***Official Reviews Thread*** Nvidia Geforce GTX Titan - Launched Feb. 21, 2013

Page 46 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

mutantmagnet

Member
Apr 6, 2009
41
1
71
It doesn't have Hyper-Q or Dynamic Parallelism like it's K20(x) brothers does it?

I'm also pretty sure it's cut on DP as well, just nowhere near as deeply as past GeForce products.

It isn't cut down in double precision at all.

Don't bother looking at reviews to find out compute features. Everybody is saying something that contradicts other reviews. I only think Anand and a german website have provided the most accurate information and they still fell short of some details.


It does have Hyper-Q but that has been cut down from 32 streams in the K20 to 8 streams.

It also has Dynamic Parallelism as confirmed by Nvidia themselves on a press sheet.

What has been removed as mentioned in the Anand review is MPI functionality, RDMA and ECC.

On top of that Titan has had VGX removed.
 
Last edited:

BoFox

Senior member
May 10, 2008
689
0
0
It isn't cut down in double precision at all.

Can't remember exactly, but I think that the Tesla versions of Fermi had 1/2 DP to SP ratio (whereas GTX 580 was cut down to 1/8).. For GFermi104/114, it was 1/12 DP to SP ratio, I think - compared against GKepler104's 1/24 ratio.

That's just my vague (most likely flawed) memory of it. The interesting thing is that Kepler K20 has 1/3 Double Precision ratio, while the workstation version of Fermi had 1/2.

Perhaps Maxwell will have full 1:1 DP to SP ratio, to keep up with the DP compute capability graph that NV presented a while ago at the launch of Fermi?
 

Vesku

Diamond Member
Aug 25, 2005
3,743
28
86
I'm guessing you are choosing not to read links.

As the link stated, it's filling a new role. I can't say it any better than the review did, if you choose to ignore, to keep pronouncing what you feel is the truth. Carry on. lol

A lot of talk about it being a workstation card, making sure people are aware there has been no official talk of Titan getting workstation drivers. Not sure I'd call it a new role, more like filling in for the 580 more completely than 680 and 690 did.
 
Last edited:

3DVagabond

Lifer
Aug 10, 2009
11,951
204
106
A lot of talk about it being a workstation card, making sure people are aware their has been no official talk of Titan getting workstation drivers. Not sure I'd call it a new role, more like filling in for the 580 more completely than 680 and 690 did.

Yes. I've seen where people are getting confused between compute capability and workstation cards which require optimized drivers. 1/3 DP isn't going to make Max or Maya run well. ;)
 

notty22

Diamond Member
Jan 1, 2010
3,375
0
0
Yes theres confusion here. Non-gamers buying this card would only be interested in another Nvidia card, it's not like any AMD card (cuda) would be a candidate. And in that, it's a less expensive option than Nvidia's workstation and server offerings.
http://www.extremetech.com/gaming/1...rings-supercomputing-performance-to-consumers
Nvidia’s GTX Titan brings supercomputing performance to consumers

The Titan’s non-gaming features are a bit more complex. GK110 is a supercomputing GPU first and foremost and includes support for features that games can’t take advantage of. The Tesla K20X is capable of spinning off its own work threads (Dynamic Parallelism), allows the CPU to spin off multiple GPU workloads simultaneously (Hyper-Q), includes a Grid Management Unit (GMU) to manage multi-threaded scenarios effectively, and includes new virtualization technology (GPU Direct).

The GTX Titan supports all these features, and is capable of running double-precision floating point operations at full GPU speed, as shown above. GK104, in contrast, limits double-precision code to 1/24 of single-precision performance. It supports compute level 3.5 (GK104 uses 3.0), and can dedicate up to 255 registers to a single thread, compared to 63 for GK104.

What do these features mean for gaming? Little to nothing. But that doesn’t make them irrelevant. CUDA programmers and developers who’ve wanted better double-precision performance from a GPU without paying for a Tesla could be very pleased with the $999 price tag on the GTX Titan.