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: GPUs Today, 03:24 PM
Replies: 18
Views: 432
Posted By gruffi
I don't know if that's really true. But only AMD...

I don't know if that's really true. But only AMD can confirm it, not Linux or S/A. And AMD's CodeXL listed Tonga under Volcanic Islands before the whole naming scheme was changed to "Graphics IP...
Forum: GPUs Today, 02:31 PM
Replies: 18
Views: 432
Posted By gruffi
According to the power consumption it should be...

According to the power consumption it should be rather the performance chip, not the mainstream chip



Only ~20W? You think memory + memory interface only needs ~30W in total? I don't think so....
Forum: CPUs Yesterday, 10:29 AM
Replies: 175
Views: 23,042
Posted By gruffi
First, we didn't talk about GPUs. Second, we...

First, we didn't talk about GPUs. Second, we didn't talk about consumers only. Third, I don't care about quarter to quarter. Such a period is too short make any conclusions. And fourth, you are still...
Forum: CPUs 11-24-2014, 08:34 PM
Replies: 175
Views: 23,042
Posted By gruffi
You just wish. The only shrinking business for...

You just wish. The only shrinking business for AMD in the last years are servers. Which isn't a surprise because the tech is outdated. K12 will target this business again. Every other business is...
Forum: CPUs 11-24-2014, 12:42 PM
Replies: 175
Views: 23,042
Posted By gruffi
Even big ARM players have to prove first they can...

Even big ARM players have to prove first they can develop server ecosystems. It's a completely different story than smartphones or tablets. AMD has a lot of expertise in the server world. That will...
Forum: CPUs 11-24-2014, 09:40 AM
Replies: 175
Views: 23,042
Posted By gruffi
Wow, what a failure. A whole 37 GFLOPS. Good we...

Wow, what a failure. A whole 37 GFLOPS. Good we have you to point this out. :rolleyes:


The original Steamroller design was on 22nm which had been canceled. That's why it was late and had less...
Forum: CPUs 11-24-2014, 05:58 AM
Replies: 175
Views: 23,042
Posted By gruffi
Don't be disappointed if Excavator is a...

Don't be disappointed if Excavator is a significant improvement anyway.
Forum: CPUs 11-23-2014, 11:07 AM
Replies: 175
Views: 23,042
Posted By gruffi
I wouldn't bet on that. They really seem to be...

I wouldn't bet on that. They really seem to be enthusiastic about the GPU improvements. I guess that's not possible if it's just Tonga like. Even if delta color compression can improve bandwidth...
Forum: CPUs 11-21-2014, 04:01 PM
Replies: 175
Views: 23,042
Posted By gruffi
Could be. Interesting, John Byrne...

Could be. Interesting, John Byrne (https://www.youtube.com/watch?v=I6zX2IqBI7A) says Carrizo will have a "brand new graphics architecture" (01:22). I guess he wouldn't have said that if it's the same...
Forum: CPUs 11-21-2014, 09:37 AM
Replies: 175
Views: 23,042
Posted By gruffi
I guess it will have a maximum of 4 CUs. So you...

I guess it will have a maximum of 4 CUs. So you won't need much bandwidth. And because it's the same package as Carrizo it will be dual channel. That's still a significant improvement over single...
Forum: CPUs 11-17-2014, 03:20 PM
Replies: 959
Views: 116,709
Posted By gruffi
Btw, AMD's packing density is amazing. Intel...

Btw, AMD's packing density is amazing.

Intel 22nm: 5.56 billion transistors on 663.5 mm˛ = ~8.38 million transistors per mm˛
AMD 28nm: 3.1 billion transistors on 244.62 mm˛ = ~12.67 million...
Forum: CPUs 11-17-2014, 02:43 PM
Replies: 959
Views: 116,709
Posted By gruffi
28nm should be cheap enough. Don't forget, <=22nm...

28nm should be cheap enough. Don't forget, <=22nm and FinFET is a lot more expensive per mm˛.
Forum: CPUs 11-13-2014, 11:11 AM
Replies: 32
Views: 3,360
Posted By gruffi
"Intel's pace of innovation" Best joke of...

"Intel's pace of innovation"

Best joke of the day. Thanks for the laugh.
Forum: CPUs 11-04-2014, 03:39 PM
Replies: 81
Views: 11,700
Posted By gruffi
Not really. That was just a demonstration of the...

Not really. That was just a demonstration of the technology on 32nm.
Forum: CPUs 11-04-2014, 09:59 AM
Replies: 50
Views: 8,325
Posted By gruffi
On AT forum they say Nexus 9 gets quite hot....

On AT forum they say Nexus 9 gets quite hot. Maybe throttling is the reason for stutters and slowdowns?
Forum: CPUs 11-04-2014, 09:57 AM
Replies: 81
Views: 11,700
Posted By gruffi
HDL was said to be used after Steamroller. So,...

HDL was said to be used after Steamroller. So, Excavator is the first core that might be designed with it.
Forum: CPUs 11-04-2014, 09:09 AM
Replies: 81
Views: 11,700
Posted By gruffi
If Carrizo's clocks are lower than Kaveri's...

If Carrizo's clocks are lower than Kaveri's clocks the IPC has to be improved by 50% or so to get 30% better power efficiency. Very unlikely. It has to be a mix of more clock and more IPC at the same...
Forum: CPUs 11-03-2014, 09:02 AM
Replies: 81
Views: 11,700
Posted By gruffi
Please stop quoting this site. It's getting...

Please stop quoting this site. It's getting ridiculous. They are using a lot of games and memory bound apps like 7-Zip or WinRAR. As I said before that's an inappropriate way to measure CPU...
Forum: CPUs 11-02-2014, 08:50 AM
Replies: 81
Views: 11,700
Posted By gruffi
Please ignore games. Games are and never were...

Please ignore games. Games are and never were good indicators of CPU performance. Games are rather an indicator of system performance. And apps like 7-zip or WinRAR are mostly memory bound. Actually...
Forum: CPUs 11-02-2014, 01:18 AM
Replies: 81
Views: 11,700
Posted By gruffi
Didn't you say the same about Steamroller which...

Didn't you say the same about Steamroller which actually increased IPC by ~15%?



True. No one here seems to know any concrete details about Excavator. So we just should wait and see. But I bet...
Forum: CPUs 11-01-2014, 09:58 AM
Replies: 81
Views: 11,700
Posted By gruffi
Where exactly? It says 24.2 GOPS (±6.0 GOPS) at...

Where exactly? It says 24.2 GOPS (±6.0 GOPS) at 2.6 GHz, or 9.51 GOPS per GHz. But yes, there are better and worse scores per GHz for Kaveri engineering samples. So, we shouldn't interpret too much...
Forum: CPUs 10-31-2014, 04:41 PM
Replies: 81
Views: 11,700
Posted By gruffi
I think you can't make any conclusions about...

I think you can't make any conclusions about performance yet. The numbers look odd. Processor arithmetic is too low for 2.6 GHz. Latency can't be that bad. Memory bandwidth seems to be single channel.
Forum: Article discussion 10-27-2014, 05:53 PM
Replies: 43
Views: 5,816
Posted By gruffi
According to your quotes I can't see a reason why...

According to your quotes I can't see a reason why it has to be Samsung. But I have to admit I can't follow your argumentation. It doesn't make any sense to me.

Actually everybody seems to have...
Forum: Article discussion 10-27-2014, 08:31 AM
Replies: 43
Views: 5,816
Posted By gruffi
The article only mentions yield problems with...

The article only mentions yield problems with 20nm, not 14nm. And this is quite odd because Samsung seem to have 20nm only for internal use, not mass production.
Forum: Article discussion 10-24-2014, 08:21 AM
Replies: 12
Views: 1,618
Posted By gruffi
I would be really surprised if it's Microsoft....

I would be really surprised if it's Microsoft. Back in the beginning of the year they were still sceptical (http://www.theregister.co.uk/2014/01/10/microsoft_arm_adoption/) about ARM. Though I think...
Showing results 1 to 25 of 500

 
Forum Jump

All times are GMT -5. The time now is 04:18 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.