Overclock.net banner

1 - 20 of 37 Posts

·
Registered
Joined
·
260 Posts
Discussion Starter #1
Source

this is absolutely not confirmed,it seems the page has been taken down.look at the first search and read it yourself.

it seems cloud gaming is the "big gaming news"
 

·
Banned
Joined
·
3,607 Posts
Quote:
Originally Posted by Alatar View Post

Double GK104 Tesla cards? What?
blinksmiley.gif
Seems odd, but I don't really see why not. Generally dual GPU cards have Nvidia's flagship GPU on them and create a lot of heat and use a lot of power... this doesn't. Plus, you don't have to worry about microstutter or anything with these... it's a compute card. I don't see why the industry wouldn't adopt this if it's priced right.

To me, a GK110 announcement seemed too early. I guess they kind of announced it by saying a DP card would be out in the fall, but that's in line with everything else we've heard anyways.

No wonder why Nvidia can't keep GTX 680 in stock though... they've got their chips spread out over so many products...
 

·
Registered
Joined
·
73 Posts
Not to our folding pockets. This is trully for companies who can afford it. I'll stick to 500 series as long it can compute better than the 600 series equivalents.
I do have a GTX 560, so I guess I can say NVIDIA R&D decieved my crunching expectations, and probably will take years before I buy NVIDIA again, maybe just some GT440 or so to populate my rig's remaining slots.
 

·
Premium Member
Joined
·
65,162 Posts
Quote:
Originally Posted by Homeles View Post

Seems odd, but I don't really see why not. Generally dual GPU cards have Nvidia's flagship GPU on them and create a lot of heat and use a lot of power... this doesn't. Plus, you don't have to worry about microstutter or anything with these... it's a compute card. I don't see why the industry wouldn't adopt this if it's priced right.
To me, a GK110 announcement seemed too early. I guess they kind of announced it by saying a DP card would be out in the fall, but that's in line with everything else we've heard anyways.
No wonder why Nvidia can't keep GTX 680 in stock though... they've got their chips spread out over so many products...
GPGPU generally require double precision floating point performance. The GK104 have 1/24th the DP performance as SP....
 

·
Banned
Joined
·
3,607 Posts
Quote:
Originally Posted by DuckieHo View Post

GPGPU generally require double precision floating point performance. The GK104 have 1/24th the DP performance as SP....
Is that hardware level? Or did they just disable some of the FP64 units?
 

·
Registered
Joined
·
1,315 Posts
any more information on the quad GPU card?
Quote:
Separately, Nvidia aims to give a boost to cloud-based gaming services with its new VGX product, a board that packs four GPUs and a 16 Gbyte frame buffer memory. It supports hardware virtualization and a dedicated H.264 video encoder to help overcome network latency that could otherwise slow game data carried over long distances.
if this is using Kepler chips it could also be why they are having stock issues
 

·
Banned
Joined
·
12,758 Posts
It wouldn't make sense to make GK104-based Tesla cards, the GK104 is terrible for compute performance, worse than the GF110, maybe worse than GF104s like the GTX560 even.
 

·
Registered
Joined
·
1,315 Posts
Quote:
Originally Posted by lordikon View Post

It wouldn't make sense to make GK104-based Tesla cards, the GK104 is terrible for compute performance, worse than the GF110, maybe worse than GF104s like the GTX560 even.
unless nVidia hamstrung the geforce parts in compute performance
 

·
Premium Member
Joined
·
65,162 Posts
Quote:
Originally Posted by Homeles View Post

Is that hardware level? Or did they just disable some of the FP64 units?
Hardware...

http://www.anandtech.com/show/5699/nvidia-geforce-gtx-680-review/2
Quote:
The CUDA FP64 block contains 8 special CUDA cores that are not part of the general CUDA core count and are not in any of NVIDIA's diagrams. These CUDA cores can only do and are only used for FP64 math. What's more, the CUDA FP64 block has a very special execution rate: 1/1 FP32. With only 8 CUDA cores in this block it takes NVIDIA 4 cycles to execute a whole warp, but each quarter of the warp is done at full speed as opposed to ½, ¼, or any other fractional speed that previous architectures have operated at. Altogether GK104's FP64 performance is very low at only 1/24 FP32 (1/6 * ¼), but the mere existence of the CUDA FP64 block is quite interesting because it's the very first time we've seen 1/1 FP32 execution speed. Big Kepler may not end up resembling GK104, but if it does then it may be an extremely potent FP64 processor if it's built out of CUDA FP64 blocks.
 

·
Premium Member
Joined
·
65,162 Posts
Why is a "supercomputer" even needed?

They just need to test a lot of configurations and load the different settings in a distributed database...
 

·
Meow!
Joined
·
8,487 Posts
Quote:
Originally Posted by DuckieHo View Post

Why is a "supercomputer" even needed?
They just need to test a lot of configurations and load the different settings in a distributed database...
Cloud computing can emulated different PC configuration, I think it's way cheaper than buying 10 brands of X79, Z77, X58, Z68, P67, 990 980 970 board, 10x 3960X 3930K FX chips, Phenom II Phenom I and etc. Swapping cards in and out.

Let's not forget, some older chip aren't available to acquire anymore. Phenom II stops its production line already.

Time and Money wise I believe.
 

·
9 Cans of Ravioli
Joined
·
20,784 Posts
Quote:
Originally Posted by chuckycheeze View Post

i've always gamed in a cloud...
cigar.gif
thumb.gif
 

·
Premium Member
Joined
·
65,162 Posts
Quote:
Originally Posted by djriful View Post

Cloud computing can emulated different PC configuration, I think it's way cheaper than buying 10 brands of X79, Z77, X58, Z68, P67, 990 980 970 board, 10x 3960X 3930K FX chips, Phenom II Phenom I and etc. Swapping cards in and out.
Let's not forget, some older chip aren't available to acquire anymore. Phenom II stops its production line already.
Time and Money wise I believe.
Trust me... it would be cheaper and faster to buy the physical components.

Emulation is running hardware in software. To do that, you require intimate and often proprietary knowledge of the hardware. Then it would a skilled engineers weeks to years to emulate something with even decent accuracy. Then you have to decide how accurate is accurate enough... then you have the performance overhead...

To emulate a CPU, you are looking to emulate just one recent generation CPU.... you are looking at a few years and $20M+.
 
1 - 20 of 37 Posts
Top