Geforce GTX Titan 6GB

[ATH]

A new $1,000 card which will probably be the new fastest single GPU card. No benchmarks for a few more days though:

http://www.tomshardware.com/reviews/...view,3438.html

http://hothardware.com/Reviews/NVIDI...Debuts/?page=1

Rather, this thing incorporates a GPU currently found in the Tesla K20X (which HP will sell you for $7,700), a cooler clearly derived from the marquee GeForce GTX 690, and a staggering 6 GB of GDDR5 memory. As you’ll soon see, the combination generally falls between a GeForce GTX 690 and Radeon HD 7970 GHz Edition in our benchmarks. That means:

1. Pay the same $1,000 for a GeForce GTX 690 if you only want one dual-slot card and your case accommodates the long board. It remains the fastest graphics solution we’ve ever tested, so there's no real reason not to favor it over Titan.
2. The Titan isn’t worth $600 more than a Radeon HD 7970 GHz Edition. Two of AMD’s cards are going to be faster and cost less. Of course, they’re also distractingly loud when you hit them with a demanding load. Make sure you have room for two dual-slot cards with one vacant space between them. Typically, I frown on such inelegance, but more speed for $200 less could be worth the trade-off in a roomy case.
3. Buy a GeForce GTX Titan when you want the fastest gaming experience possible from a mini-ITX machine like Falcon Northwest’s Tiki or iBuyPower’s Revolt. A 690 isn’t practical due to its length, power requirements, and axial-flow fan.
4. Buy a GeForce GTX Titan if you have a trio of 1920x1080/2560x1440/2560x1600 screens and fully intend to use two or three cards in SLI. In the most demanding titles, two GK110s scale much more linearly than four GK104s (dual GeForce GTX 690s). Three Titan cards are just Ludicrous Gibs!

[firejuggler]

but , what are the words about TF speed?

[TheJudger]

http://www.mersenneforum.org/showpos...postcount=2212
The Tesla K20 is a GK110 with 13 SMX (2496 cores) running @ 705MHz (up to 758MHz "application clock").

Oliver

[chalsall]

Originally Posted by TheJudger

http://www.mersenneforum.org/showpos...postcount=2212
The Tesla K20 is a GK110 with 13 SMX (2496 cores) running @ 705MHz (up to 758MHz "application clock").

And this suggests...?

[TheJudger]

The K20 (application clock set to 758MHz) is ~6-7% faster than my stock GTX 470 using the barrett76_gs kernel on some recents exponents from gpu72.
Assuming linear scaling (mfaktc is known to scale very well on CC 1.x and 2.x, only little information on CC 3.x):
GTX Titan vs. K20: (876MHz * 14 SMX) / (758MHz * 13 SMX) = 1.24.
Assuming "average clock" of 876MHz for the Titan, 758MHz application clock for the K20, 14 SMX enabled on Titan, 13 SMX enabled on K20. A SMX is a group of 192 cores. So a GTX Titan is 1.07 * 1.24 = 1.33 times faster than my old GTX 470. A GTX 580 is ~1.5 times faster than my GTX 470...
Of course this is only my guess but I'm affraid that is the truth...

Oliver

[chalsall]

Originally Posted by TheJudger

Of course this is only my guess but I'm affraid that is the truth...

Thank you for your very informed guess.

[ATH]

I know the GTX 580 is better than the 680, but did anyone test GTX 690? Is it just like 2x680? Combined it must be better than one 580?


Btw my new "if-I-won-the-lottery"-dream is to get a Haswell in June with 3x GTX Titan's heh. I would have to win >$150,000 though to want to spend so much on 1 computer.

[ixfd64]

The GK110 was promised to be more compute-oriented, so the mfaktc throughput may be better than expected. But then again, the only way to know for sure is to actually run mfaktc on it.

[chalsall]

Originally Posted by ixfd64

The GK110 was promised to be more compute-oriented, so the mfaktc throughput may be better than expected. But then again, the only way to know for sure is to actually run mfaktc on it.

One thing I found interesting on the TomsHardare review was on the "Compute Performance And Striking A Balance" page. It appears that the Windows UI for the driver allows you to increase DP performance beyond the nominal artificial limits intended to constrain power consumption and temperature.

Oliver... Your thoughts on this would be welcomed.

[sdbardwick]

And they de-Nerfed the FP64 abilities.
Titan: 1/3 FP32
680: 1/24 FP32
580: 1/8 FP32

[chalsall]

Originally Posted by firejuggler

6 GB of non-shared memory mean decent P-1 power, right?

Very decent.

Sadly, we don't have a GPU P-1 program....