backtop

Blog: Hardware Barcelona Launch Clock Speeds Changing?
David Graham (Blog) - August 24, 2007 1:48 PM
Print E-mail del.icio.us 32 comment(s) - last by nukunukoo.. on Sep 11 at 9:44 AM
AMD is changing its "Barcelona" launch plans again

Let's come right out and say it:  when AMD announced that the launch clock speeds of Barcelona would start at sub-2.0 GHz, people were frustrated.  This long-awaited monolithic quad core release from AMD would be the pinnacle of their core design technology and would usher in a new era of powerful computing. 

Sound about right?  So, imagine the collective chagrin when a 1.6 GHz base clock speed was announced with top-end clock speeds ending around 2.0 GHz.  In the interim, reports surfaced about AMD having issues with top-end core speeds, which was quickly countered by the 3.0 GHz Phenom demonstration, leakage, and other processor related functions. 

We knew AMD was hard at work on perfecting the Barcelona core even before the official production launch and that the promise of HT3 in this generation of processors was going to be eschewed in favor of the more widely integrated HT1.x version. 

Now, word is coming that AMD might drop the 1.6 GHz from the launch line-up in favour of a 2.1 GHz or 2.2 GHz core launch speed.  Perhaps the fabrication process is going much smoother than anyone realized and some of the leakage problems have been ironed out? Who knows but, looks like things are picking up for AMD.

Comments     Threshold


This article is over a month old, voting and posting comments is disabled
Basis for this Information?
By TomZ on 8/24/2007 2:40:29 PM , Rating: 2
Is this information based on rumors/speculation, or is there real information coming from a reliable source?

I realize that this is a blog post, not a news post, but I think it is important to be able to judge the reliability of this information.

In any case, I sure hope it's right. If AMD launches Barcelona at 1.6GHz, I think many people will have to wonder why they bothered. Having good efficiency would be nice, however, if the overall performance is so deficient compared to similar Intel parts, then it's not really going to be very well-received. I think that launching at 2Ghz and higher is a must.




RE: Basis for this Information?
By davegraham (blog) on 8/24/2007 2:47:38 PM , Rating: 5
It's reliable as it came from an inside source at AMD. I work very closely with these guys and, for the record, this falls outside the scope of my NDA with them.

cheers,

Dave


RE: Basis for this Information?
By Amiga500 on 8/25/2007 3:35:36 PM , Rating: 2
Is this change related to the yields of the product, or the speed of the product?

Or can you answer that?


RE: Basis for this Information?
By Griswold on 8/24/2007 2:50:25 PM , Rating: 2
I dont think anyone ever said it will be a launch only at 1.6GHz. It was always 2GHz and some "sources" claimed 1.6 as the low and 2.0 as the high.


RE: Basis for this Information?
By TomZ on 8/24/2007 2:55:55 PM , Rating: 2
No, not really. The 1.6GHz came from the demos all the AMD partners did of Barcelona in May/June with processors that ran at only 1.6GHz.

http://www.dailytech.com/AMD+Partners+Demonstrate+...

This, coupled with widespread reports of problems running at 2GHz, is why people were thinking that it might launch at 1.6Ghz and be ramped up later.


RE: Basis for this Information?
By Anh Huynh on 8/24/2007 3:01:41 PM , Rating: 4
AMD publicly stated the top out speed would be 2.0 GHz in an official press release, with lower end models.

http://www.dailytech.com/AMD+to+Launch+Barcelona+S...

1.6 GHz was the speed of most of the Barcelona samples sent to partners and demonstrated at Computex.

So far, the launch plan for Barcelona, according to the last roadmap, has 1.7 GHz as the lowest launching clock speed.


Great work AMD
By HrilL on 8/25/2007 4:19:07 PM , Rating: 1
If the 2.2ghz with a 40% gain is true. That should put it on par with a 3-3.2ghz core 2 quad if not even be better since it will have more memory bandwidth =). I hope all this is true. If it is things might be changing for the better for AMD. But Only time will tell.




RE: Great work AMD
By nukunukoo on 8/26/2007 1:06:54 AM , Rating: 1
Unfortunately, that will not be the case. Barcelona does have architectural advantages over the Core 2, like being a True Quad Core (however benchmarks show very little performance gains on multithreaded apps). It also has a shorter pipeline (12 vs 14) but again, the advantages are minimal. The on-chip memory controller's advantages are offset by using a smaller cache size than Intel's. And the Core 2's 4 pipes and macro fuse simply is one powerful feature.

The architecture has an uphill battle with Core 2 and the only way they can compete is with aggressive pricing and probably introducing multicores using more than just 2 chips to the masses (4x4x4x4). It makes sense since that will help AMD sell more chips and would really make me seriously consider going back to AMD since I do a huge amount of animation rendering, and use highly multithreaded applications.

At the same clock, based on initial benchmarks, Barcelona is still 5-10% slower than Core 2. I don't know where you got your numbers.

Alas, Penryn is in the horizon and at 45nm, will break the 4GHz barrier for overclockers this year. So essentially, Barcelona is a stopgap solution. Considering that the latest Core 2 iteration is 5% faster on average and the use of SSE 4 nearly doubles video encoding (!!!), I'd say that Shanghai had better be good!

In the near future, Expect Intel's Nehalem/Westmere to bring back Hyperthreading (but a more efficient version, using a larger Trace Cache, a shorter pipeline than Pentium 4's, and an L2 code cache) and adopting an on-chip memory controller. And, unfortunately, that is the generation that Shanghai is up against!

I'm no AMD basher, I like AMD, I started with AMD, but they better get much more aggressive than the roadmap they currently propose.


RE: Great work AMD
By OneEng on 8/26/2007 7:58:52 AM , Rating: 5
First, about the only thing anyone cares about with Barcelona is multi-threaded server and workstation applications. It is a processor squarely targeted at the server and workstation market (the high end profit stuff which is smart).

In these types of applications, scaling is nearly perfect with the number of cores (for most applications).

Second,

I don't know where you got YOUR numbers. When has there EVER been anything that showed Barcelona being slower clock for clock than Core 2?

The NEAREST thing we have is this: http://www.theinquirer.net/default.aspx?article=41...

In this DIRECT comparison and one in which Penryn should have the MOST advantage possible.

In this situation, Barcelona still holds a 10% IPC advantage over Penryn. Since Penryn holds a 5% general IPC advantage over Cloverton, then Barcelona would have 15% on Cloverton.

Now, this benchmark is NOT a server workstation benchmark.

From here: http://www.anandtech.com/IT/showdoc.aspx?i=3039

In typical server/workstation loads, the CURRENT K8 only lags Core 2 by around 15% vs the 30% K8 lags in the typical desktop application.

For the original target market (server and workstation), I am guessing that Barcelona will hold a 25-30% IPC advantage over Cloverton and a 15-20% advantage over Penryn (I think that the added cache will make more difference in server workloads than it does on desktop ones).

The 2.3Ghz suggested top clock in this blog would then equate to a 2.8-3.0Ghz Cloverton or a 2.6-2.8Ghz Penryn.

This also makes the 2.3Ghz max release make a great deal of sense. It matches up very well with Intel's best CURRENT quad core speeds despite the fact that Intel will trump this with their 3.16Ghz Penryn release.


RE: Great work AMD
By OneEng on 8/26/2007 8:02:14 AM , Rating: 2
Oops, I got the wrong top clock. Just adjust for 2.2 vs 2.3.


RE: Great work AMD
By nukunukoo on 8/26/2007 6:54:06 PM , Rating: 2
I have always agreed that AMD's core-scaling in the server and workstaion market have been truly impressive.

Unfortunately, there is a fundamental flaw in your argument. The benchmarks you gave DOES show AMD getting the lead but the tests specifically stresses server-specific inter-core transactions.

This does not necessarily translate to real-world performance. Another case in point: AMD's Athlon 64's memory benchmarks are generally faster than the Core 2. Such transfers parallel the benchmarks you mentioned since, in effect, deals with communications between subsystems (where Opterons generally do better, no surprise there). But even with those memory benchmarks it still does not mean the current Athlon 64 is faster than the Core 2.

Basically what I am saying is that the K10 architecture (server, workstation, desktop, whatever... they all share the same fundamental architecture. AMD's K10 Desktop is still a Barcelona-based core.) won't be faster that the current and tweaked Core 2 at launch. Period.

I believe this is the better hint of things to expect:

http://www.dailytech.com/Quick+and+Dirty+AMD+K10+C...

It goes to reason that Barcelona is a fine-tuned and tweaked version of the Current Athlon 64 which is quite commendable. But I absolutely doubt that those tweaks will translate into a 40 to 60% performance increase that the Core 2 is enjoying against the Athlon now.

But this issue is best addressed when Barcelona is launched. By then you can prove me wrong that Intel's latest iteration is slower, even slightly, which I seriously doubt.


RE: Great work AMD
By nukunukoo on 8/26/2007 6:56:36 PM , Rating: 2
Errata: my bad, make that 25 - 40% performance improvement.


RE: Great work AMD
By nukunukoo on 9/11/2007 9:44:10 AM , Rating: 2
Now that the Barcelona is out, initial tests prove my prediction correct and your estimates wrong, actually, way wrong.


RE: Great work AMD
By athlonotaku on 8/27/2007 11:17:16 AM , Rating: 2
quote:
The on-chip memory controller's advantages are offset by using a smaller cache size than Intel's.


I suppose that'd be a disadvantage if the achitectures were the same, but they are not. Its a well known fact that the Core 2 architecture is cache hungry, while AMD's architecture doesn't benefit nearly as much with a large cache. Which could explain why AMD only gives each individual core 512kb of L2 cache on Barcelona.

Other than that little tidbit, I agree with pretty much everything else.


hum
By AMD Opteron on 8/24/07, Rating: 0
RE: hum
By Devo2007 on 8/25/2007 5:43:25 AM , Rating: 3
A 1.6GHz quad core does not equal 6.4GHz, since you would virtually never see a gain of 100% per additional core added.


RE: hum
By Devo2007 on 8/25/2007 5:49:37 AM , Rating: 2
No edit button -- whoops.

Basically, you can't compare a 1.6GHz quad core to a 3.2GHz dual-core (the dual core would win out on any benchmark you throw at it, if the CPUs are otherwise identical). Just like SLI/Crossfire, you'll rarely (if ever) see twice the performance by adding another core.


RE: hum
By 3kliksphilip on 8/25/2007 12:03:51 PM , Rating: 2
Although not a 100% gain, I like the idea of one core doing the virus killer and desktop whilst you play a game on the other core whilst 2 cores... hey, what are they doing? I wish that I had a little program where I could see what each core is doing. I could even disperse the information my own way and see if there's any performance gain. I know that they'll probably do a little of everything each, but I get excited about the thought of more than one processor. By the time we get to 64 or 128 chips I believe that the results could be truly remarkable, especially for gamers.

Simplified, I just like the idea of splitting the work into loads of small pieces. I never thought it was possible (I mean, surely fitting 2 processors is just the 'cheap' way compared to one super fast processor?) but I'm pleased that it's being done. It'll get to the point where something goes wrong but it'll only affect one of the cores, leaving the other 1, 3, 7 or what ever active. Great invention. The only step up is where the R700 is going, with loads of slots for small graphics cards, allowing for you to fit as many as you like. As you can see, I'm looking forward to this.


RE: hum
By Muirgheasa on 8/25/2007 5:43:36 AM , Rating: 2
Even though that really isn't how it works, at all. Quad core 1.6GHz does not, nor will it ever, equal 6.4GHz. Neither will 3.2 x 2 ever equal 6.4. It just doesn't doesn't scale like that. 1.6 x 4 is simply that: 1.6, times 4 (but not multiplied out. Just 1.6 x 4)


RE: hum
By RussianSensation on 8/28/2007 10:33:37 AM , Rating: 2
You realize there are programs that take full advantage of additional cores like rendering, video/audio applications, Folding@home, SETI@home just to name a few. In those applications the performance scales 97-98% (of course the fact that the processor isn't natively quad core is why it's not 100%. You can also check out Valve particle benchmark to see that Q6600 is 95% as fast as E6600 2.4ghz.

My Q6600 @ 3.5ghz (3.5x4 = 14ghz) is 3x faster than my friend's E6600 2.4ghz (2.4x2 = 4.8ghz) at SETI@home. You can believe it or not believe it but that's reality.


We can only hope so
By grenableu on 8/24/2007 2:19:42 PM , Rating: 2
1.6 ghz will be a real downer for AMD. 2.2 ghz would be much better.




RE: We can only hope so
By Master Kenobi (blog) on 8/24/2007 3:23:15 PM , Rating: 2
Indeed. At 1.6Ghz there would be little point, but at 2.2 they will fit in well enough against the Xeon line.


RE: We can only hope so
By porkpie on 8/24/2007 5:34:45 PM , Rating: 2
Depending on price, of course.


RE: We can only hope so
By Master Kenobi (blog) on 8/26/2007 2:40:02 PM , Rating: 2
Of course.


This new old story has about 8 days now...
By fikimiki on 8/25/2007 6:00:11 PM , Rating: 2
* Opteron 2348, 1.9GHz, 95W - $320;
* Opteron 2350, 2.0GHz, 95W - $390;
* Opteron 2352, 2.1GHz, 95W - $450;
* Opteron 2354, 2.2GHz, 95W - $610;
* Opteron 2356, 2.3GHz, 95W - $795;
* Opteron 2358, 2.4GHz, 95W - $1180;
* Opteron 2360, 2.5GHz, 120W - ?;
* Opteron 8348, 1.9GHz, 95W - $790;
* Opteron 8350, 2.0GHz, 95W - $1025;
* Opteron 8352, 2.1GHz, 95W - ?;
* Opteron 8354, 2.2GHz, 95W - $1190;
* Opteron 8356, 2.3GHz, 95W - $1550.




By PaxtonFettel on 8/26/2007 7:20:10 AM , Rating: 2
Um , where are those numbers from?


By Roy2001 on 8/24/2007 7:47:07 PM , Rating: 2
Why you post misleading comparison chart based on simulated unexist product?




Not necessarily a good thing,
By PaxtonFettel on 8/25/2007 4:32:48 AM , Rating: 2
if it delays things even more whilst they refine the silicon to higher speeds. Might they not be better off starting out at the low speeds and then ramping up very quickly after launhc. I know that'll make it the opposite of most CPU launches (eg highend first then the rest) but it might help them get the full range out sooner.




Hey dave!
By Mime on 8/25/2007 1:46:47 PM , Rating: 2
ah ha! So this is where dave hangs out these days. :) Long time no see dude.

To get that kind of jump in clock speed from a monolithic quad core chip yields must be pretty good indeed. I guess we'll have to wait and see how things shape up come launch day.




Barcelona is out
By nukunukoo on 9/11/2007 9:38:23 AM , Rating: 2
Well, like I said previously, readers saying that Barcelona will outperform Intel's current Core 2 by 10 to 30% is greatly exaggerated. @OneEng, your estimates were quite off (actually, way off). Though I am glad that initial 3DS Max scores look promising, but not even close clock for clock. Don't get me wrong: I was a "loyal" AMD user but when the Core 2 came out, I upgraded out of necessity.




At Circuit City
By AntiV6 on 8/26/07, Rating: -1
RE: At Circuit City
By zornundo on 8/27/2007 2:59:44 PM , Rating: 1
Heck, I saw a dude wearing an AMD polo-type shirt outside a local CompUSA this past weekend. He could have been an AMD rep, but I didn't feel like pestering him any :P

Barcelona blah blah - the real question is what speed will phenom launch and how much $$$ will it be?


"Young lady, in this house we obey the laws of thermodynamics!" -- Homer Simpson

DailyTech Poll
Which web browser do you use on your primary personal machine? 






44 Comments

Latest By David Graham










botimage
Copyright 2009 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki