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 Yesterday, 09:32 AM
Replies: 415
Views: 60,338
Posted By gruffi
But this is a Carrizo thread.

But this is a Carrizo thread.
Forum: CPUs 03-28-2015, 07:09 AM
Replies: 415
Views: 60,338
Posted By gruffi
Carrizo? It looks more like the Kaveri Refresh.

Carrizo? It looks more like the Kaveri Refresh.
Forum: GPUs 03-24-2015, 07:24 PM
Replies: 20
Views: 1,640
Posted By gruffi
Irrelevant. You didn't get the point of my...

Irrelevant. You didn't get the point of my analogy, searching for poor excuses now.


No. He didn't mention any specific GPUs.



Both are based on the same architecture.
Forum: GPUs 03-24-2015, 06:05 AM
Replies: 20
Views: 1,640
Posted By gruffi
That's irrelevant for the point I made. Nowadays...

That's irrelevant for the point I made. Nowadays the shader architecture of GPUs is also designed for general workloads. Or what do you think GPGPU means?


You don't need to repeat me.

...
Forum: GPUs 03-23-2015, 08:23 AM
Replies: 20
Views: 1,640
Posted By gruffi
Why do we have an FPU in our CPUs when many apps...

Why do we have an FPU in our CPUs when many apps only use integer code? Just because it's not useful to develop different CPUs for every single type of workload.

It's much easier for companies...
Forum: CPUs 03-23-2015, 07:50 AM
Replies: 76
Views: 11,472
Posted By gruffi
Nothing like that has been stated so far from...

Nothing like that has been stated so far from AMD. But AM1 also wasn't intended for longevity. It's just a low cost / low power platform to offer Cat processors for desktop systems. A short-lived...
Forum: CPUs 03-19-2015, 10:50 AM
Replies: 1,076
Views: 144,857
Posted By gruffi
No. I compare what originally was defined as a...

No. I compare what originally was defined as a core and what a core actually is. You just repeat the same nonsense as the marketing. That's a mistake.


You said something that is wrong. Period.
...
Forum: GPUs 03-19-2015, 10:13 AM
Replies: 25
Views: 1,978
Posted By gruffi
Nvidia lost most professional market share to...

Nvidia lost most professional market share to AMD. One of the most significant wins was the Apple deal. Some sources claim this gives AMD alone ~30% market share. Including Intel, Nvidia's market...
Forum: CPUs 03-18-2015, 08:40 PM
Replies: 1,076
Views: 144,857
Posted By gruffi
Irrelevant if we talk about density at the same...

Irrelevant if we talk about density at the same node.


There is no mistake. I compare core against core.


I don't care what you said yesterday or whenever. You said "Bulldozer is a small...
Forum: GPUs 03-18-2015, 08:08 PM
Replies: 25
Views: 1,978
Posted By gruffi
Which advantage do you mean? Don't forget, we...

Which advantage do you mean? Don't forget, we live in 2015 and not in 2011 anymore.



It's more like 70-30 nowadays. Nvidia is constantly losing GPGPU market share. First because CUDA has not...
Forum: CPUs 03-18-2015, 08:51 AM
Replies: 1,076
Views: 144,857
Posted By gruffi
Stop posting those made up numbers unless you can...

Stop posting those made up numbers unless you can back it up with real numbers. Intel claims a lot. But the truth is, 14nm Broadwell shows only small density advantages compared to 28nm (!) Carrizo....
Forum: GPUs 03-18-2015, 08:16 AM
Replies: 25
Views: 1,978
Posted By gruffi
IBM will not be very happy if Volta is 2018+....

IBM will not be very happy if Volta is 2018+. Another case of backing the wrong horse?
Forum: GPUs 03-18-2015, 08:11 AM
Replies: 211
Views: 21,416
Posted By gruffi
Leaks show otherwise. At 1600p R9 390X might be...

Leaks show otherwise. At 1600p R9 390X might be even some percent faster than Titan X. I guess computing-wise Fiji will trounce everything from Nvidia. Once again.
Forum: GPUs 03-18-2015, 08:07 AM
Replies: 18
Views: 1,442
Posted By gruffi
You don't even need to go that far. 10x from...

You don't even need to go that far. 10x from Nvidia marketing means 2-5x in reality.
Forum: CPUs 03-16-2015, 07:50 PM
Replies: 1,076
Views: 144,857
Posted By gruffi
I don't know where you see any sign here. Intel...

I don't know where you see any sign here. Intel has an 18-core monolithic die on 22nm. So, 16 high performance cores should be no problem on 14nm and a reasonable die size. This would actually mean a...
Forum: CPUs 03-11-2015, 12:11 PM
Replies: 415
Views: 60,338
Posted By gruffi
I don't think so. Carrizo is explicitly...

I don't think so. Carrizo is explicitly mentioned, Tonga isn't. This GCN revision is from March 2015. Tonga is older.
Forum: CPUs 03-05-2015, 04:41 AM
Replies: 1,076
Views: 144,857
Posted By gruffi
Well, it's not that easy as you might think. It...

Well, it's not that easy as you might think. It depends a lot on the platform. The C++ standard just guarantees this:

So, theoretically on a certain platform float can be the same size as double....
Forum: CPUs 03-04-2015, 03:24 AM
Replies: 1,076
Views: 144,857
Posted By gruffi
Also? Double precision floating-point...

Also? Double precision floating-point instructions is the most important part of SSE2. You should say, "it also adds integer instructions". That's more accurate. Most SSE2 integer instructions are...
Forum: CPUs 03-03-2015, 11:47 AM
Replies: 1,076
Views: 144,857
Posted By gruffi
No, it isn't. Historically, it is just one core...

No, it isn't. Historically, it is just one core because there is only one pipeline. That's how the engineers defined it and how it was designed. There aren't also two of everything throughput the...
Forum: Article discussion 03-03-2015, 03:48 AM
Replies: 12
Views: 1,706
Posted By gruffi
So, it seems DirectX 12 and GLnext are just...

So, it seems DirectX 12 and GLnext are just Mantle 1.0 under the hood. While AMD continues Mantle development as a testbed for improvements and new innovations. Not really surprising at all.

Btw,...
Forum: CPUs 03-03-2015, 03:32 AM
Replies: 1,076
Views: 144,857
Posted By gruffi
HTT today is mediocre at best. It's nowhere near...

HTT today is mediocre at best. It's nowhere near the scaling of CMT or CMP. And it was even worse back then. It could in fact decrease performance. I wouldn't call that a good thing.


That's...
Forum: CPUs 03-02-2015, 07:20 AM
Replies: 1,076
Views: 144,857
Posted By gruffi
He never mentioned the word "small" in the...

He never mentioned the word "small" in the context of core size. He wanted to say that they know how to do dense designs (Cat) and how to do high frequency designs (BD). And they will take the best...
Forum: CPUs 03-01-2015, 01:00 AM
Replies: 1,076
Views: 144,857
Posted By gruffi
No. We don't know anything about the size of the...

No. We don't know anything about the size of the core. Please stop spreading this fud. Keller only said it will be a high performance core. So, it's likely not a small core compared to other small...
Forum: Article discussion 02-27-2015, 09:54 PM
Replies: 257
Views: 22,042
Posted By gruffi
I don't think so. They failed once with that very...

I don't think so. They failed once with that very hard (Larrabee). Even their process advantage couldn't save them. CPU for graphics is just too inefficient. So, the "long-term goal" is already...
Forum: CPUs 02-22-2015, 03:06 PM
Replies: 1,076
Views: 144,857
Posted By gruffi
Well, at least you can give credit to Intel for...

Well, at least you can give credit to Intel for widening the pipeline to 256-bit and a new coding scheme. The instructions itself are quite similar, of course. OTOH, Intel64 was indeed just a renamed...
Showing results 1 to 25 of 500

 
Forum Jump

All times are GMT -5. The time now is 07:12 PM.


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.