Tesla K20 GPU Compute Processor Specifications Released

[HumanSmoke]

Now we are taking marketing slides as facts

You may be. I'm just pointing out that you can't parse a simple sentence without including personal bias.

In fact you are the one who introduced the Nvidia information

Update:Nvidias marketing slides put DGEMM efficiency of K20 at 80%

So if anyone is treating marketing slides as fact...it's you.

This coming from the ***** who didnt even know who ran GPGPU.ORG

Site reports university research paper. Linked.

Troll trys to sidetrack by highlighting a non-issue to divert attention from the fact that their straw man argument is shown to garbage. Achievement unlocked !!!

 

[Mussels]

trolling wont be tolerated.


any insults past this post will result in immediate infractions.

 

[Recus]

Xzibit, everyone (developer) know Tesla performance since it runs on supercomputers.

 

[Xzibit]

Xzibit, everyone (developer) know Tesla performance since it runs on supercomputers.

It's not about knowing its about reading and not presenting cherry picked information from a employees blog/site to disparrage a competitor.

It only took me 1 minute to recognize who ran the site and a couple minutes to read the research paper.
The paper didnt have anything to do with Nvidia rather AMD Radeon HD 7970 and its effeciency in GEMM. What was the purpose in introducing that other then to point out difference in effeciency between Nvidia and AMD. It make it all the worse being presented from a blog/site from a Nvidia employee.

For what other purpose than to show that AMD Radeon HD 7970 runs at 90% and not 100% would that link be relative. If he wanted to make the efficiency he could have choosen a number of article on Mr. Harris site. After all Nvidia is his focus but he delibrately pointed to a AMD specific article from all the vast information on a Nvidia dedecated blog/site from a Nvidia employee.

I stuck to just the DGEMM efficiency so I provide a quote by a research paper done by people who run these server on the company this thread is based on, Nvidia. I didnt go out of my way to try to disparrage this thread by providing information of its competitor.

He just wasnt intrested in providing information but disparraging the competition which does a dis-service to the forum. The evidence is his responce in this quote below..

*By your reasoning the AMD FirePro W9000 (3.99 TF SP, 1 TF DP) should be four times faster than a Quadro 6000 (1 TF SP, 515 GF DP)...after all, numbers don't lie right?
No...
No...
No

Nothing to do with efficiency in GEMM in anyway. Just baiting.

Thats all...

 

[HumanSmoke]

For what other purpose than to show that AMD Radeon HD 7970 runs at 90% and not 100% would that link be relative

To show that theoretical throughput does not equal real-world performance. Which is why I said:

Any comparison probably depends on actual performance efficiency rather than hypothetical. Unless you know what K20 brings to the table, a theoretical comparison is largely useless.

Note that the statement is vendor agnostic

If he wanted to make the efficiency he could have choosen a number of article on Mr. Harris site. After all Nvidia is his focus but he delibrately pointed to a AMD specific article

Because the person I responding to used AMD as a focal point...

in other words it can almost match tahiti

...and it was the first hit on Google. Sometimes a cigar is just a cigar.

Just for the record, I was the one that introduced DGEMM as an example. There are numerous other examples of real-world vs theoretical throughput- I chose DGEMM because being open source it favours no vendor