the performance difference is huge. If you can accept that, then why is it so hard to accept that AMD may have been able to pull off a huge change and performance boost with the move to K8L? I'm not saying for certain just how large the performance improvements are or will be by the time the new processors are launched, but the potential is there.

Now, if AMD can get the performance up so they are even in terms of performance compared to Intel when you compare core to core performance, then the overall system architecture from AMD may very well allow AMD to regain the performance crown across the board.

That's really what many people don't seem to take into account. With Intel, it's CPU performance while using the same ancient system architecture. If Intel doesn't have an advantage in terms of CPU performance, then AMD will rocket past them in overall system performance the way the Athlon 64 did to the Pentium 4.

So, doubt the numbers that AMD put forward for now, but don't doubt that K8L may very well allow them to regain the performance crown.


No claims have actually been made
By HaZaRd2K6 on 1/25/2007 4:32:17 PM , Rating: 5
quote:
...We expect ... Barcelona to outperform Clovertown by 40 percent...


To me, that doesn't mean AMD is claiming that Barcelona will be faster, it means they're hoping it will. Until they're close to releasing it, nobody can really say whether or not it'll outperform Clovertown.




RE: No claims have actually been made
By hstewarth on 1/25/2007 4:53:13 PM , Rating: 2
I believe by next summer Intel will have its native quad core out by then. Plus there is a lot to room to spare in the Core 2 duo and Core 2 quad base processor. Sandra on my Dual Xeon 5160 indicated that CPU allows up to 5Ghz. But I doubt that Core 2 line will go up that far - because of new line of CPU's coming out.

In general by time that AMD comes out with there new cpus, it hard to indicate performance against Intel current line. Core 2 line is not like the Netburst line which was last year at its end of life time, the new Intel architexture is at beginning of its life cycle. Who knows maybe even by end of year, Intel will have 8 core ( Clovertown style ) out there.

But both Intel and AMD pushing technology forward is a good thing, because this means better performance and lower prices for consumers.


RE: No claims have actually been made
By hstewarth on 1/25/2007 5:00:35 PM , Rating: 2
The following link is explains the new 45nm Pemryn processors - which are expected 2H of 2007. These are native quad cores mention above.

http://hwzone.co.il/english-reviews/intel_apc_2006...


By politicalslug on 1/25/2007 7:39:54 PM , Rating: 2
The article you mentioned does say anything about Intel producing native quad-core processors. In fact neither does their roadmap. Not specifically anyway. Penryn is nothing more than a manufacturing shrink (65nm to 45nm) of the current CORE architecture. Native quad-core is like their next architecture, but that won't be on the market until mid to late 2008 at the earliest.


RE: No claims have actually been made
By politicalslug on 1/25/2007 7:43:47 PM , Rating: 2
The article you mentioned doesn't say anything about Intel producing native quad-core processors. In fact neither does their roadmap. Not specifically anyway. Penryn is nothing more than a manufacturing shrink (65nm to 45nm) of the current CORE architecture. Native quad-core is likely to be incorporated in their next architecture, but that won't be on the market until mid to late 2008 at the earliest.


By hstewarth on 1/26/2007 11:11:01 AM , Rating: 2
I am getting from the following article that Native quads from Intel are coming shortly after the Penryn

http://www.tomshardware.com/2005/12/04/top_secret_...

But Intels process of improvements are rapitly changing since the Core 2 release and dismiss of the Netburst logic.

The important thing to remember here, by the time these AMD processors are release, Intel is going to have a completely new breed of processors out with with die shrink which means more cache and likely faster components.

Right now until the cpus are actually all of this including perfromance measure of future cpus - is just speculation.


RE: No claims have actually been made
By Marlowe on 1/25/2007 7:56:00 PM , Rating: 2
Hmm.. I think Penryn is the 45 nm shrink of Clovertown, still with seperate dies.. Wolfdale is the 45 nm shrink of conroe..

:)


By coldpower27 on 1/26/2007 12:24:19 AM , Rating: 2
Not quite, Penryn refers the shrink of Merom them mobile core, and like Merom, the base name of the 45nm Core 2 architecture line, where Merom was the base name for the 65nm Core 2's.

The shrink of Clovertown is called Harpertown and is still a MCM module approach.

Though the native vs MCM debate is only 1 of many factors to consider.


By Trisped on 1/25/2007 5:37:59 PM , Rating: 2
They are not even hoping that it is faster, just outperforming, which could mean lower power use to do the same amount of work.

With the estimate so high, I believe that we are being assured it will outperform the Clovertown, but by how much and in what ways remains to be seen. I expect that since die reductions will soon be here for Intel, AMD is putting a lot of work into getting the most out of their processor so AMD can compete in the coming years.


By Viditor on 1/25/2007 10:40:10 PM , Rating: 2
Obviously AMD can't know for sure how well Cloverton will do...
But they do say that Barcelona will be 80% faster in FP than current Opterons on a core for core basis at the same clockspeed. That should make Barcelona massively better in FP than Cloverton, but I imagine that integer performance will be much closer (maybe 10% faster).


By verndewd on 1/28/2007 10:49:47 PM , Rating: 2
isnt that the same thing as intel expecting netburst to scale to 10ghz?We have been hearing alot from AMD about expecting this and that,lol,I hope to expect some more from them in the future.
It cannot be easy competing against such brilliant minds at intel,with so many resources and think tanks.what makes it hard on AMD is everyone knowing about the IDF pamphlet;Its that kind of activity that brings the skeptic out in people.


Opteron
By cochy on 1/25/2007 4:27:46 PM , Rating: 1
Just as I mentioned in a previous article. This is just more news out of AMD which confirms in my mind their changing focus from the desktop market to the server and integrated graphics/chipset (ATI) market. Opteron is their bread and butter. Makes perfect sense for them to come and defend their market share this way. Plus I don't doubt their optimism isn't out of place, the Opteron platform is a real good server platform.

As for them losing focus on the desktop market, well all we gotta do is look at QuadFX. Ugh.

Anyway who knows. The future looks interesting, that's the one thing that is for sure.




RE: Opteron
By Targon on 1/25/2007 4:38:02 PM , Rating: 5
Quad-core is the direction things are moving in, the same way the move from one core to two was the direction things were moving in only a year or so ago.

AMD always has a focus on BOTH the server and consumer markets. The big thing that needs to be brought up with this announcement is that when AMD launches their quad-core processors, it's not "just" the current dual-core chips extended to be native quad-core. Each core will have a new design called the K8L. This is the thing that will make AMD more competitive.

Now, the K8L design improvements are NOT limited to quad-core chips, but will also be used in dual-core processors released at around the same time. If the K8L design is as good as the information that has been released(and leaked), then those new dual-core chips will make it where the new AMD chips will compete very well with the Core 2 Duo processors available at that time.

Yes, AMD is around five months early in talking about this stuff, but Intel was also very early in talking about the Core marchitecture back when the Athlon 64 was making the Pentium 4 look like a nice frying pan rather than real competition.


RE: Opteron
By Regs on 1/25/2007 7:56:11 PM , Rating: 2
Hey good comment. Please though, enough people call this the K8L all ready thinking its a rebranded K8. Hah.

Leave it at barcelona.


RE: Opteron
By Tsuwamono on 1/25/2007 9:35:57 PM , Rating: 2
Its actually K10 really. K8L was what "The Inquirer" in all their knowing decided to call it. K10 is what AMD calls it


RE: Opteron
By cochy on 1/25/2007 10:29:40 PM , Rating: 2
What was K9? Did they skip 9?


RE: Opteron
By Viditor on 1/25/2007 11:06:10 PM , Rating: 3
Yes, they skipped K9 (it was a dog anyway)...:)


RE: Opteron
By KristopherKubicki (blog) on 1/26/2007 12:33:29 AM , Rating: 2
K8L was actually dubbed by AMD's Henri Richard:

http://www.dailytech.com/article.aspx?newsid=1220


RE: Opteron
By Spoelie on 1/26/2007 8:34:55 AM , Rating: 2
No, K10 was scrapped as well. K8L was not termed by the inquirer but by intel, theinq just made it public thinking it was AMD's codename.

Especially:
http://uk.theinquirer.net/?article=33906


RE: Opteron
By Regs on 1/25/2007 4:43:25 PM , Rating: 3
What's the difference between a desk top processor and a workstation processor these days? Server however is more diverse depending on the set up.

Not only is AMD's bread for money product is the Opteron, but the server segment is their strength. They have to capitalize on opportunities as much as possible.


RE: Opteron
By koglo on 1/25/07, Rating: -1
RE: Opteron
By koglo on 1/25/07, Rating: -1
RE: Opteron
By Goty on 1/25/2007 6:19:04 PM , Rating: 1
AMD has a history of releasing new architectures and revisions on server platforms first. Just because the first quad-core chip out of the foundry is an Opteron doesn't mean a thing. If you'll recall, the Opteron was out a number of months before the A64.


Where's the rest of the AMD new?
By Phynaz on 1/25/2007 4:44:44 PM , Rating: 2
Like losing half a billion bucks in one quarter? Gross margins down to 36%




RE: Where's the rest of the AMD new?
By hstewarth on 1/25/2007 4:55:56 PM , Rating: 2
I think the stock results are reasons for these statements. They know they hurting because of Core 2 and hoping to keep customers and more importantly investors still interested in the company.

Purchasing ATI was likely shown as bad decision - I feel this would isolate customrs - limits there choices.


RE: Where's the rest of the AMD new?
By Targon on 1/26/2007 5:01:18 AM , Rating: 2
Pure investments to get technology are almost always misunderstood by investors, unless it's a license agreement for technology. The problem with most investors is that they really don't understand the products produced by tech companies, and they also don't understand innovation in a segment.

AMD has shown that they have been, and continue to look for ways to improve overall system performance on an architectural level. AMD started with the very obvious addition of a memory controller to the CPU. That's a huge change from the overall system design you see used by Intel. Now, what performance benefits would be possible if you make a video card that has a direct connection to the CPU?

That's not quite the same question as the benefits of an integrated GPU on the same chip as the CPU. Put a video card into a HTX(HyperTransport) slot, and if the slot is tied directly to the CPU, you have something similar to a graphics co-processor, but as a card with it's own memory. Because the GPU and CPU are now talking together faster than having to go through a PCI Express slot, the possible performance benefits where graphics performance is concerned is huge.

Intel may be looking at how to take advantage of the AMD/ATI merger, but AMD may be and probably is looking at how to improve how system components work together for improved performance.

I bet that the investors don't understand this sort of approach, and figure the CPU market doesn't change much, in the same way that things don't change much in the automotive industry where you only get a significant change every 5-7 years from a car company.


RE: Where's the rest of the AMD new?
By smitty3268 on 1/25/2007 5:07:43 PM , Rating: 1
Actually, I believe AMD made a profit this quarter. Unless you add in the cost of buying ATI.


RE: Where's the rest of the AMD new?
By hstewarth on 1/25/2007 5:20:17 PM , Rating: 2
ATI was major part of the loss revenue but not all of was from ATI

http://money.cnn.com/2007/01/23/technology/amd_ear...



RE: Where's the rest of the AMD new?
By smitty3268 on 1/25/2007 5:24:42 PM , Rating: 2
So basically, they made .01 per share in profit, then lost .05 per share to employee stock-based compensation, then lost 1.04 per share by buying ATI. So no, it wasn't all of it but it was the vast majority.


RE: Where's the rest of the AMD new?
By hstewarth on 1/26/2007 6:16:02 PM , Rating: 2
What is most important in the recent stock actions. why did AMD sink to its lowest levels in about 2 years. It may not be related to ATI - maybe the fact that Netburst days are over with Core 2 technology. Or something else completely.


By Viditor on 1/26/2007 10:01:48 PM , Rating: 2
quote:
What is most important in the recent stock actions

I would say that the reason they sank so low was that they completely missed their predictions from just last month at the Analysts Day...
Intel's numbers were pretty dismal as well, but that was already priced into the stock.
If you look at the charts of AMD and INTC, you'll see that both had a dip down to the $16 range in July.
Also, while C2D had a huge negative effect on AMD over the year, it hasn't increased Intel's price at all (in fact Intel is below where it was a year ago).

BTW, Netburst is on it's way out, but it's not over...
Only 40% of Intel's shipments are C2D.


RE: Where's the rest of the AMD new?
By Ringold on 1/25/2007 5:21:13 PM , Rating: 2
Half a billion? Their revenue was 1.85 billion, EPS of .22, up 33% from the previous year, with $587 million profit BEFORE ATI acquisition costs. Including those costs, a loss of 47 million. Gross margins also reported as 40%, not 36%. That's down from 57% just a couple years ago, though.

With acquisition costs over and integration costs lowering AMD will return to profitability in its quarterly reports immediately. Though in practice, profitability never actually stopped.


By Ringold on 1/25/2007 5:23:38 PM , Rating: 2
Woops -- I was looking at non-GAAP margins. You're right, margins were 36.


40% seems very high
By lazyinjin on 1/25/2007 5:26:29 PM , Rating: 2
I bet this 40% increases is in performance-per-watt(the native quad vs. double dual-core approach will lower power consumption and along with performance increases due to core changes), which is a good thing but certainly not as as impressive a number as it would seem. But I hope I'm wrong--- I would like to jump from single to quad rather than take an intermediary step at dual.




RE: 40% seems very high
By Lonyo on 1/25/2007 8:54:54 PM , Rating: 2
Nah, it's all very real.
Level 3 cache + up to 40% in some circumstances basically means they've found apps which will overflow the regular cache (of both processors), but the handy extra L3 cache in Barcelona allows a 40% improvements in performance, because they can fit the necessary in there, whereas Intel can't, and need to fetch from the main memory. :P


RE: 40% seems very high
By lazyinjin on 1/25/2007 10:17:23 PM , Rating: 2
you are probably true, but i was just being pessimistic especially with all the hype about four cores lately. i would really like to see AMD keep things close, though, to keep Intel on their toes and my wallet happy!


RE: 40% seems very high
By Schmeh on 1/25/2007 11:39:43 PM , Rating: 2
Problem with the L3 cache being the reason for the 40% is that if you add up all cache of a Intel's Cloverton you get 8MB of cache. Granted it is split in two between the two die, so we'll say there is 4MB of cache. Barcelona has 4MB of cache as well. Each core has a dedicated 512KB L2 (for a total of 2MB) and there is the shared 2MB L3 cache. So essentially they have the same amount of cache.


RE: 40% seems very high
By IntelUser2000 on 1/26/2007 1:41:19 AM , Rating: 3
quote:
Nah, it's all very real.
Level 3 cache + up to 40% in some circumstances basically means they've found apps which will overflow the regular cache (of both processors), but the handy extra L3 cache in Barcelona allows a 40% improvements in performance, because they can fit the necessary in there, whereas Intel can't, and need to fetch from the main memory. :P

Not really. They have been specific.

http://www.behardware.com/news/8484/amd-to-demonst...

"About performances, AMD announces respectively 13% and 46% improvements with TCP OLTP and SPECfp compared to a Xeon 5355"

I assume they meant to say TPC not TCP.

K8L vs. Clovertown. In SpecFP, it has 46% performance improvement over Clovertown, and SpecFP(they likely mean SpecFP_Rate as they had similar graphs comparing K8L and Clovertown in SpecFP_Rate with similar 40% improvement) loves bandwidth. K8L has huge bandwidth advantage, get it now?? Clovertown performs only 31% faster in SpecFP_Rate than Woodcrest despite twice the amount of cores:
http://www.spec.org/cpu2000/results/rfp2000.html

Xeon 5150 2.66GHz 4 core/2 CPU: http://www.spec.org/cpu2000/results/res2006q3/cpu2...

77.1

Xeon X5355 2.66GHz 8 core/2 CPU:
http://www.spec.org/cpu2000/results/res2006q4/cpu2...

101

In comparison, 2 core/1 CPU Xeon 5150 gets 42.8, which means going from 1 CPU to 2 CPU gives 80% performance increase.

Looking at that Clovertown should get 138.8, not 101, but due to having same bandwidth and twice the number of cores, Clovertown has poor scaling.


RE: 40% seems very high
By hstewarth on 1/26/2007 6:08:35 PM , Rating: 2
One of the problems with these results is that they are not all on the same compiler and OS. Different compilers cause different results or possible different OS. A lot on done Windows Server 2003 which may not scale as good as other OS.

Since we are talking about a non-release cpu, it hard to verify that 40% is a true factor. Without know what they are comparing against and how - it hard to prove anything.


Mid 07?
By gramboh on 1/25/2007 6:27:13 PM , Rating: 2
What the hell does mid-2007 mean? We are almost into February and no firm dates or benchmarks? Come on, get serious AMD. Won't Peryn be out by "mid 2007"?




RE: Mid 07?
By Viditor on 1/25/2007 10:30:16 PM , Rating: 2
quote:
What the hell does mid-2007 mean?

AMD has stated that shipping begins in Q2 near mid year. The word is that launch is slated for week 2 of June, though they could push it sooner if needed. Production has reportedly already begun...
quote:
Won't Peryn be out by "mid 2007"?

No, it begins production then...release is for the end of 2007.
According to Intel's slides from the most recent IDF, they are expecting availability in Q1 08, but they should be able to push that forward at least a month...
http://techreport.com/etc/2006q4/fall-idf/index.x?...


RE: Mid 07?
By Gatt on 1/26/2007 12:41:05 AM , Rating: 2
That's an extremely confusing report, as one thing indicates 2008 for Penryn, and just a little further down it says...

quote:
Paul Otellini revealed more detail in his opening keynote last week, claiming that the first 45nm microprocessor design (presumably Penryn) will be completed next quarter, with 45nm products scheduled to ship in the second half of 2007.


Very confusing.


RE: Mid 07?
By Viditor on 1/26/2007 9:33:50 AM , Rating: 2
It's actually not as confusing as it seems...
If Intel ships in Dec 07 (second half of 2007), then it won't be available till Q1 08. Remember that there is about a 1-2 month lag between shipping and availability.
That said, I personally expect that Intel will be able to move that up to a Dec availability if they think it's strategically a good thing to do.


RE: Mid 07?
By hstewarth on 1/26/2007 11:16:39 AM , Rating: 2
Please keep in mind, that release actually come early then some believe especially now that Netburst days are over and Core 2 production. I personally thought that Intel quad cores came earlier than expected.

Ways things went 2H of last year, I would not doubt that we will be talking 8 core cpus a year from now.


40%
By casket on 1/25/2007 11:51:02 PM , Rating: 2
I guess 40% is the projection. Intel is not sitting still though. I imagine in 6 months... intel's chip will be an extra 40% than today's quad-core. Clock-speeds don't stand still. AMD is not saying they will have a 40% lead in 6-months from now.




RE: 40%
By Viditor on 1/26/2007 12:21:09 AM , Rating: 2
quote:
AMD is not saying they will have a 40% lead in 6-months from now

Actually, that's exactly what they're saying...
Intel isn't sitting still, but they don't have any real upgrades scheduled for another year (45nm).


RE: 40%
By Targon on 1/26/2007 5:05:43 AM , Rating: 2
The key is architecture changes, not just a die shrink. If Intel only has a die shrink with a few tweaks for another year or two, then AMD has a pretty good chance to catch up. It's been proven time and time again that just increasing the clock speed of a chip does NOT help if your CPU architecture is inferior(the problem AMD has when competing with the Core marchitecture these days).

So, the move to 45nm will help Intel, but if AMD has a better CPU design with the K8L, then the die shrink will only help to a limited extent. Knowing Intel, they will put 1 gig of cache memory on their processors if they find that they are being beaten in terms of performance, at least until they get a new CPU design ready.


RE: 40%
By Khato on 1/26/2007 12:44:21 PM , Rating: 2
Psssst, here's a little scenario for you. Ever wonder how Intel pulled in the Conroe release quite so much as they did? Sure it was always scheduled for 2H '06, but it launched some 3-6 months ahead of the actual date. Here's a hint: with major architectural changes, it's usually performance issues that are the last to get worked out. If a product works well enough, you can forego the performance steppings and move up the release date and then have plenty of time to tweak the design for a future revision.

Those 1333 MHz FSB Conroes are gonna be purrdy.


RE: 40%
By Viditor on 1/26/2007 10:19:33 PM , Rating: 2
quote:
Ever wonder how Intel pulled in the Conroe release quite so much as they did?

They announced how they did it...
They shelved almost half of their working projects and put all of their teams on rushing Core and Core2 out the door.
Whitefield with CSI was originally supposed to ship in 2006...

quote:
If a product works well enough, you can forego the performance steppings and move up the release date and then have plenty of time to tweak the design for a future revision

There is certainly quite a bit of truth in this...
I think that this was the original plan for Pressler, but they were able to get C2D out quicker than estimated so they dropped the major refreshes they had planned.


Deja Vu All Over Again
By Dactyl on 1/25/2007 5:57:03 PM , Rating: 2
I remember when AMD had the performance crown, Intel's chips were priced down to be competitive, and Intel was bragging about a new architecture that would outperform AMD by about 40%.

That wasn't so long ago! Maybe AMD can pull an Intel and get back into the lead, at least for the server and high-end gaming markets.




RE: Deja Vu All Over Again
By IntelUser2000 on 1/25/2007 6:12:27 PM , Rating: 3
quote:
I remember when AMD had the performance crown, Intel's chips were priced down to be competitive, and Intel was bragging about a new architecture that would outperform AMD by about 40%.


40% over NETBURST.


RE: Deja Vu All Over Again
By Dactyl on 1/27/2007 12:58:20 AM , Rating: 3
Intel claimed Woodcrest would beat Opteron by 33%[1]. There isn't that much difference between Opteron and Athlon or between Woodcrest and Conroe. I certainly remember people debating whether or not Conroes would be significantly better than Athlons, or that Intel was just making stuff up.

It's true that Intel claimed that Core 2 had 40% better performance and 40% less power usage than Netburst[2]. Together, that's more than 40% better. That would be about twice as good (1.4 x 1.4 == 1.96). I was aware of those claims and I wasn't trying to refer back to them.

I should have said that if you took Intel's bragging, and extrapolated that to a comparison of Conroe vs. X2 (which is what we cared about at the time!), Intel was implicitly claiming it's chips were somewhere roughly around 50% better than AMDs. There was a lot of debate about whether Intel was telling the truth, before[3] Anand's exclusive IDF 2006 benchmarks[4] came out and we had direct comparisons.

[1] http://www.dailytech.com/article.aspx?newsid=1148
[2] http://www.anandtech.com/tradeshows/showdoc.aspx?i...
[3] http://www.anandtech.com/printarticle.aspx?i=2503
[4] http://www.anandtech.com/tradeshows/showdoc.aspx?i...


With no hard dates....
By VooDooAddict on 1/26/2007 12:19:52 PM , Rating: 2
With no hard dates I won't be waiting for K8L.

My personal gaming system is still socket 754. My home office system is Dual Xeon 2.66ghz ([i]Single core Netburst[/i]).

I feel waited long enough for quad core. Unless I see hard dates and review numbers by the expected Intel price drop; I'll be going with [i]Clovertown.[/i]




RE: With no hard dates....
By antiterror13 on 1/26/2007 4:35:18 PM , Rating: 2
Windows XP Prof only support 2 CPUs .. so I suppose it would be useless for XP and because majority of computers in the world run XP ... is quad core useful ?

how about Vista ? ... does it support 4 CPUs


RE: With no hard dates....
By hstewarth on 1/26/2007 6:11:56 PM , Rating: 2
Actually Windows XP supports two physical cpus - so 2 quad cores will function under Windows XP.

Issue with # of cpus only comes into play with with 4 socket systems ( or more ) like AMD 8000's and Intel 7000's series which are typically use in servers.


Hopes too high for AMD
By TheDoc9 on 1/25/2007 5:45:21 PM , Rating: 2
Since AMD was on top for 2+ years in performance, I think we've all set the standard a little high. It's unlikely that AMD can really pull 40% off of this revision. As long as the performance is similar though I'll buy it myself.




RE: Hopes too high for AMD
By coldpower27 on 1/26/2007 12:36:59 AM , Rating: 2
I am skeptical at this point, a 40% figure from AMD's mouth doesn't mean very much, I want to see some actual performance figures, not abstract ones like 3.6x floating point power in comparison to Opteron Dual Core K8.


The price tag will put it out of reach!!!
By keepitreal on 1/26/2007 1:14:52 AM , Rating: 2
The price tag AMD will put on it. We will only be reading about.

I'm in wait and see mode for the 40% faster in real life test by expects.




By hstewarth on 1/26/2007 11:17:56 AM , Rating: 2
I wonder if the same as 64 bit being twice as fast as 32 bit.


L3
By fxnick on 1/28/2007 9:00:00 PM , Rating: 2
Not to many PC users know about L3 but it should help AMD out alot.
Before the G5 Macs every powermac had 2Mb L3 Cache, and i remember benchmarks showing L3 on VS L3 off made a big difference.




RE: L3
By verndewd on 1/28/2007 10:42:18 PM , Rating: 2
What are we talking about on the l3?I remember my barton had an l3.I just need enough info to connect the dots of the claim on l3.
How much perf gains are we talking here,and why?What exactly will the l3 do to the chip to make it 40% faster?
Thats where i am drawing a blank on K8L.How will an additional cache do so much with so much more routing?lets compare it to how intels cache works so i can understand the l3 claim.


Right away Sir
By crystal clear on 1/26/2007 5:52:04 AM , Rating: 2
AMD get ready has another K.O. punch coming soon-

Quote-
Roni Friedman and the Penryn processors
(credit: Israel Hadri)

Friedman has also publicly showcased, for the first time in the world, two of Intel's next-generation processors produced at 45nm (codenamed "Penryn") and told us that they are currently undergoing final phases of testing and debugging at Intel's development center in Haifa. These processors were designed in the USA (with the assistance of Israeli engineers) and are due to be launched in the second half of 2007, with no expected changes of schedule (according to Friedman).


http://hwzone.co.il/english-reviews/intel_apc_2006...




whats the perf average?
By verndewd on 1/28/2007 6:38:51 PM , Rating: 2
quote:
"We expect across a wide variety of workloads for Barcelona to outperform Clovertown by 40 percent," said AMD's corporate vice president for server and workstation products, Randy Allen.


This is the key to what is bieng said,I was telling my dad that the average over all applicable beches would not hit 40%.There are obvious benches that will see that,but i am thinking an avg of 20-25% acrossed the breadth of benches,at the most,as an average.If like big if,like supersized if,what they are saying is acually fact.
We can go through the last 12 months and find alot of verbal reason to doubt this claim;but i would like to be posotive about barcelona.




"I f***ing cannot play Halo 2 multiplayer. I cannot do it." -- Bungie Technical Lead Chris Butcher

Related Articles
AMD Hits Rough Waters
January 13, 2007, 2:00 PM
AMD 2007 Server Roadmap
December 14, 2006, 9:47 AM
Update: AMD Demonstrates Native Quad-Core
November 30, 2006, 2:27 PM













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