First bulldozer benches from AMD - surprising results (beats 980x)

Page 3 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Idontcare

Elite Member
Oct 10, 1999
21,110
59
91
So AMD has an inferior architecture yet it performs better when an application is actually written for the future. So what you're really saying is that AMD is ahead of it's time and that poor programming results in favor of Intel CPU's?

The existence of poor programming (or compilers as it were) is what Intel's hyperthreading attempts to capitalize on.

Hyperthreading is sort of a litmus test for code being overly optimized for Intel architecture.

If hyperthreading actually causes a loss in performance (as is the case with LinX/IBT) then that is a good indication that the code has been optimized for the chip. If hyperthreading improves performance then that is indicative of some rather poorly optimized code.
 

Abwx

Lifer
Apr 2, 2011
11,515
4,301
136
The problem is that DiRT 3 is a completely multi-threaded game that favors AMD's architecture, hence why you see better results with their CPUs even if they're inferior. Look at the Phenom II X4 980 vs Core i5 2500(K) in the game and you'll see what I mean.

Not only does it favor AMD's inferior architecture, but it also takes advantage of six threads, hence why the Phenom II X6 and this do so well in it.
.

All this is completely contradicted by the bench you posted above....
 

LOL_Wut_Axel

Diamond Member
Mar 26, 2011
4,310
8
81
So AMD has an inferior architecture yet it performs better when an application is actually written for the future. So what you're really saying is that AMD is ahead of it's time and that poor programming results in favor of Intel CPU's?

Do you actually understand what you're saying? Having more cores does not equal having a better architecture. How is simply adding more cores to a slow architecture being "ahead of your time"?

Look at the image above. The Core i5-2500K is clearly faster than the Phenom II X4 980, yet loses against it in DiRT 3.
 

LOL_Wut_Axel

Diamond Member
Mar 26, 2011
4,310
8
81
All this is completely contradicted by the bench you posted above....

Not sure if serious.

The Phenom II X6 1100T has a clock speed that's 400MHz lower than the X4 980, yet it's faster. The game is taking advantage of those two additional cores.

The Core i7 920 and i5 750 are faster in gaming than the X4 and X6, so that seems in line, but the i5-2500K performs noticeably worse than the i5 750, which should not be the case.
 

Zstream

Diamond Member
Oct 24, 2005
3,395
277
136
Do you actually understand what you're saying? Having more cores does not equal having a better architecture. How is simply adding more cores to a slow architecture being "ahead of your time"?

Look at the image above. The Core i5-2500K is clearly faster than the Phenom II X4 980, yet loses against it in DiRT 3.

Yes, I understand. What you're not understanding is that poor programming equals a positive for Intel or as IDC pointed out (hyperthreading). Now, I think it's a good move to help the programmers but how you easily dismiss the architecture for AMD is mind boggling.

Even with all the optimizations in the world, the Phenom II X4 should NEVER beat the i5 2500k. So, this tells me, if a program is properly written for a given architecture it can flourish.
 

Abwx

Lifer
Apr 2, 2011
11,515
4,301
136
Not sure if serious.

The Phenom II X6 1100T has a clock speed that's 400MHz lower than the X4 980, yet it's faster. The game is taking advantage of those two additional cores.

1100T has turbo up to 3.7ghz , same frequency as a 980.....
 

Tuna-Fish

Golden Member
Mar 4, 2011
1,474
1,964
136
Looks like AMD still needs to figure out what's causing the low Cinebench scores and solve it before the retail chips ship for revenue.

Media encoders often make much use of sse shuffle instructions. There is only one pipe capable of shuffle in the BD FPU, so any benchmark that is heavily dependent on shuffle should run at the same speed as phenom 2 does, clock for clock and core for core. So running only shuffles, BD should lose to thuban by 33%.

This is probably not fixable without major design changes. On the other hand, fixing it is probably unnecessary. Very few loads suffer from this, and the ones that do can probably be recompiled to use amd's 'permute' instruction (from XOP), which is somewhat more generic and powerful than the normal sse shuffles. Still, any program that doesn't get a recompile, will not benefit from that. And that means all commodity video encoding software that people already have.
 

LOL_Wut_Axel

Diamond Member
Mar 26, 2011
4,310
8
81
Yes, I understand. What you're not understanding is that poor programming equals a positive for Intel or as IDC pointed out (hyperthreading). Now, I think it's a good move to help the programmers but how you easily dismiss the architecture for AMD is mind boggling.

I'm not dismissing anything, I'm just stating a fact. AMD is inferior to Intel on CPU architectures. Bulldozer, by design, has an inferior architecture to Sandy Bridge.

Even with all the optimizations in the world, the Phenom II X4 should NEVER beat the i5 2500k. So, this tells me, if a program is properly written for a given architecture it can flourish.

Yes, that's true. But given the fact that, like I said before, the 1100T has an improvement over the 980 even with a 400MHz lower clock speed, and AMD was working with Codemasters to optimize. That means the game takes proper advantage of AMD's CPUs. It also takes good advantage of Nehalem. What is completely unoptimized is Sandy Bridge, however.

^^
 
Last edited:

LOL_Wut_Axel

Diamond Member
Mar 26, 2011
4,310
8
81
1100T has turbo up to 3.7ghz , same frequency as a 980.....

Yes, but the game, as stated before, is multi-threaded. Therefore, no boost. Even if it weren't, it's clearly nitpicking. The game places Sandy Bridge BELOW Nehalem and Phenom II in performance.
 

drizek

Golden Member
Jul 7, 2005
1,410
0
71
BlueBlazer, wouldn't we expect the 2600K to be higher performing based on price?

The 2600K is nearly a year old. People just starting to look at getting a new system are already going to be considering waiting for Ivy Bridge. You should assume that the price of a 2600K is more like $250 rather than whatever the MSRP is.
 

blackened23

Diamond Member
Jul 26, 2011
8,548
2
0
You Intel fannys just need to stay out of BD threads. I can't believe there are people here that does everything they can to show that BD is a failure. Its only a stinking cpu people, not your mother or anything. Intel doesn't even like you anyways. :p

Yeah, its pretty ridiculous. Its almost as if they dont want intel to have competition -- i'm pretty hopeful that BD turns out to be good. I'm certainly not going to foam at the mouth and try to trash it constantly, i'll wait and see.
 

Atreidin

Senior member
Mar 31, 2011
464
27
86
There sure is a lot of conjecture being stated as "fact". There is no meaningful data to say for sure one way or the other what processor is better. It is also ridiculous to assume that price is an indicator for performance, especially when AMD has in the past sold better processors for less money.

For the people who keep saying over and over again, as fact, that bulldozer is slow and awful, I'll just say that there is a chance you are correct. Maybe even a very good chance. But if you are right, you are just lucky, because you know NOTHING for sure.
 

blackened23

Diamond Member
Jul 26, 2011
8,548
2
0
Should that be...



:thumbsdown:

What i'd like to see is if anyone has proof that Dirt 3 is in any way, shape, or form optimized for AMD. What we know is that it bears the "nvidia - the way its meant to be played monkier" and supports specific nvidia features.

Where's the proof that dirt 3 favors amd? So they had promotions involving dirt 3. So what? Nvidia has promotions with Batman: AA and Batman sure as hell isn't optimized for nvidia (to my knowledge). Does anyone have specific proof that Dirt 3 is coded with specific amd features in mind? I seriously doubt it.
 

LOL_Wut_Axel

Diamond Member
Mar 26, 2011
4,310
8
81
Should that be...



:thumbsdown:

I swear, do people here even read? A game SHOULD NOT place a Phenom II X4 980 being faster than a Core i5-2500K, nor a Core i5-750 being faster.

AMD worked close with Codemasters and the game has been optimized for it. They optimized it for Nehalem, too, but left Sandy Bridge completely unoptimized. Out of all, though, it's made to take every advantage it can out of Phenom II, followed by Nehalem.

It's easy to compare when you have a game that you gave the developers the incentive to make it run better on your CPUs. I want to see how it does in a normal game.
 

inf64

Diamond Member
Mar 11, 2011
3,863
4,539
136
Truth be told,anything that is optimized for Nehalem will run the same(read:not worse) on SB. SB is just like a super charged Nehalem.It has up to 10% better IPC,on average, and has higher clock. So something is wrong in the test suite and I suspect maybe SB's Turbo was the culprit.
 

LOL_Wut_Axel

Diamond Member
Mar 26, 2011
4,310
8
81
What i'd like to see is if anyone has proof that Dirt 3 is in any way, shape, or form optimized for AMD. What we know is that it bears the "nvidia - the way its meant to be played monkier" and supports specific nvidia features.

Where's the proof that dirt 3 favors amd? So they had promotions involving dirt 3. So what? Nvidia has promotions with Batman: AA and Batman sure as hell isn't optimized for nvidia (to my knowledge). Does anyone have specific proof that Dirt 3 is coded with specific amd features in mind? I seriously doubt it.

CPU2.png
 

drizek

Golden Member
Jul 7, 2005
1,410
0
71
It isn't. It's just a game with good DX 11 features that AMD uses to show that they have better DX 11 support than Intel and, to a lesser degree, nvidia.
 

blackened23

Diamond Member
Jul 26, 2011
8,548
2
0
Still no proof that it was coded with anything specific to AMD. Still a lot of conjecture with no evidence. Just because it bears the amd gaming evolved or nvidia - the way its meant to be played monkier means nothing. Its just a marketing tool to sell cards, period. :)

There are tons of games with the nvidia monkier that had no features specific to nvda, same for amd.
 
Last edited:

LOL_Wut_Axel

Diamond Member
Mar 26, 2011
4,310
8
81
There sure is a lot of conjecture being stated as "fact". There is no meaningful data to say for sure one way or the other what processor is better. It is also ridiculous to assume that price is an indicator for performance, especially when AMD has in the past sold better processors for less money.

For the people who keep saying over and over again, as fact, that bulldozer is slow and awful, I'll just say that there is a chance you are correct. Maybe even a very good chance. But if you are right, you are just lucky, because you know NOTHING for sure.

Umm... when? Even during the Pentium 4 area, they were both quite comparable in price, with AMD being better in some areas and Intel in others. AMD MASSIVELY spanked Intel in IPC, but Intel had extremely high clock speeds to counter with. In efficiency AMD was better because of the high-IPC, low clock speed design.

And we already know A LOT about Bulldozer at this stage. We know about the module design, we know very approximate prices, we know the clock speeds, and we know how many cores it has. All these things combined mean that it's still at a huge disadvantage compared to Intel when it comes to IPC, hence they needed more clock speed and lower priced to counter the fact that the architecture has lower IPC and when running two threads on a module you lose performance.

Again, simple maths:

AMD FX-8120: ~$230, 3.1-4.0GHz, Eight-Core.
Intel Core i5-2500K: $220, 3.3-3.7GHz, Quad-Core.

Do you think AMD is just out to give us much higher performance for free? There's a reason they need twice the cores as Intel in the same price range...