AMD Bulldozer Engineering Sample CPU Overclocked to 4.63GHz

[sangyup81]

Considering the resources Intel has plowed into developing not only Sandy Bridge's architecture but also the 32nm process tech that enables it's clockspeeds and power-consumption...the only thing that we should expect to be able to beat Sandy Bridge is Ivy Bridge.

Miracles can happen, but Sandy Bridge is no Pentium 4 Netburst step-back situation. Miracle denied IMO.

The only thing AMD can hope to do is beat Intel in Price/Performance in their i3 segment and perhaps their lower end i5s. Of course, all Intel would have to do is lower their prices (which I hope they do!)

 

[AnandThenMan]

Considering the resources Intel has plowed into developing not only Sandy Bridge's architecture but also the 32nm process tech that enables it's clockspeeds and power-consumption...the only thing that we should expect to be able to beat Sandy Bridge is Ivy Bridge.

Miracles can happen, but Sandy Bridge is no Pentium 4 Netburst step-back situation. Miracle denied IMO.

AMD has beat Intel several times before, using a small fraction of the R&D budget. It can happen again.

 

[Diogenes2]

Several ?

 

[996GT2]

AMD has beat Intel several times before, using a small fraction of the R&D budget. It can happen again.

Not since the day Core 2 was released, and that was 5 years ago.

AMD has been playing a perpetual game of catch-up to Intel ever since the Core 2 days, and they haven't been winning it, either.

 

[nanaki333]

Several ?

slot a vs slot 1
p4 vs athlon xp
pd vs a64 x2
i would argue that k6-2 and k6-3 were better than pentium back in the day too if you were a gamer (as much as you could be back then anyway).

amd got way too comfy after the a64 series though and were blindsided by c2d. since then it's been intel's game

 

[bridito]

Just ploughed through this:

http://www.eweek.com/c/a/Desktops-a...s-Will-Feature-TDP-Capping-Technology-834387/

...and my eyes are blurry. OK, so let me get this correct: In a server farm, you can adjust a BD Opteron so that it uses less electricity without diminishing performance. So... it begs the question, why not just set it at that max performance min electricity consumption setting at the factory? Is there an advantage to operating the TDP Power Cap at higher power draws? If there is, I sure can't figure it out by that article!

 

[Vesku]

It's saying with BD Opterons you won't have to be a giant customer and work directly with AMD to set your CPUs maximum TDP at something lower than the max rating from the factory. I'm not big into datacenter type operations but it seems reasonable that each facility and perhaps certain areas of the facility have a maximum amount of power they can deliver. I've read that the obsession with keeping everything as close to 16C-18C ambient has gone away for the most part and so power delivery and UPS capacity has become the primary concern in the dense computing world.

Edit: Oh, and that once they set their max TDP they can self experiment to find the frequencies they are comfortable with operating within the new TDP envelope.

 

[GammaLaser]

Just ploughed through this:

http://www.eweek.com/c/a/Desktops-a...s-Will-Feature-TDP-Capping-Technology-834387/

...and my eyes are blurry. OK, so let me get this correct: In a server farm, you can adjust a BD Opteron so that it uses less electricity without diminishing performance. So... it begs the question, why not just set it at that max performance min electricity consumption setting at the factory? Is there an advantage to operating the TDP Power Cap at higher power draws? If there is, I sure can't figure it out by that article!

I would read this blog article by JF. It explains it better than the eweek article.

http://blogs.amd.com/work/2011/06/28/tdp/

 

[Markfw]

slot a vs slot 1
p4 vs athlon xp
pd vs a64 x2
i would argue that k6-2 and k6-3 were better than pentium back in the day too if you were a gamer (as much as you could be back then anyway).

amd got way too comfy after the a64 series though and were blindsided by c2d. since then it's been intel's game

I would agree here...

 

[Minerva]

IPC aside a stable o/c to 4.63 GHz on that chip would be an impressive feat.

 

[996GT2]

IPC aside a stable o/c to 4.63 GHz on that chip would be an impressive feat.

Why? It's been long known that Bulldozer was designed with high clock speeds in mind rather than the high IPC of the Sandy Bridge processors.

Just about every single 2500K or 2600K out there can do 4.63 GHz with ease. Bulldozer needs to be able to overclock higher than that to be competitive.

Also, did you see the voltages used in the OP? That BD chip needed 1.5V to reach 4.63 GHz. My 2500K will do 5 GHz prime-stable at that voltage, and there are many 2500/2600Ks out there that will do 5 GHz stably at lower voltages.

 

[Diogenes2]

What makes you think anything about the OP was factual ?

 

[podspi]

Is everybody here talking about singlethreaded or multithreaded performance, and which Bulldozer series? I don't expect any BD to be faster than SB in singlethread performance, but I expect FX-8 to be faster than the 2600K in multithreaded code. If it isn't AMD is in for a world of hurt, unless Fusion catches on in a super-huge way, very quickly.

Or, OEMs just stop caring and start slapping AMD CPUs into computers just 'cause.

 

[Ken g6]

AMD is in for a world of hurt, unless Fusion catches on in a super-huge way, very quickly.

I think AMD has basically said as much; they've said that they're betting huge on Fusion, anyway.

 

[drizek]

I've read that the obsession with keeping everything as close to 16C-18C ambient has gone away for the most part and so power delivery and UPS capacity has become the primary concern in the dense computing world.

I still remember how cold it was in my computers class in middle school.

Anyway, reading the JFA piece ,this seems to be exactly like what they are doing with their video cards. It makes sense to be doing this moving forward. Was it Intel with IB that is going to allow laptop chips to exceed TDP when docked?

 

[AtenRa]

Come on now, would you really be surprised that Bulldozer is not faster than Sandy Bridge? That's simply what I've been expecting all along, which is why I didn't even bother waiting for BD and just upgraded straight to Sandy Bridge for my build earlier this year.

Since AMD Phenom II X6 1100T can beat Intel Core I5 2500K in almost all multithreaded applications at the same price point, i would say BD will be much faster at the same apps even than SB 2600K.

I'm sure you all would like a link to support what i say,

http://www.anandtech.com/show/4310/amd-phenom-ii-x4-980-black-edition-review/1

No mater how fast your IPC will be, more cores are better in multithreaded applications and since BD will have both higher IPC and more Cores (than Phenom II) we really have to wait and see the real silicon and what is capable off.

 

[Skurge]

Since AMD Phenom II X6 1100T can beat Intel Core I5 2500K in almost all multithreaded applications at the same price point, i would say BD will be much faster at the same apps even than SB 2600K.

I'm sure you all would like a link to support what i say,

http://www.anandtech.com/show/4310/amd-phenom-ii-x4-980-black-edition-review/1

No mater how fast your IPC will be, more cores are better in multithreaded applications and since BD will have both higher IPC and more Cores (than Phenom II) we really have to wait and see the real silicon and what is capable off.

Good point.

 

[ydnas7]

but even in those Samples the AMD 6 core tended to be bested by 45nm Nehalem 4C8T

 

[Idontcare]

Since AMD Phenom II X6 1100T can beat Intel Core I5 2500K in almost all multithreaded applications at the same price point, i would say BD will be much faster at the same apps even than SB 2600K.

Isn't the 2500K basically an intentionally dumbed down Sandy? Hyperthreading is disabled, right? Reduced cache?

One would hope a 6C thuban can out-do a 4-core no-HT chip in multi-threaded apps unless the IPC per core on a SB is 50% higher than that of the IPC of a Stars core.

That said, if you are on a lower-end budget (where $100 more for a fully-functional SB 2600K is too much) and you use multi-threading apps then the obvious winner for your dollars is the X6.

 

[Abwx]

One would hope a 6C thuban can out-do a 4-core no-HT chip in multi-threaded apps unless the IPC per core on a SB is 50% higher than that of the IPC of a Stars core.

Without AVX , SB is good in INTEGER or in FP but not in both
simultaneously given that it has no more than 3 exe ports shared
by boths INT and FP instructions.

In fact, in such situations , a X6 would trounce even a 2600K if it was
not for the softs primarly optimised for Intel s processors that opportunly
always have one if not two generations advance in SSE instructions.
.

Hopefully this will change with BD , and surely that many improvements
when compared to a X6 will not come from architectural changes , but
rather from its ability to execute all existing and even yet to come new instructions.

 

[Idontcare]

In fact, in such situations , a X6 would trounce even a 2600K if it was
not for the softs primarly optimised for Intel s processors that opportunly
always have one if not two generations advance in SSE instructions.
.

For the absolutely newest release software I can see there being some legitimacy to the argument that software is "optimized" to the hardware...but the fact that the CPU does so well in apps that were compiled before the release of Sandy Bridge I would argue that this is actually proof that Intel optimized Sandy Bridge for existing software.

Cause and effect, which is the cart and which is the horse?

Did Intel profile hundreds and thousands of existing apps to then create hardware that was better optimized to process those apps (with an eye towards improving performance of future apps as well, i.e. AVX) or did they create hardware that only looks good because of its performance on new apps?

 

[sm625]

I'm sure if you profile the 100 most commonly used apps and optimize your uarch for those apps, then it will perform better for all apps. It makes me wonder how much better Intel's profiling resources are compared to AMD's. I'm thinking it is like 100:1.

 

[Phynaz]

Hopefully this will change with BD , and surely that many improvements
when compared to a X6 will not come from architectural changes , but
rather from its ability to execute all existing and even yet to come new instructions.

Seriously?

Bulldozer can execute instructions that haven't been created yet?

 

[Idontcare]

Seriously?

Bulldozer can execute instructions that haven't been created yet?

Dude, haven't you been paying attention lately?

Bulldozer is the Chuck Norris of CPU's.

It is so awesome that it designed itself, traveled to the future to determine what instruction set it needed, then went back in time to kick the P4's ass just for sh!ts and giggles, then when it came back to the future it realized that with all its extra site-seeing (what kind of self-respecting Chuck Norris CPU wouldn't take the time to go back in time to see the original release of the K7 Athlon?) that it missed its own launch date by a few months. (but it is so awesome that it doesn't even care, it will kick ass all the same whenever we are graced with its presence, yeah, its simply that awesome...Chuck Norris awesome!)

 

[996GT2]

Without AVX , SB is good in INTEGER or in FP but not in both
simultaneously given that it has no more than 3 exe ports shared
by boths INT and FP instructions.

In fact, in such situations , a X6 would trounce even a 2600K if it was
not for the softs primarly optimised for Intel s processors that opportunly
always have one if not two generations advance in SSE instructions.
.

Hopefully this will change with BD , and surely that many improvements
when compared to a X6 will not come from architectural changes , but
rather from its ability to execute all existing and even yet to come new instructions.

Strong troll