Bulldozer

Aastii

VIP Member
I don't see a single reason why performance won't increase when Microsoft get their act together and make Windows understand how to use the processor's technology, considering it is proven Windows can't work out when and how to use each module effectively. Once they have fixed it, I want to see bulldozer's true performance
 

jonnyp11

New Member
I mentioned this before somewhere else, but I want to see a bench with the affinity set to use the first 2 modules. Then re-run the bench with the affinity set to use every other "cpu." Wouldn't this mean that it would use 1 thread per module, which is what they are supposed to do to get better performance? From my understanding that would at least boost performance, and possibly offer a glimpse to what the patches will give us.
 

Perkomate

Active Member
Right now with their roadmap, it sounds like they are. I sure hope not.

If they could give us 8 true cores with the similar performance to the new intel chips that would be great.

I'm just going to try to stay optimistic at this point.

Hell, if they even got close to Intel's performance per core, then they could just survive by sticking more cores on a die.
That's how they've done it so far, isn't it?
 

SuperDuperMe

New Member
Do we even have a slight inclination as to when a patch will be out?

I like the idea of an 8 core, but i like performance much more :p, I want bulldozer but i think im going to end up getting thuban :/ So confused!

I just dont know wether to risk it with bulldozer or stay safe with old tech :p
 

BassAddict

New Member
Do we even have a slight inclination as to when a patch will be out?

I like the idea of an 8 core, but i like performance much more :p, I want bulldozer but i think im going to end up getting thuban :/ So confused!

I just dont know wether to risk it with bulldozer or stay safe with old tech :p

There is absolutely nothing wrong with Bulldozer.. It has the potential to outperform the 2600K. Microsoft just needs to get the fix out there.

Why would you go with old tech? Doesn't make any sense to me. This is just a small bump in the road that will be fixed soon.
 

StrangleHold

Moderator
Staff member
There is absolutely nothing wrong with Bulldozer.. It has the potential to outperform the 2600K. Microsoft just needs to get the fix out there.

Why would you go with old tech? Doesn't make any sense to me. This is just a small bump in the road that will be fixed soon.

You have to be joking. The L1 is smaller and slower. The L2 is slower. Since the L1 is smaller it relies more on the L2 which is slower. The core itself is slower in IPC then the Phenom II.

The Patch does nothing to correct the above problem. Even if the patch corrects the problem with scheduling. It will still be slower then the Phenom II with 1 to 6 threads depending if your comparing it to the Phenom II X4 or X6. This thing is a dud until the L1 and L2 is fixed and the IPC is corrected.
 
Last edited:

jonnyp11

New Member
You have to be joking. The L1 is smaller and slower. The L2 is slower. Since the L1 is smaller it relies more on the L2 which is slower. The core itself is slower in IPC then the Phenom II.

The Patch does nothing to correct the above problem. Even if the patch corrects the problem with scheduling. It will still be slower then the Phenom II with 1 to 6 threads depending if your comparing it to the Phenom II X4 or X6. This thing is a dud until the L1 and L2 is fixed and the IPC is corrected.

what are the chances of those being fixed with the 8170's release since i think it may have been you or one of the other guys that said that the 8170 and the other cpus coming out Q1 '12 were supposed to be on their next stepping.
 

StrangleHold

Moderator
Staff member
Dont really see a stepping correcting the L1 or L2 problem. Unless they jump up to a C0/1/2 stepping. But I see Piledriver coming out before then.
 

2048Megabytes

Active Member
Dont really see a stepping correcting the L1 or L2 problem. Unless they jump up to a C0/1/2 stepping. But I see Piledriver coming out before then.

So Zambezi is a failure when compared to Thuban, Deneb and Rana core processors. They did worst on Zambezi series when compared to the Phenom I generation it looks like.

Hopefully AMD will do better on the next series of processors.
 

spynoodle

Active Member
You have to be joking. The L1 is smaller and slower. The L2 is slower. Since the L1 is smaller it relies more on the L2 which is slower. The core itself is slower in IPC then the Phenom II.

The Patch does nothing to correct the above problem. Even if the patch corrects the problem with scheduling. It will still be slower then the Phenom II with 1 to 6 threads depending if your comparing it to the Phenom II X4 or X6. This thing is a dud until the L1 and L2 is fixed and the IPC is corrected.

But if each module runs only 1 thread in a 1-4 threaded program, then it would at least be faster than the Phenom II, wouldn't it? Yes, Bulldozer's not that good, but it's not that bad either.
 

jonnyp11

New Member
not sure what you mean, even on an 8 threaded program the 8150 looses to the 2500k and 2600k a good number of times, sometimes the 1100t i think too. but if you meant what i said before as in making a quad threaded app run on all 4 modules with one thread per module instead of running the 4 threads on 2 modules, then idk, i mentioned this before but nobody commented.
 

2048Megabytes

Active Member
But if each module runs only 1 thread in a 1-4 threaded program, then it would at least be faster than the Phenom II, wouldn't it? Yes, Bulldozer's not that good, but it's not that bad either.

But why would you buy a Zambezi core when Thuban and Deneb cores processors are better? Zambezi are more expensive or just the same cost so I would see no reason to buy a Socket AM3+ processor now.
 

xxmorpheus

Member
This cpu almost certainly requires a fresh windows install. Games are crashing way too much on 1 gpu with temps in normal operating ranges. Even at stock cpu speed my pc is crashing. I think it needs fresh windows install and clean install of chipset drivers. Ill update this weekend
 

jonnyp11

New Member
hey can you do what i was saying, un a bench with affinity on the first 4 threads then on every other thread so it's one hread per module to see the increase/difference.
 

StrangleHold

Moderator
Staff member
So Zambezi is a failure when compared to Thuban, Deneb and Rana core processors. They did worst on Zambezi series when compared to the Phenom I generation it looks like.

Hopefully AMD will do better on the next series of processors.

But if each module runs only 1 thread in a 1-4 threaded program, then it would at least be faster than the Phenom II, wouldn't it? Yes, Bulldozer's not that good, but it's not that bad either.

I think the problem is, they just didnt have enough money to upgrade the architecture beause it really was developed like 4 years ago. If it came out right after the Phenom I on 45nm, it would have been a killer, but big. They did a trade off between IPC and clock speed trying to over take Intel and didnt get enough of either. I say with newer steppings it will get better on IPC and wattage.

If I was AMD I would have kept my mouth shut, came out with a 32nm Phenom III, to cut teeth on 32nm. And released the Bulldozer under the Piledriver name with corrections on the original design.
 
Last edited:

spynoodle

Active Member
But why would you buy a Zambezi core when Thuban and Deneb cores processors are better? Zambezi are more expensive or just the same cost so I would see no reason to buy a Socket AM3+ processor now.

Well, Thuban is definitely faster per thread, but I would venture to bet that an entire module is faster than a Thuban core. Therefore, in a quad-threaded app, Zambezi would win.
 

StrangleHold

Moderator
Staff member
Well, Thuban is definitely faster per thread, but I would venture to bet that an entire module is faster than a Thuban core. Therefore, in a quad-threaded app, Zambezi would win.

Dont really agree. The Thuban core is just flat out faster on IPC then bulldozer even on a single thread. Anything with 6 or less threads the Thuban will win. It has to beat Thuban using clock speed. The big mistake they made with Zambezi was not to let a single thread run on both sets of pipelines on a module. Zambezi is a four year old architecture and shows. If this came out three years ago it would be a win. If I was AMD, I would not put much more effort in Zambezi, maybe put out a B3 stepping to save some face. But dump all the money in Piledriver and release better IPC and wattage processor and let 4 or less threads run on both sets of pipelines on the modules.
 
Last edited:

linkin

VIP Member
Too late now IMO. The damage is already done. These sort of fixes need to be done before release and not after.
 

Okedokey

Well-Known Member
^ this.

The sad thing is that if AMD was able to deliver on its promises we would all have some bloody good talking point ;)

As it stands however, this is a real game changer. A small(ish) company like AMD/ATi cannot afford to have a failed architecture. It costs too much, and with Ivy Bridge just around the corner, I am seriously doubting the impact PD will have too.

Great chip if you run a server, however this was not the intended consumer so it was a major fail.
 
Top