SemiAccurate Forums  

 
Go Back   SemiAccurate Forums > Search Forums
Reload this Page Search Results
Register FAQ Members List Social Groups Calendar Search Today's Posts Mark Forums Read

Page 1 of 9 1 234567 Last »
Showing results 1 to 25 of 221
Search took 0.01 seconds. Search: Posts Made By: gruffi
Forum: Article discussion Today, 08:59 AM
Replies: 8
Views: 366
Posted By gruffi
No, they haven't.

No, they haven't.
Forum: CPUs 03-31-2013, 05:40 AM
Replies: 432
Views: 58,590
Posted By gruffi
I've never questioned that, for the moment. It...

I've never questioned that, for the moment. It might slow down in the future. Anyway, it still doesn't say anything about CPU usage.
Forum: CPUs 03-30-2013, 09:13 AM
Replies: 432
Views: 58,590
Posted By gruffi
Then you should stop quoting me. That was not my...

Then you should stop quoting me. That was not my context.


That's completely irrelevant for my statements unless you compare load/usage. More graphics performance doesn't automatically mean more...
Forum: Misc Technology 03-30-2013, 08:49 AM
Replies: 1,047
Views: 65,308
Posted By gruffi
No. GCN uses vector units for FP. Only integer...

No. GCN uses vector units for FP. Only integer units are scalar.
Forum: CPUs 03-29-2013, 09:47 AM
Replies: 102
Views: 11,491
Posted By gruffi
It depends. Whether you look on the node itself...

It depends. Whether you look on the node itself or the difference between two nodes.

...
65 = new full node
55 = half node of 65
45 = new full node
40 = half node of 40
32 = new full node...
Forum: CPUs 03-29-2013, 09:21 AM
Replies: 432
Views: 58,590
Posted By gruffi
And I thought you will stop here. :rolleyes: Some...

And I thought you will stop here. :rolleyes: Some people seem to be as inconsistent as provocative. High-end and low-end smartphone market is irrelevant. We talked about cores. And, for example, 8...
Forum: CPUs 03-28-2013, 08:14 AM
Replies: 102
Views: 11,491
Posted By gruffi
I don't know what they tried to show. But they...

I don't know what they tried to show. But they are doing 1D scaling. It doesn't work this way. It's about area, which means 2D. From 55 to 40nm is actually a full node with a theoretical reduction of...
Forum: CPUs 03-28-2013, 08:00 AM
Replies: 432
Views: 58,590
Posted By gruffi
No. High-end gaming was and will always be...

No. High-end gaming was and will always be desktops/consoles. You cannot have the same power envelope in a tablet or smartphone. You HAVE TO scale down hardware on SFF.


It doesn't work your way....
Forum: CPUs 03-27-2013, 11:22 AM
Replies: 102
Views: 11,491
Posted By gruffi
You ignore differences beyond pure math, for...

You ignore differences beyond pure math, for example SOI vs bulk. I saw a slide from AMD (or was it Glofo?) some time ago which showed ~40% reduction in size from 32 to 28nm, IIRC. Can't find this...
Forum: CPUs 03-27-2013, 09:25 AM
Replies: 432
Views: 58,590
Posted By gruffi
Likely yes. ;) I thought we talk about...

Likely yes. ;)


I thought we talk about gaming? You can't have high-end gaming hardware in smartphones or tablets by definition. So, it doesn't change anything on my statements. You probably...
Forum: CPUs 03-26-2013, 12:29 PM
Replies: 102
Views: 11,491
Posted By gruffi
AMD uses twice the cores in their server design,...

AMD uses twice the cores in their server design, Trinity (2 CUs) -> Orochi (4 CUs). So, if Kaveri has 3 CUs then the 28nm server design might have up to 6 CUs. As simple as that. ;) 12 CUs for the...
Forum: CPUs 03-26-2013, 09:43 AM
Replies: 432
Views: 58,590
Posted By gruffi
Wouldn't it just prove my point? Your...

Wouldn't it just prove my point?


Your main game loop is typical single-threaded. But that's trivial and doesn't need fast cores. The rest (game logic, transformations, UI, AI, physics, etc) is a...
Forum: Article discussion 03-26-2013, 08:46 AM
Replies: 99
Views: 10,994
Posted By gruffi
5-10% is okay, in line with expectations. I...

5-10% is okay, in line with expectations. I wouldn't call it impressive. Impressive would be at least 20-25%.
Forum: CPUs 03-25-2013, 01:57 PM
Replies: 102
Views: 11,491
Posted By gruffi
Probably a 28nm shrink of Komodo? Anyway, I think...

Probably a 28nm shrink of Komodo? Anyway, I think none of it is still relevant.



I said, there is an official server roadmap. Because you were asking. The rest was just my personal speculation....
Forum: CPUs 03-25-2013, 12:38 PM
Replies: 432
Views: 58,590
Posted By gruffi
So, big.LITTLE cores don't have asynchronous...

So, big.LITTLE cores don't have asynchronous frequencies? Think about it again. That's no argument. It's about waste, not about a useful amount of transistors that can be implemented in any...
Forum: CPUs 03-24-2013, 12:09 PM
Replies: 432
Views: 58,590
Posted By gruffi
No, it doesn't. If you want performance, go...

No, it doesn't. If you want performance, go exclusively for bigger cores or a lot of smaller cores. If you want low power, go exclusively for smaller cores. Big.LITTLE mixes both without achieving...
Forum: CPUs 03-23-2013, 01:57 PM
Replies: 432
Views: 58,590
Posted By gruffi
And that's right for throughput designs. No need...

And that's right for throughput designs. No need to bother with big cores. You should completely rely on small and efficient cores then. The higher number of cores can easily outweigh the loss in...
Forum: CPUs 03-23-2013, 11:47 AM
Replies: 102
Views: 11,491
Posted By gruffi
Because it makes no sense. It was just my...

Because it makes no sense. It was just my personal speculation, based on rumors we know so far. Learn to read and to differentiate!


No, you do. One Bulldozer CU is the same as one Bulldozer...
Forum: CPUs 03-22-2013, 08:22 AM
Replies: 102
Views: 11,491
Posted By gruffi
He was just fooling around. :) Probably he read...

He was just fooling around. :) Probably he read Kaveri and thought I was implying that Kaveri is a server chip. But I never said that. I just said "Considering Kaveri ..." because it's the chip...
Forum: CPUs 03-22-2013, 07:22 AM
Replies: 432
Views: 58,590
Posted By gruffi
No, it isn't. Big.LITTLE is always efficiency,...

No, it isn't. Big.LITTLE is always efficiency, not performance. For example, the silicon you need for 4 small cores you could also use for 1 or 2 of the bigger cores. Which offer more performance in...
Forum: CPUs 03-21-2013, 06:52 AM
Replies: 432
Views: 58,590
Posted By gruffi
I think the article is not that bad. The basic...

I think the article is not that bad. The basic message, performance is not everything. And that's absolutely right! Nvidia is known as a company to fight for the performance crown no matter what it...
Forum: Article discussion 03-21-2013, 06:27 AM
Replies: 99
Views: 10,994
Posted By gruffi
Intel went from 32nm to 22nm FinFET. Lowering TDP...

Intel went from 32nm to 22nm FinFET. Lowering TDP at the same clock isn't that big of a surprise then, right? ;) Haswell increases TDP again.
Forum: CPUs 03-20-2013, 09:25 AM
Replies: 102
Views: 11,491
Posted By gruffi
Because I talked about server CPUs? :rolleyes:

Because I talked about server CPUs? :rolleyes:
Forum: Article discussion 03-20-2013, 09:09 AM
Replies: 99
Views: 10,994
Posted By gruffi
I don't know who brought up this number. But...

I don't know who brought up this number. But personally I ignore synthetic tests and games to get a meaningful impression of CPU performance. So, I see 9 apps which aren't influenced by meaningless...
Forum: CPUs 03-19-2013, 08:13 AM
Replies: 102
Views: 11,491
Posted By gruffi
Considering Kaveri, I would assume it's a...

Considering Kaveri, I would assume it's a Steamroller based CPU with 6+ CUs and a lot of cache. :D Still not specific enough?
Showing results 1 to 25 of 221
Page 1 of 9 1 234567 Last »

 
Forum Jump

All times are GMT -5. The time now is 09:49 PM.

Contact Us - SemiAccurate Forums - Archive - Top

Powered by vBulletin® Version 3.8.7
Copyright ©2000 - 2013, vBulletin Solutions, Inc.
SemiAccurate is a division of Stone Arch Networking Services, Inc. Copyright © 2009 Stone Arch Networking Services, Inc, all rights reserved.