Reviews - NVIDIA GeForce GTX TITAN in SLI

[Jimbo75]

You don`t understand how average is calculated. If one discard the WoW results and 3DMark, one Titan is 35% faster than one 7970GHz. If one include WoW and 3DMark, its 36.2%. Since its 18 games that is tested in total, that game won`t count as much. We have Wow which is 60% difference, we have the average which is 36% and we also have Sleeping Dogs which was around 12%. It all average out in the end.

If TPU did a review with only Batman, Assassin Creed 3 and Borderlands 2, people would have reacted too. That was my whole point of this review. Poor choice of games by Hexus which in total will tip the scale over to one side, in this case, AMD. It doesn`t show the complete picture

What about Hardware.fr?

http://translate.googleusercontent.c...O6HZL9jKHePS5Q

12% faster over 14 games including plenty of Nvidia titles...then again they also did the correct thing and tested the Titan after it had been through a run, so that it wasn't getting any artificial boost benefit from being cold.

[CAT-THE-FIFTH]

Oh dear I think Rollo has come back with a vengeance!

It seems so.

[Hicks12]

If you are gonna write down percentages, you better know what you are talking about. Understanding basic math would be a start.

At 1600p, Titan SLI exactly 2x (200%) faster than 7970GHz CF. A single Titan is beating 7970GHz by 35%.

Oh so I dont understand basic maths? Thats great however I think it is in fact you who has either misread the results or cannot count. Again going by the TPU chart we see that:
1. 1x titan is 26% faster than a 7970 Ghz - Costs ~£500 more than 7970

2. 2x titans are 74% faster than a single 7970ghz

3. 2x titans are 33% faster than 2x 7970ghz - Costs ~£1000 more than 2 7970s

4. 3x titans are 54% faster than 3x 7970ghz - Costs ~£1500 more than 3 7970s.

I think you can see here that the 7970 is much better value especially for only a hit of 26%! why pay 165% more for so little and i am struggling to see how 2 titans vs 1 7970 is fair game? 2 7970s vs 1 titan is fair as the amd set still costs less! the titans still arent double the speed!.


You dont see to understand averages, if anything this should show nvidia in a better light, you say about biased games yet TPU has loads and its a fair mix (probably more Nvidia though) and whats the point if you get 200fps in one game when the other setup gets 120fps, you wont notice any difference so its not relative performance!.


I think you're being rather ignorant and petty considering even your sources are saying its bad lol!.

[huaxshin]

Oh so I dont understand basic maths? Thats great however I think it is in fact you who has either misread the results or cannot count. Again going by the TPU chart we see that:
1. 1x titan is 26% faster than a 7970 Ghz - Costs ~£500 more than 7970

2. 2x titans are 74% faster than a single 7970ghz

3. 2x titans are 33% faster than 2x 7970ghz - Costs ~£1000 more than 2 7970s

4. 3x titans are 54% faster than 3x 7970ghz - Costs ~£1500 more than 3 7970s.

I think you can see here that the 7970 is much better value especially for only a hit of 26%! why pay 165% more for so little and i am struggling to see how 2 titans vs 1 7970 is fair game? 2 7970s vs 1 titan is fair as the amd set still costs less! the titans still arent double the speed!.


You dont see to understand averages, if anything this should show nvidia in a better light, you say about biased games yet TPU has loads and its a fair mix (probably more Nvidia though) and whats the point if you get 200fps in one game when the other setup gets 120fps, you wont notice any difference so its not relative performance!.


I think you're being rather ignorant and petty considering even your sources are saying its bad lol!.

Oh my god.

Did you just take 100%-74% = 26%?
148%-74% = 74%?

You can`t do that. Percentage doesn`t work like that

Its 100%/74% = 1.35 = 35%
148%/74% = 2 = 200%

I`m done here. Nice visiting you people here at Hexus. Some are quite hostile but I guess thats how the www just is. I`ve said my opinion about this review. Thats all I wanted. Up to Hexus wether to discard it or do some changes. I`m out. Bye

[Hicks12]

Not when its relative? there is nothing else needed, the titan has already been made a reference point.

Sorry but you were the one coming here and spouting nonsense that you cant even validate, if you came on here and gave a constructive response then yes the response would be difference I merely pointed out the difference.

Point out *if* my maths is flawed but dont be a prat about it.

[CAT-THE-FIFTH]

Not when its relative? there is nothing else needed, the titan has already been made a reference point.

Sorry but you were the one coming here and spouting nonsense that you cant even validate, if you came on here and gave a constructive response then yes the response would be difference I merely pointed out the difference.

Point out my flawed maths but dont be a prat about it.

Rollo's Fermi damage control comes to mind. You remember how fruitful replying to that was.

[Jimbo75]

I`m done here. Nice visiting you people here at Hexus. Some are quite hostile but I guess thats how the www just is. I`ve said my opinion about this review. Thats all I wanted. Up to Hexus wether to discard it or do some changes. I`m out. Bye

Ok so...

You didn't acknowledge my first point which was that out of the 6 games, Titan (and the 680) were outperforming their average in 4 of them.

On my second point about TPU's scores you suggest that I don't know how to calculate averages. Trust me, I know how to calculate averages - it is you who can't see what the benchmark is saying.

My 3rd point was that other websites tested more games and had Titan's lead even lower than here at Hexus. Hardware.fr tested 14 games including some very favourable Nvidia games, and found that Titan was only 12% faster at 1600p. You ignored this point again.

You didn't answer the point that I raised over most websites testing "cold" cards instead of warm cards. PCGH, hardware.fr and even Anandtech mentioned that results can and do vary depending on the cards temperature.

The long and short of it is, Titan is probably ~20% faster than the 7970 GHz edition. Some sites will have it nearer 10%, some nearer 30%.

[Terbinator]

In regards to Crysis performance, can you guys try a longer scene (not 30seconds?) and not on the first level, I have been reading reports that the first level is terrible and that there is something wrong in the game as other levels are much better, worth a look maybe?.

It is to do with the rope physics (i know lol) in effect.

[CAT-THE-FIFTH]

Even on the blue forum ,there seems to be newish poster,basically saying Geforce Titan is the Bugatti of cards,etc and worth the price,and saying the whole place is AMD biased. LOL!

It sounds familar,don't you think??

Edit!!

Geforce Titan has gone up in price:

http://www.scan.co.uk/shop/computer-...82688-cores%29

I thought these were £50 to £70 cheaper on launch day??

[Jimbo75]

Well let's not jump to conclusions that might be unfair on the new guy. I myself have been less than happy with Hexus's choice of games at varying points in the past.

However if anyone is going to go off on a rant then they at least need to be willing to see it out and answer the questions posed to them. Huaxshin avoided every single point I made - not one was answered - ergo I can only conclude that there was no real belief behind his accusations.

[CAT-THE-FIFTH]

Well let's not jump to conclusions that might be unfair on the new guy. I myself have been less than happy with Hexus's choice of games at varying points in the past.

However if anyone is going to go off on a rant then they at least need to be willing to see it out and answer the questions posed to them. Huaxshin avoided every single point I made - not one was answered - ergo I can only conclude that there was no real belief behind his accusations.

None of use ever accused Hexus of being "biased" though. It questions the journalistic integrity of people like Tarinder and DR,who have always been happy to take suggestions on board.

Even with Dirt:Showdown they reduced the settings to make it fairer for AMD cards and in fact they mentioned Nvidia marketing did not want the game!

The only reason why they upped the settings for Geforce Titan is to see,if the added compute performance was the reason for the relatively poor performance of the GTX680 in this case. In this case it appears,that a Nvidia card with very good DirectCompute performance will run it fine.

[brasco]

Scaling, power and acoustics are impressive for such a large chip, looking forward to the tri-sli results, other sites have shown that's where it starts to dominate - I thought for a sec who the hell would pay 2.5K for a few extra FPS in gaming? Then I realised it's the folk who want 7680 x 1600 and just to have the latest tech. The sports car analogies are pretty apt, but I don't think it's the Bugatti, Nvidia doesn't have any heritage Maybe a GTR vs an equally fast Hotrod - more a taste thing rather than top speeds?

As a tangential question about the GPU reviews, recently lots of reviewers have been bringing up the importance of frame times and microstutter, especially with Sli/Xfire, is that something Hexus would consider adding to the charts?

Geforce Titan has gone up in price:

http://www.scan.co.uk/shop/computer-...82688-cores%29

I thought these were £50 to £70 cheaper on launch day??

They were, I got mine for £850, maybe just upped it whilst stock is low?

[CAT-THE-FIFTH]

Scaling, power and acoustics are impressive for such a large chip, looking forward to the tri-sli results, other sites have shown that's where it starts to dominate - I thought for a sec who the hell would pay 2.5K for a few extra FPS in gaming? Then I realised it's the folk who want 7680 x 1600 and just to have the latest tech. The sports car analogies are pretty apt, but I don't think it's the Bugatti, Nvidia doesn't have any heritage Maybe a GTR vs an equally fast Hotrod - more a taste thing rather than top speeds?

They actually called "normal" cards a Toyota Supra and the Titan a Bugatti(apparently it is a quality thing) and the pricing is fine. However,Rollo did something similar too. Considering that normal cards such as the GTX580,etc can last for years,I am not sure what it was all about.

On a side note how do you think the Geforce Titan will do against a GTX580 your work?? Is a 2X increase possible??

As a tangential question about the GPU reviews, recently lots of reviewers have been bringing up the importance of frame times and microstutter, especially with Sli/Xfire, is that something Hexus would consider adding to the charts?

I thought the GTX690 was fine though,regarding stutter issues??? AFAIK,isn't that one area where Nvidia seems,better with,ie, dual card setups with scaling,etc??

They were, I got mine for £850, maybe just upped it whilst stock is low?

Maybe!!

[brasco]

On a side note how do you think the Geforce Titan will do against a GTX580 your work?? Is a 2X increase possible??

Not too sure in gaming, I'm only on dual 1920x1200 and only game on one, so probably quite a good bump but not as much as if I was driving 3 x 1600p.
With the CUDA stuff I should see x 2 in real terms, I'd be very happy with that

I thought the GTX690 was fine though,regarding stutter issues??? AFAIK,isn't that one area where Nvidia seems,better with,ie, dual card setups with scaling,etc??

They tend to do better yeah but any multi-gpu setup has produced quite a lot of stutter / lag. The 690 does pretty well and the sheer FPS at that level probably mitagates any visible issue.
7970 Xfire doesn't fair well, AMD are working on drivers to fix it though.

Here's your reading for the next hour

Frame Rating Pt1
Frame Rating Pt2
Frame Rating Pt3

[CAT-THE-FIFTH]

Thanks.

Here's your reading for the next hour

Frame Rating Pt1
Frame Rating Pt2
Frame Rating Pt3

I already read it before!!

It will be interesting to see how the official 13.3 drivers perform,as they are meant to be doing something different regarding memory management. The proof of the pudding is in the eating though!!

BTW,this is an interesting read too:

http://techreport.com/blogs.x/swasson

[watercooled]

As I've said on another thread, I really don't think the sports car analogy is relevant to GPUs, unless all cars are essentially exactly the same except for powerplant, but even so this isn't a good analogy as we'd be assuming everything else would remain the same after upgrading the engine, which would not be the case - weight would increase, the powertrain would struggle with increased torque, wheels would need to be bigger etc. There's really nothing fancy about more expensive GPUs, they're not better quality or anything, as far as a user is concerned they're all the same except performance changes a bit. And TBF spending about 3 times the price on titan vs 680 or 7970 won't yield a noticeable improvement in a lot of games - humans are not capable of perceiving tiny changes in FPS.

If you're going to use it for CUDA or something, then I don't think it's too unreasonable as it's significantly cheaper than the 'pro' compute cards.