Sign in with
Sign up | Sign in
Your question
Closed

Tesla C2070 vs GTX 590

Last response: in Overclocking
Share
June 9, 2011 7:57:10 AM

Tesla C2070 vs GTX 590 have the same Fermi archicteture but GTX 590 has 1024 CUDA cores and Tesla C2070 has 448. Does that mean that GTX 590 outperforme C2070 more then two times even in supercompiuting? I mean it is the same archicteture. I am intrested in other opinions.

More about : tesla c2070 gtx 590

a b U Graphics card
a b K Overclocking
June 9, 2011 2:33:13 PM

Those Tesla C2070's are meant more for thing such as CAD or Health science stuff. Basically anything that requires calculations that -isn't- gaming.

Yes you will get more performance for gaming out of the GTX 590 than the C2070.

But it will not be 2x the performance even though it has 2x cuda cores.

If your going for gaming - get the GTX.

If your doing workstation calculations - get the tesla.
a b U Graphics card
a b K Overclocking
June 9, 2011 2:34:30 PM

One more thing. Tesla's have no vide out put FYI.
Related resources
June 9, 2011 3:59:28 PM

Chainzsaw said:
Those Tesla C2070's are meant more for thing such as CAD or Health science stuff. Basically anything that requires calculations that -isn't- gaming.

Yes you will get more performance for gaming out of the GTX 590 than the C2070.

But it will not be 2x the performance even though it has 2x cuda cores.

If your going for gaming - get the GTX.

If your doing workstation calculations - get the tesla.


But is the same archicteture! Tesla only has more GDDR5 memory. And talking about a workstation solution. C2070 has a DVI output. I need more tecnical explenation wich is beter. Your ansver will not suffice.
a b U Graphics card
a b K Overclocking
June 9, 2011 4:13:23 PM

"But is the same archicteture! Tesla only has more GDDR5 memory. And talking about a workstation solution. C2070 has a DVI output. I need more tecnical explenation wich is beter. Your ansver will not suffice. "

Google is your friend. Good luck.
June 9, 2011 6:54:40 PM

Thanks, I am trying but I can't find the answers to my questions. As you see, you are the only how comented on my question.
a b U Graphics card
a b K Overclocking
June 9, 2011 7:20:15 PM

Ok. If i'm going to help you out, you need to give me more information.

Are you researching for a project?

Why do you need the information?

Is this for gaming? Or is this or AutoCad or computational requirements?

What is your budget (if you plan on buying one)?

What is your system specifications (if you plan on buying)?

BTW the original TESLA didn't have video output. The newer versions do have DVI outputs. So in a sense we are both right about the video output.

Both cards may use the same architecture, but are geared for different uses, not to mention the price differences are staggering.
Anonymous
a b U Graphics card
a b K Overclocking
June 15, 2011 3:14:37 AM

One big difference between Tesla and GeForce is that Tesla has enhanced double-precision floating point enabled. It's something in the order of 4x the performance there. It also offers ECC and larger memories (3GB/6GB) than GTX cards.

If you don't need double precision or large memory, you probably don't need Tesla.
August 27, 2011 7:41:21 PM

Dumb question:
How do you know if your app requires double prescision floating accuracy?
Im asked to research on a gpu based cluster as a part of project and deciding the card seem to be a bottleneck.4 GTx 590s or 4 Tesla C2070s
Anonymous
a b U Graphics card
a b K Overclocking
November 17, 2011 1:46:44 AM

if you don't know, then I don't know what to say.

I use gtx480's for scientific computing. I'd much rather have C2070's, and I know why. I don't have them because they are enormously costly, and I can use 32 bit floats. tesla cards have more memory, and various bits and bobbed turned on for management. also the 64bit floats (but it runs at half speed) and error correcting RAM.

what do you mean "your app"? aren't you writing code? don't you have the source code? doesn't it at least say in the docs what hardware it needs?

why are making a gpu cluster??

if you have the money, get the teslas. that's about 12k. the 590's maybe 2.8k

you have to know how to prgram them. how busses, coalescing, caches etc work.

that 1101=13=F1

it's not that simple. you do get what you pay for. but a ferrari is no good to you if don't drive stick.

a b U Graphics card
a b K Overclocking
November 17, 2011 1:55:00 PM

Someone should close this thread...the OP hasn't even responded at all.
December 6, 2011 3:37:37 AM

Just to complete the thread with more tech details,
the core speed is roughly the same on both card but:
with GTX590 the double-precision run at 1/8th of the speed of single-precision float.
while on C2075 the double-precision run at 1/2th of the speed of single-precision float.

GTX590 got 1024 core, in 2 GPU.
DDR5 1707Mhz, Graphic Clock 607Mhz, Processor 1215Mhz

I just want to tell GTX580 would be more near to C2075 based on cores available (with only 1 GPU)
GTX580 got 512 core,
DDR5 2004Mhz, Graphic Clock 772Mhz, Processor 1544Mhz,

Memory bandwidth is very important in case of small data collection to process.

Although GeForce are much cheaper, I agree "chainzsaw", Tesla are for pure computing while GeFore are more Game Industry focused, if you don't need so much memory and single precision is enough for your need GeForce and Tesla will give you same perf.

Personaly we built a cluster of 30 computers with a GTX590 in each node, we got an improvement of 7x faster than our standard PC math library. Data to process are small chunk of 20MB to 50MB.

cheers

Anonymous
a b U Graphics card
a b K Overclocking
March 8, 2012 6:56:40 PM

joemanu said:

Personaly we built a cluster of 30 computers with a GTX590 in each node, we got an improvement of 7x faster than our standard PC math library. Data to process are small chunk of 20MB to 50MB.

cheers


7x faster than what? A cluster of another 30 computers with Teslas? Just asking because the topic is Tesla vs GTX590.

Also, OP, you are 100% right except for one thing,
which some of the elitist "go google it your self" posters failed to tell you.

Drivers. It's all in the software. You're 100% right, that the two are exactly the same FERMi architecture.
Which is why it is confusing to some why the Tesla costs >$2k.
Tesla products cost as much as they do because of three things:
-- The software written for them (it's all thanks to the driver, which is only usable with Tesla cards)
-- nVidia's ass kissing customer support
-- The ridiculous amount of RAM they give with them (supplies a lot more room and safety while rendering)

I guess since the difference is on a software level, not physical, it is possible to "fool" your computer into
believing that you're using a Tesla instead of a GTX, but you'd literally fry your GeForce and lose the warranty.
It's not worth it unless you have a ton of liquid nitrogen and an elite cracker (cracka lol) who can write a proper driver.
March 12, 2012 3:02:03 PM

For gaming gtx 590, for cad animation and other stuff c2070
This is the simpest answer
April 17, 2012 2:59:03 PM

Chainzsaw said:
Those Tesla C2070's are meant more for thing such as CAD or Health science stuff. Basically anything that requires calculations that -isn't- gaming.

Yes you will get more performance for gaming out of the GTX 590 than the C2070.

But it will not be 2x the performance even though it has 2x cuda cores.

If your going for gaming - get the GTX.

If your doing workstation calculations - get the tesla.


Well yes, that is the standard answer. Everyone knows that Tesla is designed and marketed for computation and does not have a video output, etc.

That being said, is there any empirical support for the assertion that Tesla is actually better for computation per dollar? Exactly how much better in the case of single-precision computation?

Has anyone actually done pure computation benchmark comparisons between various GTX and Tesla models?
a b U Graphics card
a c 197 K Overclocking
April 17, 2012 3:47:32 PM

This topic has been closed by Jsc
!