Hardware Nvidia GeForce GTX Titan First Look

Three, I believe.

The test suite mirrors that of Custom PC, in which they use Skyrim, BF3, Crysis 2, The Witcher 2 and Heaven 3.0.

Perhaps Harry or Matt can confirm?

 

Oh. Heaven 3.0. 2560x1600 res, GTX580 now equals 846. That looks more like it I guess!

 

We updated the heaven settings test for this review; i'll fix the copy. Settings were

2,560 x 1,600

API DX11
Tessellation Normal
Shaders High
Anisotropy 16x
Anti-Aliasing 4 x

We're still using Heaven 3.0 though, although I doubt a version change would mess up much.

 

Would have been nice to have seen the 580 GTX added to the results to see what kind of leap could be expected.

 

Just look at the 660Ti results, they're close enough.

 

I expect 2x GTX 580's 3GB to perform the same, if not a little faster than the Titan.

 

Just bagged me an ASUS.

 

Only one

 

Evga has its hydrocopper ones up on its site and seen pics across net now.

Im in the market for a card soon, so looks like this one will do.

 

How much though? £1000+???

 

Yeah, Im guessing loads, as they havent disclosed how much as yet as far as i can tell.

Googling like mad at the moment for info he he.

Might avoid luxuries like rent heat and food, so i can save up for one.

 

Most Sli CF users would give up SLI or CF tommorow if they could get a single card to offer the performance.

SLI CF is still way to dependent on drivers for many.

680 sli cost most on this forum the best part of £850 without watercooling.

This is now the only single gpu card on the market that can do bf3 ultra at 2560 res with 60+ fps id consider it a win for nvidia.

If nvidia can release a card this powerful next gen that offers the performance id like to see then id be sorely tempted to buy it. If i did not own 680 sli id be tempted to buy it.

Not many Multi GPU users can say they have never had a single problem whilst if you ask most single card users they have had close to 0.

Does not make any dif if 580 sli or 7950 cf is slightly faster for many having one single card that offers that sort of performance is a benifit worth paying for.

Thats the market nvidia has aimed at and it will sell to that market.

 

would a pair of 7870 le in CF actually match the Titan?

 

Not true, I will always go with multiple cards and I know a lot of others that will always go down this route too, as SLI has been a joy.

You have just as much chance of having driver problems on a single card situation as you would when running multiple cards. I only ever get the odd Drivers have stop responding and have recovered, when I OC to far.

Just to add, BF3 @ 2560x1440 maxed out with 4xMSAA runs @ 80FPS avg with 2x GTX 580 3GB

People are still reading post regarding SLI back when the 7800 GTX series was out, when SLI was not at it's optimal standard's and where drivers were a bit more flaky.
This isn't the case in today's driver and SLI support, as more people are opting for multiple GPU configurations.
NVidia are very good at bringing out new SLI profiles when new games come out, and they always aim to improve performance in newer driver releases.

 

Got to say I am with TG on this one. SLI issues are long gone, and if new issues arrive they are short lived. I have SLI for years now and not had an issue for 2years + and I play almost every major new release of games in one form or another.

580 3GB SLI are currently the best bang for buck by a country mile.

The Titan does however open a new market, along side the 690 gives ITX a high end gaming potential. I am more interested in is, will the Titan release prompt more "top end" ITX boards to hit the market - that really get my pulse going.

Having my gaming rig under one arm and my Ipad in the other would be a dream!

 

We have to agree to disagree on the PCI-E 2 Vs 3, sorry.

My point is this - the FPS may change slightly using the x79 set-up, but the order of which card is fastest and where would not change - and that is where I feel your critique is flawed - the graphs would scale and the status quo remains.

 

Well that's fine, I'm basing my opinions on results such as these. That's with 4 cards, with two it's also a similar story, although not quite as drastic. Again this isn't really going to be a big of an issue unless you're using current gen top-end cards, like the ones tested here.

Although you're right in that the card order wouldn't be likely to change. The magnitude of the differences could be large for all we know at present. For a benchmark review, you shouldn't introduce a bottleneck, full stop.

 

I really don't understand the importance here, you concede the result order would not change - which is probably the single most important thing to take away here.

The resolution in those tests you've linked start at 3600*1920 which is much higher than the highest resolutions in the review (6.95m pixels Vs 6.22m)....the overall effect of bandwidth limitations should be reduced in a similar fashion.

Lest we forget that order of the cards would not change, it could be as easy at looking at those percentages (in your link) and applying them to get a reasonable guess at what the FPS might be if all lanes were PCI-E 3 x16 instead of x8.

I must of exagerated when I said the 0.05% of the pc population who have a system that actually cares about performance at that resolution, it's closer to 0.001%....it's a poor example, produced by people who want to sell PCI-E 3 products.....I'd still buy a single titan for my P67 rig, if it were half it's current price that is

Also when you find a review that meets your criteria - please, send me a link

 

I think there's a rather important consideration that has to be made though, that the numbers themselves are very important. Let's take a hypothetical situation, you have two cards A and B. A costs 15% more than B, yet performs 10% better than B. Card A is found to be bottlenecked, and when that is removed, it performs 17% better than B. Those kinds of figures could quite easily sway somebody to make a purchase on either, but the full potential has to be shown to make it a fair situation.

The reason why I don't really think the order is that important is that it was a given for the most part. We expected the Titan to perform better than current single cards, but we didn't know by how much it would. We expected it to have reduced performance against a dual GPU solution, but again not by how much. Now a single card won't have this issue. But once we start hitting 2+, potentially the Titan is more severely bottlenecked than the other cards by comparison, but we don't know, it may not be.

Also, plenty of reviewers hit those standards, all that's really required is to remove the CPU and mobo from the equation. Aka just use a socket 2011 CPU and voila. Why not just stick it on a similar system used to test 2011 coolers? I mean it's not as if Bit-tech doesn't have access to the parts, they chose the Z77 mobo and i7. Of course, there's always the new in-thing of measuring frame latency that's popping up, but that's another topic. I just think it's a bit odd to go for something that could potentially negatively affect results. It also doesn't make sense from a relevancy standpoint as people who want more than one Titan simply won't be using Z77. A single Titan is the cost of a decent Z77 setup, let alone two or three of them.

 

X79 has a hell of a lot more bandwidth to give with the extra PCIe lane.

Here's my old results running 3 GTX 580 1.5GB cards on a Gigabyte UD7 that does X16 x8 x8
And the Asus Rampage IV does x16 x16 x8.

I see a big difference in performance. So the Z77 is holding back those 3 GTX Titans.

990X @5GHz Gigabyte UD7


3960X @5GHz Asus Rampage IV


990X @5GHZ Gigabyte UD7


3960X @5GHZ Asus Rampage IV