SemiAccurate Forums  

 
Go Back   SemiAccurate Forums > Search Forums

Showing results 1 to 25 of 500
Search took 0.02 seconds.
Search: Posts Made By: gruffi
Forum: CPUs 05-22-2015, 05:40 AM
Replies: 116
Views: 18,310
Posted By gruffi
What? Jaguar/Puma(+) is a 2-way OoO low power...

What? Jaguar/Puma(+) is a 2-way OoO low power core, designed for efficiency. Core 2 was a 4-way OoO high performance core. A performance comparison today is okay. But your conclusion is completely...
Forum: GPUs 05-20-2015, 01:44 PM
Replies: 42
Views: 3,141
Posted By gruffi
Becomes? TPU is pro Nvidia since a long time. ...

Becomes? TPU is pro Nvidia since a long time.



It's just 130 days or so. But, who cares? I don't want to update a graphics driver every month. I think longer update intervals are better for...
Forum: CPUs 05-19-2015, 07:18 AM
Replies: 50
Views: 7,787
Posted By gruffi
I didn't ask for user experiences. I wanted to...

I didn't ask for user experiences. I wanted to know what the failure exactly is or what's causing it.



I think you are wrong. Slides about Carrizo say otherwise. Actually the power management...
Forum: GPUs 05-18-2015, 10:27 AM
Replies: 263
Views: 29,742
Posted By gruffi
Interesting that R9 290X is almost 20% faster...

Interesting that R9 290X is almost 20% faster than GTX 780 Ti. Sponsored by Nvidia FutureProof™ technology. :rolleyes:
Forum: CPUs 05-18-2015, 10:14 AM
Replies: 50
Views: 7,787
Posted By gruffi
And the power management failure is what exactly?...

And the power management failure is what exactly?


HDL wasn't used for Kaveri. :rolleyes:
Forum: CPUs 05-17-2015, 04:22 AM
Replies: 413
Views: 27,495
Posted By gruffi
Okay. We summarize, Steamroller's integer...

Okay. We summarize, Steamroller's integer performance is better than Piledriver and Bulldozer, not worse. ;)
Forum: CPUs 05-16-2015, 11:46 AM
Replies: 413
Views: 27,495
Posted By gruffi
Don't include all the compiler flags. Especially...

Don't include all the compiler flags. Especially MMX and AVX are unnecessary. Dhrystone is all about strings. Only SSE(4) might be useful. And why do you include flags like -mfpmath? I thought you...
Forum: CPUs 05-16-2015, 06:47 AM
Replies: 413
Views: 27,495
Posted By gruffi
True. I have corrected your nonsense before. ...

True. I have corrected your nonsense before.


No. It's the same number of physical cores, pipelines and threads.


Okay. But Steamroller is still faster per clock. ~10% on average as AMD...
Forum: CPUs 05-16-2015, 06:38 AM
Replies: 413
Views: 27,495
Posted By gruffi
Then you shouldn't wonder if a core with almost...

Then you shouldn't wonder if a core with almost twice as much logic is also a lot faster. Speaking of "unmitigated disaster" if a much smaller "core" is slower is just clueless trolling.
Forum: CPUs 05-16-2015, 06:34 AM
Replies: 413
Views: 27,495
Posted By gruffi
@Thevenin You are wrong with your testing....

@Thevenin

You are wrong with your testing. You have to exclude all the SIMD stuff. Otherwise you might test the FPU and not just the integer cluster. Don't forget, Steamroller has one less FPU...
Forum: CPUs 05-15-2015, 03:49 PM
Replies: 413
Views: 27,495
Posted By gruffi
Bulldozer architecture: 8 FLOPs per thread ...

Bulldozer architecture: 8 FLOPs per thread
Sandy/Ivy Bridge architecture: 8 FLOPs per thread
Haswell architecture: 16 FLOPs per thread

Bulldozer has the same theoretical FP throughput as...
Forum: CPUs 05-15-2015, 10:03 AM
Replies: 413
Views: 27,495
Posted By gruffi
You talked about integer performance of...

You talked about integer performance of Steamroller compared to Bulldozer/Piledriver. I'm still waiting for any proof. So far we have one integer benchmark that shows ~6% performance advantage for...
Forum: CPUs 05-15-2015, 04:23 AM
Replies: 413
Views: 27,495
Posted By gruffi
Sandy/Ivy Bridge have the same theoretical FP...

Sandy/Ivy Bridge have the same theoretical FP throughput as Bulldozer. Only Haswell is a different ballpark. Bulldozer is faster with proper software support than Sandy/Ivy Bridge. Which means...
Forum: CPUs 05-14-2015, 04:39 PM
Replies: 413
Views: 27,495
Posted By gruffi
Haswell is irrelevant because it is a newer...

Haswell is irrelevant because it is a newer design and has two 256-bit FMA units. If you argue this way then Sandy/Ivy Bridge's FP performance is rubbish as well. With proper FMA support Bulldozer...
Forum: CPUs 05-14-2015, 10:40 AM
Replies: 413
Views: 27,495
Posted By gruffi
Even if there is a frequency hit, I don't think...

Even if there is a frequency hit, I don't think that's the reason. I guess they just wanted to point out that the architecture is much stronger. They probably don't even know final frequencies yet.
...
Forum: CPUs 05-14-2015, 04:37 AM
Replies: 413
Views: 27,495
Posted By gruffi
I would say IPC is very much the same for one...

I would say IPC is very much the same for one thread. The problem for Steamroller is, true cores still scale better.



Sure. :rolleyes:


It was originally >20%. And that are percentage...
Forum: CPUs 05-13-2015, 05:00 PM
Replies: 413
Views: 27,495
Posted By gruffi
It's not my speedup. The numbers are from AMD. ...

It's not my speedup. The numbers are from AMD.


Irrelevant calculation. Alexko already gave you the correct calculation if you want to know your relative error. It's over 50% or over 35%...
Forum: CPUs 05-13-2015, 12:34 PM
Replies: 413
Views: 27,495
Posted By gruffi
But Excavator means the whole module and not just...

But Excavator means the whole module and not just that fake marketing "core". I wouldn't ignore the possibility of two Excavator threads for now. Especially because it came from a marketing guy. Who...
Forum: CPUs 05-13-2015, 06:47 AM
Replies: 413
Views: 27,495
Posted By gruffi
Why am I not surprised that you don't even...

Why am I not surprised that you don't even understand the math behind percentage points. You were at least 20% off. Period.


As I said before, he didn't mention if the Zen core was inclusive SMT....
Forum: CPUs 05-12-2015, 05:00 PM
Replies: 413
Views: 27,495
Posted By gruffi
Your criteria is irrelevant. The terms "big core"...

Your criteria is irrelevant. The terms "big core" and "small core" have a clear meaning. And it's not physical size. Stop being that ignorant!


No. Your "prediction" was at least 20% off. You...
Forum: CPUs 05-12-2015, 04:15 AM
Replies: 413
Views: 27,495
Posted By gruffi
No. Your predication was at least 20% off. ...

No. Your predication was at least 20% off.


Power consumption and power efficiency do not depend on physical size or transistor count only. You ever heard of frequency and voltage? A smaller core...
Forum: CPUs 05-12-2015, 03:55 AM
Replies: 413
Views: 27,495
Posted By gruffi
That's basically what you said and what I said...

That's basically what you said and what I said before. You can't even understand what you say, let alone what others say. :rolleyes:


You still don't understand what big and small cores are....
Forum: CPUs 05-11-2015, 05:51 PM
Replies: 413
Views: 27,495
Posted By gruffi
And will they have 4-way SMT? Probably not. The...

And will they have 4-way SMT? Probably not. The next Xeon Phi is supposed to be based on Silvermont cores. Silvermont has no SMT.


That's basically what I said. Diminishing returns of an already...
Forum: CPUs 05-11-2015, 05:41 PM
Replies: 35
Views: 2,634
Posted By gruffi
Sure. :rolleyes:

Sure. :rolleyes:
Forum: CPUs 05-11-2015, 12:15 PM
Replies: 413
Views: 27,495
Posted By gruffi
We talk about a high performance OoO core. Phi...

We talk about a high performance OoO core. Phi just isn't that. It uses low power in-order cores at the moment. Vulcan is not SMT. It is FGMT. It's just like I said. The more SMT threads, the wider...
Showing results 1 to 25 of 500

 
Forum Jump

All times are GMT -5. The time now is 02:49 AM.


Powered by vBulletin® Version 3.8.7
Copyright ©2000 - 2015, vBulletin Solutions, Inc.
SemiAccurate is a division of Stone Arch Networking Services, Inc. Copyright © 2009 Stone Arch Networking Services, Inc, all rights reserved.