backtop


Print 45 comment(s) - last by mindless1.. on Sep 16 at 7:50 PM

FX CPU ran at 8.429 GHz with liquid nitrogen cooling

AMD has been content to let Intel offer the pricier and often faster processors to the enthusiast desktop crowd for a while now while AMD focused on the lower price crown. In fact, the company’s value-oriented Fusion processor has been a great performer for the company.

Today, however, AMD announced that its FX 8-core desktop CPU has grabbed a Guinness World Record for the highest frequency for a computer processor to date. The processor was able to run at 8.429 GHz, beating the previous record of 8.309GHz.

The team had to resort to some extreme measures to cool the processor enough to hit that record-breaking frequency. Air or water-cooling was out and Team AMD FX had to resort to liquid nitrogen for cooling. The team included some overclocking gurus along with AMD folks in the effort. 

The clock speed of the processor was validated using CPU-Z and will go into the record book. Although AMD was able to overclock the FX CPU to "well above 5GHz" using air or liquid cooling systems that cost under $100, the team used a three-phase system to make its record run.

 “The record-breaking processor speed that resides in the AMD FX CPU clearly demonstrates performance gains for the new AMD Bulldozer multi-core architecture, which will provide x86 computing power for this CPU and future AMD Accelerated Processing Units,” said Chris Cloran, corporate vice president and general manager, Client Group at AMD. “Along with world-record frequencies, the AMD FX processor will enable an unrivaled enthusiast PC experience for the money – extreme multi-display gaming, mega-tasking and HD content creation.”

AMD has a video posted that shows the overclocking process and the record run.



Comments     Threshold


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

Cool, but...
By Motoman on 9/13/2011 1:38:56 PM , Rating: 1
...while I'm all for AMD vs. Intel, and think that this is a nifty bit of achievement, we all know it's not the pure clock speed that makes a CPU "fast."

Sure, they got the clock speed up to stratospheric heights, but that's not a sure indicator that it was fast than the Intel chip that previous held the record...in terms of actually doing work.

It probably didn't run at that speed long enough to do any "work," but it would have been interesting to at least see some benchmarks along the way...although even that is kind of academic. Instructions per clock scale linearly with clock speed...so if Chip X has an IPC greater than Chip Y at 2Ghz, it'll still have a higher IPC at 8Ghz.




RE: Cool, but...
By AMDftw on 9/13/2011 2:00:05 PM , Rating: 1
I want to go see the demo. It was held here in Austin, TX. There is a video of it on PC perspective website. What i didn't understand is why in the hell did they use Win XP...? Their is rumors of a benchmark @8.3 but couldn't find a score or vid on it. I'm pro AMD and that's it. I have tried (3 times)Intel but just didn't seem to run as smooth as my AMD systems. I don't want to hear anyone bashing my post because I didn't post nothing bad about Intel its just my opinion.


RE: Cool, but...
By Kurz on 9/13/2011 2:15:59 PM , Rating: 2
Well... honestly it comes down the stability of the motherboard. Those issues with intel boards probably because of the BIOS/UEFI, though if you were to try those systems again and update the firmware and the BIOS I bet it'll run much smoother.


RE: Cool, but...
By AMDftw on 9/13/2011 2:25:20 PM , Rating: 1
for Intel I've use EVGA, Intel, And Gigabye. I built a E8400, Q6800 and I7 920. I'm a firm believer of updated BIOS and software and out of all of them the E8400 was the best. I still ended up switching back to AMD no matter if Intel was faster. I guess using AMD from the Thunderbird era kind of grew on me.


RE: Cool, but...
By bug77 on 9/13/2011 5:03:01 PM , Rating: 2
Ironically, the Thunderbird era is when the legend of AMD's "instability" was born. All because of VIA's KT266 chipset which in turn evolved into the legendary KT400(A).


RE: Cool, but...
By Clauzii on 9/13/2011 6:38:56 PM , Rating: 2
The KT266 chipset was already fixed with the KT266A. The fault was that the KT266 got instable with the AGP at x4. Run @ x2 is ok in most cases.

I still have a KT266A with a 9600Pro @ AGP x4. Solid as a rock.


RE: Cool, but...
By Alexvrb on 9/13/2011 11:29:51 PM , Rating: 2
Yeah and before that you had the KT133 which was junk, and was fixed with the KT133A. Noticing a pattern? The only VIA chipset from those days that was able to skip needing an "A" revision was maybe the KT333. Which in and of itself wasn't that much different in design from the KT266A.

Although to be fair, VIA did clean up their act over time, and they were still *usually* better than SiS. Not to mention even Nvidia had quite a few problematic chipsets.


RE: Cool, but...
By silverblue on 9/14/2011 2:16:55 PM , Rating: 2
I loved my KT266A board. It was stable, had plenty of BIOS options, and outperformed the unstable K7S5A board (SiS 735) which died on me. The only bad thing VIA could do at that point would be to release bad drivers (and they did; one release damaged hardware, but at that point I'd gotten an nForce 2 board).


RE: Cool, but...
By FaceMaster on 9/13/2011 5:20:54 PM , Rating: 3
But... using a certain company's processor doesn't change things like, say, driving a different car does. Do you really use a random PC and go 'URGH I don't like this one, it's clearly powered by Intel!' or anything like that?


RE: Cool, but...
By Targon on 9/14/2011 10:36:16 AM , Rating: 1
When you run into problems caused by low end Intel chipsets, yes, I DO feel like that. Intel may make a great CPU, but their chipsets leave a LOT to be desired.


RE: Cool, but...
By FaceMaster on 9/14/2011 12:53:35 PM , Rating: 2
Can't say I've ever had that problem myself, could you link me to information about this since I've been unaware of it until now?


RE: Cool, but...
By Belard on 9/13/2011 6:39:17 PM , Rating: 2
I'm okay with my Intel system and my ThinkPads all come with Intel... hopefully on my next ThinkPad purchase, it'll come with an A-Series CPU in a T-Series body.

But I hope to replace my Quad core intel with an AMD FX/bulldozer setup.


RE: Cool, but...
By Targon on 9/13/2011 6:26:47 PM , Rating: 2
Windows XP has a lower overhead compared to Windows XP, so for CPU related workloads would be better. For testing of combined CPU+GPU results, then a stripped down Windows 7 would be a better bet.


RE: Cool, but...
By Alexvrb on 9/13/2011 11:36:22 PM , Rating: 2
But they're not benchmarking it. They're just pushing clocks up as hard as possible and running CPU-Z to validate. I suspect running Windows XP has to do with it being slightly more forgiving of unstable overclocks than even Win7.


RE: Cool, but...
By Da W on 9/13/2011 3:35:47 PM , Rating: 2
It means a lot since:

1. from technical previews it doesn't look like bulldozer has such an extended pipeline and low IPC as netburst had.

2. almost every preview and (real or false) leaked benchmark is confident about bulldozer multitreading potential

3. almost every rumors about the cause of bulldozer delay and almost every concerns was about AMD not being able to achieve high enough clock speed. I think they proved wrong.

4. Even the 8,4 Ghz running on only 2 cores (1 module)is important since 1 module is almost the same a Intel's core i3 being a dual core with hypertreading and turbo boost disabled. A single module bulldozer could make a great low-power laptop chip. And it shows how agressive turbo mode on an 8 core chip can be.


RE: Cool, but...
By someguy123 on 9/13/2011 9:14:40 PM , Rating: 3
I don't see any preview benchmarks showing good performance. On the contrary, the only tests that haven't been proven fake are those leaked by the chinese.

In theory it would seem like the design would offer better performance for out of order tasks, but we've yet to see any real testing. I don't see how you can compare it directly to another processor like the i3 when legitimate benchmarks don't even exist.

Also, that bulldozer delay wasn't a rumor. AMD themselves said that they weren't hitting turbo spec with their last stepping.


RE: Cool, but...
By 2bdetermine on 9/14/2011 3:31:44 PM , Rating: 1
Let's see, loading windows, executed CPU-Z if that was not actual work I don't know what then...


RE: Cool, but...
By mindless1 on 9/16/2011 7:50:37 PM , Rating: 2
IPC doesn't scale linearly with clock speed when doing real "work" because you still have variable bus speeds involved, data moving to do... though I do agree, all else being equal (but when is that ever really true?), one with higher IPC at 2GHz would retain it at 8GHz, except that to hit a particular speed you may be fiddling with raising or lowering the bus speeds to hit a particular CPU clock speed.


Liquid Helium
By BioHazardous on 9/13/2011 1:47:40 PM , Rating: 5
I could've swore in the video they said they got to that record after they switched to Liquid Helium from Liquid Nitrogen...




RE: Liquid Helium
By AMDftw on 9/13/2011 2:00:39 PM , Rating: 2
yes they did.


RE: Liquid Helium
By Manch on 9/13/2011 2:42:41 PM , Rating: 2
According to Anandtech it was helium:

http://www.anandtech.com/show/4770/amd-sets-world-...

They achieved 4.8 using an ANTEC water cooler.

They also have pricing info listed in another article. Top of the line proc is 3.6/4.2turbo:

http://www.anandtech.com/show/4767/pricing-reveale...


RE: Liquid Helium
By bug77 on 9/13/2011 3:25:06 PM , Rating: 2
I was asking myself something similar: is going from 8.309GHz to 8.429 GHz really an advancement in CPU tech or in cooling tech? After all, it's less than 1.5% faster. Still a record, but not ground breaking at all.


RE: Liquid Helium
By Targon on 9/13/2011 6:30:02 PM , Rating: 3
Most people feel that Intel has been way out in front of AMD in terms of CPU design. Now, one test of a design is how far it can be overclocked and still be stable. Yes, you could say that there have been advances in cooling, but the real kicker is that Bulldozer can be clocked at that speed, and it is well beyond what the current K10.5 chips could be overclocked to, even with extreme cooling.

Remember, there have been difficulties in getting the current generation AMD processors to go much beyond the 4GHz mark, and from what they have said, they are hitting 5GHz on high-end air cooling


RE: Liquid Helium
By Reclaimer77 on 9/13/11, Rating: 0
RE: Liquid Helium
By Fritzr on 9/13/2011 9:35:32 PM , Rating: 5
AMD marketing can claim that an AMD chip is listed in the Guiness Book. In marketing, image is everything, reality takes a back seat.


RE: Liquid Helium
By Reclaimer77 on 9/14/11, Rating: -1
RE: Liquid Helium
By priusone on 9/14/2011 10:25:38 AM , Rating: 2
What version is Firefox on? v6? Geez, and here I thought it was just at v5...

So you spend a couple of bucks to make your product look 'awesome' to some beginner builders. "Should I buy this or this. Wow, a world record speed!"...

"So you're vastly overstating the impact of marketing in the CPU consumer space in my opinion."

I think your vastly underestimating human silliness.


RE: Liquid Helium
By Targon on 9/14/2011 10:32:20 AM , Rating: 2
The top speed is a suicide shot, but the 5GHz on high end air implies that with the current implementation of Bulldozer, 4GHz and 4.5GHz with an eight core chip SHOULD be a realistic expectation. And this is with the initial Bulldozer release, which will have a lot of room for improvements.

Many people have not expected much from AMD, even with the new design, so the real question here is what sort of performance Bulldozer will provide at a given clock speed. Bulldozer SHOULD be faster per-clock compared to the current Athlon 2/Phenom 2 chips, the question remains how much faster. Eight core at 4GHz does sound rather tasty at this point, even if Intel still has an edge.


RE: Liquid Helium
By Clauzii on 9/13/2011 6:46:15 PM , Rating: 3
Records are meant to be broken all the time ;)

Every record is groundbraking. It makes the competitors run too.


RE: Liquid Helium
By NicodemusMM on 9/13/2011 7:26:01 PM , Rating: 3
Please keep in mind that the 8.309GHz record was with a single core processor. The new 8.429GHz record is with an 8 core processor with more advanced features... hence greater complexity.

~ Nicodemus


RE: Liquid Helium
By someguy123 on 9/13/2011 9:10:49 PM , Rating: 2
I believe all but one module was disabled. This does allow them to claim "2" cores over one, but the relevance depends a bit on how well their module architecture works, which has yet to be revealed.


False
By Reclaimer77 on 9/13/2011 1:38:56 PM , Rating: 3
quote:
“The record-breaking processor speed that resides in the AMD FX CPU clearly demonstrates performance gains for the new AMD Bulldozer multi-core architecture


No. The only thing this demonstrates is the Bulldozer has a temperature sensor calibrated for extreme low temps so that it doesn't "cold bug" under these conditions and refuse to boot like a lot of CPU's under liquid cooling. This was a "suicide shot" record and in no way was a stable overclock or indicative of real world performance gains.

Everyone who's ever overclocked understands this. It's impressive but doesn't really mean anything. It's hype.




RE: False
By kattanna on 9/13/2011 2:06:40 PM , Rating: 2
the LN2 cooling, while impressive, is that a stunt.

but what really interested me from the video was how it was running above 5Ghz with "aggressive" air and stock water cooling, and running under load.

now that was impressive.


RE: False
By Reclaimer77 on 9/13/2011 2:15:50 PM , Rating: 1
I have to wonder if AMD is becoming the new 1990's Pentium 4? Their architecture can't really compete with Sandy Bridge, so they're jacking up the clocks to compensate?

I would rather they spent the time doing this overclock on benchmarking the chip instead. At least then we would have some kind of idea of what this thing can really do with those clocks.


RE: False
By adiposity on 9/13/2011 2:34:01 PM , Rating: 2
quote:
I have to wonder if AMD is becoming the new 1990's Pentium 4? Their architecture can't really compete with Sandy Bridge, so they're jacking up the clocks to compensate?


I was under the impression Netburst was designed to have high clock speed from the beginning. They deliberately lengthened the pipeline so they could get higher clocks with less performance per clock.


RE: False
By silverblue on 9/14/2011 2:20:51 PM , Rating: 2
...and throughout the P4's life, they enhanced the chip with a faster FSB, more cache, HT and some other tweaks to ensure performance scaled linearly. I saw some graphs on the Anandtech XP3000+ review the other day which nicely demonstrates a) how P4 was scaling along with the IPC boost shown when Northwood came out, and b) that the Athlon XP was running out of legs. Still, P4 was far too hungry and hot.


RE: False
By FaceMaster on 9/13/2011 5:26:50 PM , Rating: 2
quote:
I have to wonder if AMD is becoming the new 1990's Pentium 4? Their architecture can't really compete with Sandy Bridge, so they're jacking up the clocks to compensate?


Then again, their 'lower speed, lower price' policy has worked brilliantly in the graphics card industry. I guess you could look at their increased number of cores and higher clocks as the brute-force approach, but it could also be a simple case of finding a different solution to a problem. If it ends up being comparable with Intel's CPUs, I see no problem with the method. Especially if it means that you get the choice between good gaming chips, and those that are more suited to video editing and the like.

Just as long as they are competitive with Intel.


RE: False
By SiliconJon on 9/13/2011 2:08:36 PM , Rating: 2
Yeah, the clock speed doesn't mean much to me anymore - that became an exposed myth long ago.

How about a FLOP measurement? Or maybe a benchmark based or MURPHY or something similar?

Or still better yet: Graph500.


RE: False
By Targon on 9/14/2011 10:45:07 AM , Rating: 2
It isn't really a myth, where clock speed alone is the benchmark for performance. With the same design, high clocked chips will have a higher performance. We can assume that AMD is not going to go to a new design where you need higher clock speeds to get similar performance to the previous generation chips. So, if Bulldozer is faster on a per-clock basis compared to K10.5, and at the same time, Bulldozer is able to hit 4GHz sustained speeds all the time, not just in turbo mode, that DOES give us quite a bit of hope.

Remember, Bulldozer will be released with the high end being 8-core(4 modules) running at 3.6GHz normal, 4.2GHz turbo for the INITIAL chip release, and will be sold for $320 for the CPU. In multi-threaded applications, that should be a very solid performer, even if it does not deliver better performance per clock. Considering many are happy with the performance of a Phenom 2 x4 at 3.2GHz and above, Bulldozer SHOULD be seen as a nice step up.


RE: False
By Kurz on 9/13/2011 2:13:01 PM , Rating: 2
I heard it was just 1 module active as well.


1.21 JIGAWATTS!
By Rabscuttle on 9/13/2011 1:40:04 PM , Rating: 5
Just to beat everyone to the punch...

... But can it play Crysis!
... It's like glazing a stick of dynamite!
... Apple invented it first!




RE: 1.21 JIGAWATTS!
By FaceMaster on 9/13/2011 5:48:20 PM , Rating: 3
I think the real question is 'When is it out?'


RE: 1.21 JIGAWATTS!
By rcc on 9/14/2011 1:43:20 PM , Rating: 2
Followed (or preceeded by) "Can they mass produce it".


RE: 1.21 JIGAWATTS!
By NicodemusMM on 9/13/2011 7:32:12 PM , Rating: 2
"... It's like glazing a stick of dynamite!"

Isn't it supposed to be "lasing" versus "glazing". I don't recall pottery in Real Genius.


By superstition on 9/13/2011 4:32:59 PM , Rating: 2
I have one word for you:

mindshare

and one other:

marketing

that is all




"The whole principle [of censorship] is wrong. It's like demanding that grown men live on skim milk because the baby can't have steak." -- Robert Heinlein














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