Bulldozer ES benchmark is out!

[996GT2]

Super Pi's x87 code is completely outdated, and has been for many years. It shows nothing of modern workloads on modern architectures. It's funny that GPU benchmarks have evolved and have been updated to represent modern architectures (haven't seen any 3D Mark '01 benchmarks for quite a few years). Maybe the GPU market is just that much more progressive, or intel propaganda has seared it into enthusiasts brains that it is a good benchmarking tool since they currently do well in it.

Having said that, these are completely crippled samples, and I fully expect Bulldozer to knock it out of the park in Super Pi but it's irrelevant in modern processing unless you are concerned with legacy x87 code.

Crippled perhaps, but I don't think it's an intentional move on AMD's part.

 

[eelw]

Well hopefully some SB owners will chime in. On my 970 at 4.5GHz

1 instance approx 9.5
2 instance approx 9.6
4 instance approx 10.0
8 instance approx 12.9
12 instance approx 17.5

 

[BLaber]

These benchmarks are done using B0 stepping ES which is all broken , Anything below B1 stepping is not worth looking at.

 

[996GT2]

Well hopefully some SB owners will chime in. On my 970 at 4.5GHz

1 instance approx 9.5
2 instance approx 9.6
4 instance approx 10.0
8 instance approx 12.9
12 instance approx 17.5

In Super Pi?

Here's my 2500K @ 5.0 GHz:

7.519 Seconds for SuperPi 1M

 

[THRiLL KiLL]

SuperPi is a single threaded benchmark, though. It's kind of useless for benching modern CPUs.

i thought bulldozer was to "break down the walls" so single threaded applications would get a large speed boost...

 

[RoyG]

These benchmarks are done using B0 stepping ES which is all broken , Anything below B1 stepping is not worth looking at.

I disagree with you. From what I heard, AMD is looking to bump up the frequency, not IPC.

 

[996GT2]

I disagree with you. From what I heard, AMD is looking to bump up the frequency, not IPC.

Agreed. Plus they're realistically not going to be able to make dramatic increases to IPC since the base architecture is pretty much finalized anyway.

 

[Khato]

The fact that the 1055T (@ 2.8ghz stock) can get ~ 21 seconds, and the fact that INCREASING the clockspeed slowed down the Pi calculation implies that something is wrong.

Or it implies that their attempt at overclocking merely increased the integer core frequency and caused the rest of the module frequency to decrease (either due to limited clock crossing ratios or something else). After all, Super Pi uses x87 FPU instructions. so it could care less how fast the integer cores are running.

 

[BLaber]

I disagree with you. From what I heard, AMD is looking to bump up the frequency, not IPC.

I am ready to bet , when Bulldozer processor launches officially it will be either B1 stepping or B2


And results will be totally different to what you are seeing with that broken B0 stepping being benched

 

[996GT2]

I am ready to bet , when Bulldozer processor launches officially it will be either B1 stepping or B2

And results will be totally different to what you are seeing with that broken B0 stepping being benched

Depends on what you mean by "totally different"

10% performance increase is something I could see happen with the last-minute adjustments, but that's still not enough to compete with Sandy Bridge, much less Ivy Bridge.

Let's not forget that the 2.8 GHz Bulldozer ES still loses to the 2.8 GHz Core i7 860 in Cinebench (a multithreaded benchmark), despite having 8 real "cores" instead of 8 hyperthreaded cores.

And the i7-860 is already an outdated chip. The 2600K is considerably faster. AMD has a lot of work ahead of them to make Bulldozer competitive with Intel's current offerings...

 

[podspi]

Looks like L2/L3 cache writes are disabled...


http://www.youtube.com/watch?v=Z12e7oKPJd8 (switch to HD).

You can also see in some of the pictured benchmarks. If they are legit, cache is disabled. In light of that, the results actually aren't too bad...

 

[Mr. Pedantic]

Just wondering, why doesn't AMD switch to using an LGA for their sockets? Is it just to keep backwards compatibility with AM2/AM3? Or is it cheaper/easier to manufacture, or something?

 

[BLaber]

Depends on what you mean by "totally different"

10% performance increase is something I could see happen with the last-minute adjustments, but that's still not enough to compete with Sandy Bridge, much less Ivy Bridge.

Let's not forget that the 2.8 GHz Bulldozer ES still loses to the 2.8 GHz Core i7 860 in Cinebench (a multithreaded benchmark), despite having 8 real "cores" instead of 8 hyperthreaded cores.

And the i7-860 is already an outdated chip. The 2600K is considerably faster. AMD has a lot of work ahead of them to make Bulldozer competitive with Intel's current offerings...

I would not use Benchmark obtained from a really crippled , broken B0 stepping for setting any base line performance & then anticipate at best 10% performance improvement or more from stepping change. Stepping change can be so many other issues like turbo not working properly , TDP related issues, just not to increase IPC etc.

Also if we can believe this a review done with really good working ES then how could AMD be planning to bring FX back & set price of 320$ for the top 8 core model

 

[996GT2]

I would not use Benchmark obtained from a really crippled , broken B0 stepping for setting any base line performance & then anticipate at best 10% performance improvement or more. Stepping change can be so many other issues like turbo not working properly , TDP related issues, just not to increase IPC etc.

Here's the thing. How much do you realistically think AMD can change in the next 2 months? Yes, it really looks like the chip is crippled and broken, but I don't think the crippling is intentional. It seems like the chip really has a lot that needs to be fixed/improved, but AMD has a very short timeframe in which to do it. The benchmark results right now are horrendous for a chip that's supposed to compete with 2500K, 2600K, and Ivy Bridge when it comes out early next year.

In contrast, Intel's Sandy Bridge ES chips benchmarked pretty close to the final versions, and Anandtech ran those benchmarks 3-4 months before the official release of SB chips (August 2010). Here we are 3 months away from Bulldozer's "official" release and AMD still can't produce engineering samples that can perform half-decently?

 

[notty22]

Here's the thing. How much do you realistically think AMD can change in the next 2 months? Yes, it really looks like the chip is crippled and broken, but I don't think the crippling is intentional. It seems like the chip really has a lot that needs to be fixed/improved, but AMD has a very short timeframe in which to do it. The benchmark results right now are horrendous for a chip that's supposed to compete with 2500K, 2600K, and Ivy Bridge when it comes out early next year.

In contrast, Intel's Sandy Bridge ES chips benchmarked pretty close to the final versions.

I agree. No other way to look at things than depressing.
We are beyond original launch date, and have seen nothing in terms of clock speed or performance that bodes well.
Both things we saw with SB months before launch. Intel showed up with a rig, running 'normal' water cooling running hand break at 5ghz on a ES.

 

[BLaber]

Here's the thing. How much do you realistically think AMD can change in the next 2 months? Yes, it really looks like the chip is crippled and broken, but I don't think the crippling is intentional. It seems like the chip really has a lot that needs to be fixed/improved, but AMD has a very short timeframe in which to do it. The benchmark results right now are horrendous for a chip that's supposed to compete with 2500K, 2600K, and Ivy Bridge when it comes out early next year.

In contrast, Intel's Sandy Bridge ES chips benchmarked pretty close to the final versions, and Anandtech ran those benchmarks 3-4 months before the official release of SB chips (August 2010). Here we are 3 months away from Bulldozer's "official" release and AMD still can't produce engineering samples that can perform half-decently?

You are not aware of B1 stepping Bulldozer ES being in hands of select few for a month or more now & that was even much before computex had started , After computex what is plan that I have no info on.

 

[996GT2]

You are not aware of B1 stepping Bulldozer ES being in hands of select few for a month or more now & that was even much before computex had started , After computex what is plan that I have no info on.

B1 stepping is obviously still flawed or else AMD wouldn't have delayed the launch by another 2-3 months. If the B1 stepping chips could compete with Sandy Bridge now, you can bet AMD would have rushed them out to retailers as soon as they could. But obviously they are flawed in some major way such that they cannot compete with Sandy Bridge.

The $320 price of the highest end FX Bulldozer is another good sign that it's not going to be some magical performer. If AMD had a chip that could blow Intel's Sandy Bridge chips out of the water, they would have priced it accordingly.

Sandy Bridge ES chips 3-4 months before release were performing exactly on par with their final retail counterparts.

See the difference?

 

[Topweasel]

B1 stepping is obviously still flawed or else AMD wouldn't have delayed the launch by another 2-3 months. If the B1 stepping chips could compete with Sandy Bridge now, you can bet AMD would have rushed them out to retails as soon as they could.

Sandy Bridge ES chips 3-4 months before release were performing exactly on par with their final retail counterparts.

See the difference?

A mature platform (not counting the useless socket change) on die shrunk version of mature architecture?

 

[996GT2]

A mature platform (not counting the useless socket change) on die shrunk version of mature architecture?

Sandy Bridge is a completely new microarchitecture, not a die-shrink. Nehalem was the architecture for the first gen Core i3/i5/i5 chips (Bloomfield), and Westmere was the 32nm die-shrink. Next year, Ivy Bridge will be the die-shrink of Sandy Bridge.

This is why it's depressing news that AMD cannot get decently performing ES chips out just a few months before release, when Intel was able to do the same 3-4 months before the final SB release.

 

[Vesku]

B1 stepping is obviously still flawed or else AMD wouldn't have delayed the launch by another 2-3 months. If the B1 stepping chips could compete with Sandy Bridge now, you can bet AMD would have rushed them out to retailers as soon as they could.

Sandy Bridge ES chips 3-4 months before release were performing exactly on par with their final retail counterparts.

See the difference?

Sandy Bridge wasn't a completely new design though. Silly to compare their ES sample performance.

Will be interesting if BD launch is B1 or B2 stepping, not to mention all the info regarding this rocky launch year that will start to flow out once the NDAs start expiring.

 

[996GT2]

Sandy Bridge wasn't a completely new design though. Silly to compare their ES sample performance.

Will be interesting if BD launch is B1 or B2 stepping, not to mention all the info regarding this rocky launch year that will start to flow out once the NDAs start expiring.

Sandy Bridge is based on a completely new microarchitecture. It's not a die-shrink; that was Nehalem-->Wolfdale.

Sandy Bridge's development started in ~2005, which is about the same time development on Bulldozer began as well.

 

[Vesku]

Sandy Bridge is based on a completely new microarchitecture. It's not a die-shrink; that was Nehalem-->Wolfdale.

Sandy Bridge's development started in ~2005, which is about the same time development on Bulldozer began as well.

Rofl, you think the Sandybridge changes are in the same league as what AMD is trying to pull off with Bulldozer? Intel has no reason to take the risks AMD has to take. They at least learned from Phenom 1 to not launch until it is ready. If BD is really bad they will probably push it back again for a full respin rather than try to sell a $320 processor that performs like a $200 Intel processor.

 

[BLaber]

B1 stepping is obviously still flawed or else AMD wouldn't have delayed the launch by another 2-3 months. If the B1 stepping chips could compete with Sandy Bridge now, you can bet AMD would have rushed them out to retailers as soon as they could. But obviously they are flawed in some major way such that they cannot compete with Sandy Bridge.

The $320 price of the highest end FX Bulldozer is another good sign that it's not going to be some magical performer. If AMD had a chip that could blow Intel's Sandy Bridge chips out of the water, they would have priced it accordingly.

Sandy Bridge ES chips 3-4 months before release were performing exactly on par with their final retail counterparts.

See the difference?

Assuming B1 stepping is OBVIOUSLY flawed based on delay in bulldozers release, not really sure, there could be so many other reason other than chip being flawed.

Hope fully Bulldozers release will clear all doubts & assumptions.

 

[Topweasel]

Sandy Bridge is based on a completely new microarchitecture. It's not a die-shrink; that was Nehalem-->Wolfdale.

Sandy Bridge's development started in ~2005, which is about the same time development on Bulldozer began as well.

I am not buying the fact that its new design from the ground up. I guess it depends on how you look at it. Keeping in mind that I don't consider, PP-PII-PIII-PM-CD-C2D, to be considered whole new architectures. Heck I am not a big fan of the K8 being considered new. SB is based off a design of Nehalem architecture. Lots less effort in working within a system you already have. That means base performance expectation is at least that of the previous unit without any tweaks.

 

[996GT2]

Well hopefully some SB owners will chime in. On my 970 at 4.5GHz

1 instance approx 9.5
2 instance approx 9.6
4 instance approx 10.0
8 instance approx 12.9
12 instance approx 17.5

As requested, 4 simultaneous instances on my 2500K @ 4.5 GHz:

Around 8.6 seconds.

(The screenshot shows 2 GHz but that's because Speedstep had already kicked in by the time I got the screenshot)