SemiAccurate Forums  

 
Go Back   SemiAccurate Forums > Search Forums

Showing results 1 to 25 of 333
Search took 0.01 seconds.
Search: Posts Made By: gruffi
Forum: CPUs 07-03-2014, 08:00 AM
Replies: 540
Views: 48,099
Posted By gruffi
You wish. That's not an analysis. That's wishful...

You wish. That's not an analysis. That's wishful thinking without any facts.

It rather looks like Excavator will be another major improvement of Bulldozer. It better should if AMD want to make a...
Forum: CPUs 06-24-2014, 06:02 AM
Replies: 16
Views: 1,846
Posted By gruffi
Factor 25 until 2020 would mean an improvement of...

Factor 25 until 2020 would mean an improvement of ~71% every year. That's a lot, but not impossible.
Forum: CPUs 06-21-2014, 05:59 PM
Replies: 1,786
Views: 632,850
Posted By gruffi
A 65W mobile processor? Not gonna happen. ...

A 65W mobile processor? Not gonna happen.



Not gonna happen as well. Intel seems to have problems with their 14nm process. And their next chips don't seem to be much of an improvement over...
Forum: Article discussion 06-09-2014, 02:24 PM
Replies: 28
Views: 4,107
Posted By gruffi
I already have one and didn't need Intel for it....

I already have one and didn't need Intel for it. It's called solar watch. ;)
Forum: CPUs 06-02-2014, 05:45 PM
Replies: 101
Views: 12,943
Posted By gruffi
Sure. But keep in mind, that's still not a...

Sure. But keep in mind, that's still not a precise measurement for the processor only.


You should always take power consumption measurements with a grain of salt, especially computerbase.de....
Forum: CPUs 06-01-2014, 08:45 AM
Replies: 400
Views: 49,576
Posted By gruffi
I didn't miss it. It's just that you talked about...

I didn't miss it. It's just that you talked about chances that are irrelevant in practice because of single thread dependencies. There is an IPC wall for singlethreaded workloads. At that point more...
Forum: CPUs 05-31-2014, 08:06 AM
Replies: 400
Views: 49,576
Posted By gruffi
But resources were not really doubled. K10 has a...

But resources were not really doubled. K10 has a 3-way pipeline. Bulldozer is only 4-way, not 6-way. K10 has 64 KB L1I$ and 64 KB L1D$ per core. Bulldozer has 64 KB L1I$ (Steamroller: 96 KB L1I$) and...
Forum: CPUs 05-30-2014, 09:12 PM
Replies: 400
Views: 49,576
Posted By gruffi
No, it's pretty good in both. That has...

No, it's pretty good in both.


That has nothing to do with Bulldozer. CMT for such small synthesized cores is too complicated and has too few advantages. No one risks a significantly longer and...
Forum: CPUs 05-30-2014, 10:35 AM
Replies: 400
Views: 49,576
Posted By gruffi
No, it's actually pretty good in both. It's...

No, it's actually pretty good in both. It's Glofo's 32nm process and AMD's caches that are not good enough.


Nothing was made clear. We don't know anything about the architecture yet. There is...
Forum: CPUs 05-30-2014, 05:54 AM
Replies: 400
Views: 49,576
Posted By gruffi
No. You were comparing dual modules to quad...

No. You were comparing dual modules to quad cores. And that makes absolutely no sense. A10 with two modules is no quad core processor. Please don't repeat AMD's marketing nonsense. AMD's CMT-based...
Forum: CPUs 05-29-2014, 04:42 PM
Replies: 400
Views: 49,576
Posted By gruffi
No, you can't because there is no core in a...

No, you can't because there is no core in a traditional sense. Bulldozer's Flex FPU works differently. You can't say half of the FPU is dedicated to a "core". A "core" can also use the whole FPU. The...
Forum: Article discussion 05-29-2014, 11:13 AM
Replies: 28
Views: 4,107
Posted By gruffi
And Larrabee. :D

And Larrabee. :D
Forum: CPUs 05-29-2014, 06:47 AM
Replies: 400
Views: 49,576
Posted By gruffi
And that's wrong. You can't count per "core"...

And that's wrong. You can't count per "core" because there is no common basis for a core.

Bulldozer (w/o AVX): 8 FLOPs per module x 4 modules = 32 FLOPs (SP)
Sandy/Ivy Bridge (w/o AVX): 8 FLOPs...
Forum: CPUs 05-28-2014, 04:55 PM
Replies: 400
Views: 49,576
Posted By gruffi
No, the claim is true. Obviously you can't...

No, the claim is true. Obviously you can't interpret deltas the right way. The equation "bad + good = good" must not be true. It still can be "bad + good = bad". It's a question of weighting of each...
Forum: CPUs 05-28-2014, 02:03 PM
Replies: 400
Views: 49,576
Posted By gruffi
You mean on par with Haswell. AMD already is on...

You mean on par with Haswell. AMD already is on par with Sandy, and more. It's really odd that people still don't seem to understand that i5 is not the counterpart of Kaveri. i5 is four pipelines...
Forum: CPUs 05-24-2014, 08:12 AM
Replies: 213
Views: 28,520
Posted By gruffi
R9 295X? First Next-Gen GPU a.k.a.Tonga? HBM? ...

R9 295X?
First Next-Gen GPU a.k.a.Tonga? HBM?
Carizzo?

Or my wish, 16-core K12 prototype which is 4 times as fast as i7-4790 at 1/4 the power consumption. :D
Forum: CPUs 05-20-2014, 06:15 PM
Replies: 400
Views: 49,576
Posted By gruffi
Doesn't matter. ICC still has very poor...

Doesn't matter. ICC still has very poor optimizations for Non-Intel processors. I don't want to blame Intel for it. They are free to do with their compiler whatever they want. But people should...
Forum: CPUs 05-19-2014, 10:53 AM
Replies: 400
Views: 49,576
Posted By gruffi
As I said before, Core2, Stars/Husky and...

As I said before, Core2, Stars/Husky and Steamroller are all very close in terms of singlethreaded IPC. Cinebench says nothing. It's highly Intel optimized due to ICC. Btw, you compare IPC per...
Forum: CPUs 05-15-2014, 08:23 AM
Replies: 400
Views: 49,576
Posted By gruffi
I don't think so. FSB isn't much of a bottleneck...

I don't think so. FSB isn't much of a bottleneck in typical desktop applications. Especially with the large L2 Merom/Conroe and Penryn/Wolfdale had. Finally it's irrelevant. There is no Core2 without...
Forum: CPUs 05-14-2014, 01:29 PM
Replies: 400
Views: 49,576
Posted By gruffi
I don't see any sense in this statement. If it...

I don't see any sense in this statement. If it has better IPC than Steamroller then it has better IPC than Core2. If it will be very close to Core 2 then it will be very close to Steamroller....
Forum: CPUs 05-13-2014, 07:53 AM
Replies: 400
Views: 49,576
Posted By gruffi
It's 16 FLOPs per module (SP), which means 64...

It's 16 FLOPs per module (SP), which means 64 FLOPs per CPU (SP). The same number as Ivy/Sandy Bridge. But Ivy/Sandy Bridge can't do FMA. One AMD FMA is faster than one Intel MUL + ADD. 6 cycle...
Forum: CPUs 05-11-2014, 10:16 AM
Replies: 400
Views: 49,576
Posted By gruffi
Like Intel don't use SMT for their OoO Atoms? ;) ...

Like Intel don't use SMT for their OoO Atoms? ;)

I think it hasn't to do with CMT itself. CMT is fine. But the Cat cores are small and mostly synthesized. It's probably too complicated and takes...
Forum: GPUs 01-26-2014, 07:19 AM
Replies: 189
Views: 23,816
Posted By gruffi
Nvidia? No way. :D

Nvidia? No way. :D
Forum: CPUs 01-13-2014, 12:32 AM
Replies: 3,536
Views: 492,542
Posted By gruffi
IPC should be above Phenom 2 while still having a...

IPC should be above Phenom 2 while still having a 20-25% clock advantage.
Forum: CPUs 01-12-2014, 01:44 PM
Replies: 538
Views: 69,281
Posted By gruffi
Kaveri's L2 is quite large and fast. At least...

Kaveri's L2 is quite large and fast. At least much faster than Deneb's L3. A significant improvement over Trinity/Richland.
Showing results 1 to 25 of 333

 
Forum Jump

All times are GMT -5. The time now is 08:10 PM.


Powered by vBulletin® Version 3.8.7
Copyright ©2000 - 2014, vBulletin Solutions, Inc.
SemiAccurate is a division of Stone Arch Networking Services, Inc. Copyright 2009 Stone Arch Networking Services, Inc, all rights reserved.