BRIGHT SIDE OF NEWS About | Advertise | Contact BSN USER Login
| Register 		SUBSCRIBE Newsletter | RSS Feeds
Friday, May 24, 2013
Email this to a friend.
Your friend's e-mail:
Your Name:
Your e-mail:
Message subject:
Comments
TS Raja by: Anonymous on 12/7/2009
If motherboards are "soul" of computers,"Asus is God".
by: Anonymous on 12/7/2009

NVIDIA GeForce GTX 360 :
- 3.0 Billion Transistors on TSMC 40nm Process.
- 12 Streaming Multiprocessors (SM).
- Each SM has 2x16-wide groups of Scalar ALUs (IEEE754-2008, FP32 and FP64 FMA).
- The 12 SMs Have 512KB Shared L2 Cache.
- 384 Stream Processors (1-way Scalar ALUs) at 1500 MHz.
- 384 ALUs Total.
- 384 FP32 FMA Ops/Clock.
- 192 FP64 FMA Ops/Clock.
- SP (FP32) FMA Rate : 1.152 Tflops.
- DP (FP64) FMA Rate : 576 Gflops.
- 96 Texture Address Units (TA).
- 96 Texture Filtering Units (TF).
- INT8 Bilinear Texel Rate : 62.4 Gtexels/s
- FP16 Bilinear Texel Rate : 31.2 Gtexels/s
- 40 Raster Operation Units (ROPs).
- ROP Rate : 26 Gpixels
- 650 MHz Core.
- 320 bit Memory Subsystem.
- 4200 MHz Memory Clock.
- 168 GB/s Memory Bandwidth.
- 1280MB GDDR5 Memory.
January 2010 release date. Price : 379 USD.
In terms of performance GTX 360 will sit between HD 5870 and dual GPU HD 5970.
by: Anonymous on 12/7/2009
NVIDIA GT300/GF100 "Fermi" Architecture

NVIDIA GeForce GTX 380 :
- 3.0 Billion Transistors on TSMC 40nm Process.
- 16 Streaming Multiprocessors (SM).
- Each SM has 2x16-wide groups of Scalar ALUs (IEEE754-2008, FP32 and FP64 FMA).
- The 16 SMs Have 768KB Shared L2 Cache.
- 512 Stream Processors (1-way Scalar ALUs) at 1700MHz.
- 512 ALUs Total.
- 512 FP32 FMA Ops/Clock.
- 256 FP64 FMA Ops/Clock.
- SP (FP32) FMA Rate : 1.74 Tflops.
- DP (FP64) FMA Rate : 870 Gflops.
- 128 Texture Address Units (TA).
- 128 Texture Filtering Units (TF).
- INT8 Bilinear Texel Rate : 83.2 Gtexels/s
- FP16 Bilinear Texel Rate : 41.6 Gtexels/s
- 48 Raster Operation Units (ROPs).
- ROP Rate : 31.2 Gpixels
- 650MHz Core.
- 384 bit Memory Subsystem.
- 4200 MHz Memory Clock.
- 201.6 GB/s Memory Bandwidth.
- 1536MB GDDR5 Memory.
January 2010 release date. Price : 499 USD
GeForce GTX 380 will be 15% faster than dual GPU Radeon HD 5970.
wow by: Anonymous on 12/7/2009
wow
3 Days by: Anonymous on 12/7/2009
Wow, what a change (12/2 -12/5) 3 days makes. First you are falling all over yourself about 1TF and the next article is stamped "failed".

For all those saying Intel should buy Nvidia, please post what you are smoking. It must be good. If you think the FTC would let the #1 graphics manufacturer (Intel) buy the #2 (Nvidia) you must be under the influence of some pretty heavy narcotics.
by: Anonymous on 12/7/2009
@Anonymous on 12/5/2009

Voxels on Radeon HDs, refer to
http://www.youtube.com/watch?v=Bz7AukqqaDQ
V

Voxels with CryEngine3, refer to http://www.youtube.com/watch?v=1Kvl31g77Z8

From Beyond3D's forum, a developer was able to achieve ~1 TFlops (1000 GFlops) using Radeon HD 4870 for the same SGEMM benchmark(1). Refer to
http://forum.beyond3d.com/showthread.php?t=54842
by: Anonymous on 12/6/2009
Tim Sweeny is describing what Nvidia is already starting to do with it's products now. The G80 was the start in GPU programming flexibility, and the G200 was of course more evolved. Now with Fermi - if it ever gets released - Nvidia is once again raising the bar. By the time Intel has silicon worthy of releasing, how much farther along will Nvidia be? And this doesn't even consider the performance/TDP advantage ATI currently has over both companies.
So basically Intel Concedes by: Anonymous on 12/6/2009
The GPU game is alot tougher than one may think. Looks like that 8.5 billion dollar acquisition of ATI is paying dividends after all. I said it all along, Intel will certainly have to buy Nvidia if they want to get serious about this whole CPU-GPU thing. Any more talk out of them is a complete joke, they are pathetic on the GPU side and deserve to be ridiculed.
Intel Fanbois rallying. by: Anonymous on 12/6/2009
Over at the HardOCP forums this announcement is now a full scale flame war between the GPU fanbois(NV and ATI seem united against a common foe) and Intel fanbois. Gotta love the pointless religious zeal these idiots have.

All Intel is doing is telling everyone what they had suspected all along, that Larrabee isn't a viable product on a commercial level in today's economic environment. Of course they're calling this site unreliable FUD and telling everyone that none of this is true and that Larrabee will be coming "real soon now™"..

I used to love discussing technology in the days before the fanbois invaded many of the tech forums. Now these sites resemble armed religious camps with a take no prisoners attitude.
Now I don't care about tech at all and tell people who ask to buy whatever they can afford, it doesn't matter in the end because a new generation will always be superseding the old.

Larrabee was an experiment, nothing more, Intel made the mistake of presenting it like a real product which would soon dominate the competition. NV's CEO had every right to call it a Powerpoint slide, Intel was practicing more FUD then anyone else. No real product, a lot of promises and worst of all, missing deadlines.
They broke every good marketing rule and therefore have earned the "vaporware" designation with pride.
Larrabee=Duke Nukem Forever in the GPU world.
Larabee didn't cost a billion by: Anonymous on 12/6/2009
Theo writes: "it needed another billion to billion and a half to make it work"

There is zero logic and no facts to back up the above comment:

Intel is putting $7 billion into fab renovation for all their USA locations.

GlobalFoundries is spending $4.3 billion building a new location.

ATIC originally promised to spend $6 billion over five years on GlobalFoundries.

ATIC has recently agreed to acquire Singapore-based Chartered Semiconductor Manufacturing Co. Ltd., for a total of $3.9 billion.

So where does Theo come up with the exaggerated claims of Larrabee costing billions of dollars?
Who like Larrabee? by: Anonymous on 12/5/2009
I'm currently in a discussion with a person who stated this;

"Honnestly gamers dont know what they want. They only know whats good once its out and they tried it out.

Just look at all those stupid remarks they make about raytracing vs rasterising.
95% of them dont understand what they are talking about (even though they are convinced they do).

Devoloppers know what they are talking about and they LOOVE larrabee. Especially since they can do lots of voxels etc on it ;)"

He claims Tim Sweeney likes Larrabee.
When I gave him the above quote he said this;

"yes, but he was talking about the integrated graphics then. Big difference ;)"

So which developer LOOOOVES Larrabee?
Please enlighten me.
He hasn't answered me yet which developer loves Larrabee.
Segue by: Greg442 on 12/5/2009
Are all Intel Engineers Idiots?
xxx by: Anonymous on 12/5/2009
Comment on Cnet. About canning consumer Larrebee by sagrailo

"this may be somewhat of good news to us doing HPC development on GPUs... Namely, it seems like OpenCL is gaining some foot throughout the last time, but it was obvious, through the RapidMind and alike acquisitions, that Intel is intending to push own API for programming Larrabee. So, although it would be great to have more competition in order to further spur the innovation and to have hardware prices to drop, at least there is better chance now that Intel will be forced, once when they eventually deliver this chip, to provide support for API common in use."

That would suck for the Computer industry if it has to cope with an extra API?
That's if the Larrabee had been widely adopted.
i told you theo by: Anonymous on 12/5/2009
I told you, 2 years behind..
© 2009 - 2011 Bright Side Of News*, All rights reserved.
© 2009 - 2013 Bright Side Of News*, All rights reserved.