GeForce GTX 760 Review: GK104 Shows Up (And Off) At $250

[pauldh]

Interesting side note - Did anyone notice the HD 7850 gaming bundle is gone on Newegg and 7870 only comes with Crysis 3? Buying up to 7950 still nets you 4 free games.

AMD said to Tech Report that "the promo was always advertised as a ‘while supplies last’ campaign. Sadly, some partners are running out of supply of codes."

Source: http://techreport.com/news/24998/e-tailers-running-out-of-codes-for-never-settle-reloaded-bundles

Yes, understandable. And yet the same codes still are available for the 7950, which as of right now desperately needs that advantage. I sill say the timing and selective code issuing is interesting, right? This may purely be Newegg's decision, I am not pointing fingers.

edit: BTW, sorry to go OT.

 

[viewgamer]

The Far Cry 3 Benchmark is unfair, HBAO ambient occlusion was used for the nvidia cards & HDAO was used for the AMD cards.
HDAO is of higher quality & is significantly more demanding that HBAO, they should have used the default SSAO.
http://www.hardocp.com/article/2012/12/17/far_cry_3_video_card_performance_iq_review/7#.Ucmw3G00-Fg
This is another blow to tom's hardware credibility.

 

[eklipz330]

well you can find the 7950 boost for about $260 AR nowadays, which makes it a far better deal.

 

[mapesdhs]

I want proof from somewhere other than your site as my experience with SLi'd 560Ti's doesn't seem to match yours but it does match other sites that are considered reputable.

Hard to answer that when I don't know what your experiences are.
What are your tests results? What is it you're seeing that's bugging you
with your cards? Indeed, which model cards do you have? If your cards are
the lesser ones with clocks around the 830 range then yeah perhaps not so
impressive - that's why I bought two 900MHz EVGA Crysis Editions. 8)
Since they're the main cards in my current gaming PC though, I haven't
yet tested them much except for various 3DMark runs (06, Vantage, 11,
Firestrike, etc.) In theory I could replace them with two reference 580s
I recently obtained (though that's overkill for the games I'm playing
just now), but it'd be a bit of a squeeze in the case (Antec 300) and I
don't know if the PSU could cope (850W... maybe, would need to check).

To summarise: I need a lot more info to be able to help with your
specific query.

Btw I'm not talking here about overall 3DMark scores which can of course
be skewed by CPU performance because of the Physics/Combined tests; I mean
check the Graphics scores individually. For example, here's a typical 670
vs. my 560Tis SLI at stock speed (the person who owns this 670 is someone
from whom I bought one of my 460s, so he supplied me with some test info):

http://www.3dmark.com/compare/3dm11/5859252/3dm11/6035982

Check the graphics scores: almost identical. And with the 560Tis oc'd:

http://www.3dmark.com/compare/3dm11/5859252/3dm11/6037434

Now the 560Ti SLI Graphics score has shot up. This is just an example though.
As always, it varies depending on the task, eg. my 560Tis would likely be
hopeless for running Firestrike Extreme because that test needs more than
1GB VRAM. What game are you running? Or is it some other task? Happy to
help if I can.

Ian.

PS. One thing though, don't use the 320.18 driver, it's not so good for
cards like the 460/560.

 

[pauldh]

The Far Cry 3 Benchmark is unfair, HBAO ambient occlusion was used for the nvidia cards & HDAO was used for the AMD cards.
HDAO is of higher quality & is significantly more demanding that HBAO, they should have used the default SSAO.
http://www.hardocp.com/article/2012/12/17/far_cry_3_video_card_performance_iq_review/7#.Ucmw3G00-Fg
This is another blow to tom's hardware credibility.

Chris is here, but if I may chime in one last time....

We discussed this when adding FC3 to our bench suite and agreed this was the way to go. Do you only play with SSAO even with cards that allow way more? Doubtful!

We test both ways, but at highest details prefer to test as folks would game, which is HDAO for AMD and HBAO with NVidia. There is nothing shady going on when it is presented clearly (not hidden), and consistent from story to story.

 

[Wisecracker]

As pointed out by multiple posters in Cleve's 'Best Graphics' update article, the Radeon HD 7870 is closer to the $200 price point (before rebates with Crysis3), and the Radeon HD 7950 is hitting $280 (and below, with rebate) including the '4 game bundle' -- in between the 2 cards is the cut-down Tahiti '7870 XT' (not to be mentioned in this article, but rated above the 660ti in a previous Tom's report?).

Can't you guys afford an unpaid intern to check prices for you? :lol:

The GTX760 look like a great update for nVidia fans, but it's another swing and a miss for Tom's in the real-world department.

 

[Mousemonkey]

I want proof from somewhere other than your site as my experience with SLi'd 560Ti's doesn't seem to match yours but it does match other sites that are considered reputable.

Hard to answer that when I don't know what your experiences are.
What are your tests results? What is it you're seeing that's bugging you
with your cards? Indeed, which model cards do you have? If your cards are
the lesser ones with clocks around the 830 range then yeah perhaps not so
impressive - that's why I bought two 900MHz EVGA Crysis Editions. 8)
Since they're the main cards in my current gaming PC though, I haven't
yet tested them much except for various 3DMark runs (06, Vantage, 11,
Firestrike, etc.) In theory I could replace them with two reference 580s
I recently obtained (though that's overkill for the games I'm playing
just now), but it'd be a bit of a squeeze in the case (Antec 300) and I
don't know if the PSU could cope (850W... maybe, would need to check).

To summarise: I need a lot more info to be able to help with your
specific query.

Btw I'm not talking here about overall 3DMark scores which can of course
be skewed by CPU performance because of the Physics/Combined tests; I mean
check the Graphics scores individually. For example, here's a typical 670
vs. my 560Tis SLI at stock speed (the person who owns this 670 is someone
from whom I bought one of my 460s, so he supplied me with some test info):

http://www.3dmark.com/compare/3dm11/5859252/3dm11/6035982

Check the graphics scores: almost identical. And with the 560Tis oc'd:

http://www.3dmark.com/compare/3dm11/5859252/3dm11/6037434

Now the 560Ti SLI Graphics score has shot up. This is just an example though.
As always, it varies depending on the task, eg. my 560Tis would likely be
hopeless for running Firestrike Extreme because that test needs more than
1GB VRAM. What game are you running? Or is it some other task? Happy to
help if I can.

Ian.

PS. One thing though, don't use the 320.18 driver, it's not so good for
cards like the 460/560.

I'm pretty sure I don't need your help but just for the record my cards were 900mhz OC models and they were bested by one of the two 660Ti's that I replaced them with, so put that on your site as a fact! :lol:

 

[mapesdhs]

Does that sound right to you Chris?

I should perhaps have said two decent 560Tis, because indeed there are various
models with clocks that are kinda on the low side, but I always felt anyone buying
ref versions was a bit daft since they weren't really that much better than a good 460.

Anyway, before judging, etc., did you actually look at the 3DMark11 links I gave?


Note that I constantly caveat re VRAM capacity. If one is playing newer games and/or
using higher resolutions/detail with heavy AA, etc., then newer cards do make much more
sense (and AMD has an advantage for single-card use with their default 3GB on so many
models), but older cards are nowhere near as bad as so many people frequently claim as it
often depends on the task as to where the VRAM limit really matters, eg. I find Crysis2 runs
rather well on my system at the settings I'm using, even though when testing with a 460 2GB
it's clear the game will grab up to around 1300MB if the VRAM is available. No doubt other
games/tasks will not fare so well when the VRAM is expended based on one's chosen
settings. Also note that atm I'm playing on just a single 1920x1200 display. I recently obtained
a 2560x1440 display so I'll be adding data at that res when I can - I expect that's when I'll
see 1GB results looking not so good for some tests.

Ian.

 

[mapesdhs]

cangelini writes:
> I need to take a closer look at what Fluidmark is doing (Igor did all of the compute numbers),
> but that's a good question.

Thanks! Yeah, kinda weird isn't it?


> For multiples, yeah, only the one here.

Ah well.


> The naming does make it difficult for enthusiasts to keep things straight. Both companies are guilty of it, ...

Indeed. Sometimes I can't help but rant. :}

Ian.

 

[Fulgurant]

Is this the first time you guys have measured power consumption on the card rather than on the whole system? Loving that.

Nice review.

 

[brennok]

So if you don't care about the 4 games in the 7970 Vapor-X on Newegg for $349 AR, is the card worth $100 more than the 760?

 

[mapesdhs]

Yeah mapesdhs, I am definitely with mouse on this one.

Really? So what tests have you run yourself to explore these issues?

Take careful note of the caveats I mention with my data. It's entirely possible for site reviews of the
latest cards to show different results because they so often use the latest titles which by definition
impose VRAM loads that older cards cannot cope with. I am using a number of older tests/games
precisely because that's what so many people keep asking about on forums I read, ie. questions
about upgrades, older games, etc. However, I'm also using Firestrike Extreme and a couple of
games at 2560 res which should show the limitations of older cards.

Point is, when someone asks about card A vs. B, it's often not a simple answer. Pointing them to
a site review may give an incorrect conclusion if they're not playing comparable games at
similar resolutions, etc. I'm trying to cover some of the ground that site reviews normally leave
out because naturally they will test newer cards with the latest titles. Go check the forums, there
are people out there who want to know whether it's worth upgrading their 6000+, E8400, Q6600
or Ph2 965 system with a 780. I'm just trying to fill in the blanks where I can.

Ian.

 

[JPNpower]

Inevitably AMD will come up with a competitor that provides GTX770 performance with GTX760 speed. It's a good idea for NVidia to have a "GTX760 ti" name to launch a competitor to that.

 

[Amdlova]

more watts for nothing... thanks i will be fine with 670...

 

[Novuake]

Yeah mapesdhs, I am definitely with mouse on this one.

Really? So what tests have you run yourself to explore these issues?

Take careful note of the caveats I mention with my data. It's entirely possible for site reviews of the
latest cards to show different results because they so often use the latest titles which by definition
impose VRAM loads that older cards cannot cope with. I am using a number of older tests/games
precisely because that's what so many people keep asking about on forums I read, ie. questions
about upgrades, older games, etc. However, I'm also using Firestrike Extreme and a couple of
games at 2560 res which should show the limitations of older cards.

Point is, when someone asks about card A vs. B, it's often not a simple answer. Pointing them to
a site review may give an incorrect conclusion if they're not playing comparable games at
similar resolutions, etc. I'm trying to cover some of the ground that site reviews normally leave
out because naturally they will test newer cards with the latest titles. Go check the forums, there
are people out there who want to know whether it's worth upgrading their 6000+, E8400, Q6600
or Ph2 965 system with a 780. I'm just trying to fill in the blanks where I can.

Ian.

Ever considered the fact that scaling on older games decrease dramatically as cards get faster?
There is always an FPS wall when using old games to bench new cards, even with a very intense game like Crysis that was used till recently by many reviewers.

 

[mapesdhs]

I'm pretty sure I don't need your help but just for the record my cards were 900mhz OC models and they were bested by one of the two 660Ti's that I replaced them with, so put that on your site as a fact! :lol:

Bested at what task for grud's sake? Am I the only one with some grounding in statistical methods?
I'd be absolutely happy to add it as info on my site, but it'd be a complete waste of time doing so without
knowing what task you run, at what res, settings, detail, CPU, etc. If you run something that exceeds the
VRAM limit of the 560Tis and the task in question doesn't cope with that so well then your experience is
hardly surprising, and that's precisely my point. It depends what you're doing.

Or are you saying my 3DMark11 tests were faked? Or somehow the other guy's 670 test was broken?

Read my original post, I said two older cards can be usefully better. It depends on so many factors,
as I did indeed state. Absolute assertions like your's are pointless because they lack the detail one
needs to make use of what you've said re any buying decision. So what do you get for running
3DMark11-P/X then? How does it compare to Firestrike/Extreme?

There is enormous overlap in how multiple old cards compare to new cards, depending on so many
factors. Your assertion is false because I've already provided proof of the fact, but your assertion is
also true because of whatever task you happen to be running. That's my point, it depends on what
one is doing; saying X is better than Y in some absolute sense is just plain wrong. Just like the model
names and numbers, reality is a right old muddle.

Ian.

 

[Mousemonkey]

I'm pretty sure I don't need your help but just for the record my cards were 900mhz OC models and they were bested by one of the two 660Ti's that I replaced them with, so put that on your site as a fact! :lol:

Bested at what task for grud's sake?

Both tasks, gaming and folding.

 

[JPNpower]

tomfreak, two good 460s are somewhat less than a single 670 (VRAM limits not
withstanding), while two 560s will be similar since they're just oc'd 460s and
most of them had slower clocks than the best 460s. Two 560Tis can be usefully
better, matching a 670, or even matching a 680 if oc'd, but again VRAM capacity
may be an issue, though for all these older cards there were 2GB versions
available (but often with slower clocks; I have two oc'd 2GB 460s @ 800MHz
which run quite well). My gaming PC still has two 900MHz 1GB 560TIs which are
quicker than a 670 at stock, quicker than a 680 oc'd, at least when the 1GB limit
is not an issue. They certainly cope with Crysis2 at high detail well enough.

However, comparing to these newer 600/700 cards, the only older cards that
do still perform well by comparison (by that I mean the potential gain from SLI)
are the 570/580, again assuming VRAM capacity is not a factor, though the
3GB 580 definitely shines here. Indeed, two 580s SLI are almost identical in
performance to a single 780 for 3DMark11, eg. see:

http://www.tweaktown.com/reviews/5516/nvidia-geforce-gtx-770-2gb-video-card-review/index4.html

Here are my 3DMark11 results for two 797MHz 1.5GB 580s:

http://www.3dmark.com/3dm11/6683648
http://www.3dmark.com/3dm11/6683683

If you want lots more 460/560 data, see my site:

http://www.sgidepot.co.uk/sgi.html#PC

Ian.

PS. I don't include the 470/480 because they run too hot for my liking
and thus I wouldn't use either in SLI.

Damn, I changed my homepage from washingtonpost.com because it took 10 friggin seconds to load. Your website has far more links and info and loaded instantly... Why don't more websites do this?

 

[hixbot]

My only gripe with the 760 is the misuse of a model number which allows one to
infer it should be quicker than older cards with 'lesser' names (660, etc.) when
infact it's often slower. I really wish NVIDIA would stop releasing products that
exhibit such enormous performance overlap. Given the evolutionary nature of
GPUs, and the time that has passed since the 600s launched, one might
reasonably expect a 760 to beat the 670 too, but it never does....

I agree, the 760 should match the 670 for less price. These "new generations" should offer better bang for the buck than the last generation but with GPU and CPUs for that matter, all we see is slightly more performance for slightly more money.

Good god, where is the competition to drive prices down?

 

[Novuake]

tomfreak, two good 460s are somewhat less than a single 670 (VRAM limits not
withstanding), while two 560s will be similar since they're just oc'd 460s and
most of them had slower clocks than the best 460s. Two 560Tis can be usefully
better, matching a 670, or even matching a 680 if oc'd, but again VRAM capacity
may be an issue, though for all these older cards there were 2GB versions
available (but often with slower clocks; I have two oc'd 2GB 460s @ 800MHz
which run quite well). My gaming PC still has two 900MHz 1GB 560TIs which are
quicker than a 670 at stock, quicker than a 680 oc'd, at least when the 1GB limit
is not an issue. They certainly cope with Crysis2 at high detail well enough.

However, comparing to these newer 600/700 cards, the only older cards that
do still perform well by comparison (by that I mean the potential gain from SLI)
are the 570/580, again assuming VRAM capacity is not a factor, though the
3GB 580 definitely shines here. Indeed, two 580s SLI are almost identical in
performance to a single 780 for 3DMark11, eg. see:

http://www.tweaktown.com/reviews/5516/nvidia-geforce-gtx-770-2gb-video-card-review/index4.html

Here are my 3DMark11 results for two 797MHz 1.5GB 580s:

http://www.3dmark.com/3dm11/6683648
http://www.3dmark.com/3dm11/6683683

If you want lots more 460/560 data, see my site:

http://www.sgidepot.co.uk/sgi.html#PC

Ian.

PS. I don't include the 470/480 because they run too hot for my liking
and thus I wouldn't use either in SLI.

Damn, I changed my homepage from washingtonpost.com because it took 10 friggin seconds to load. Your website has far more links and info and loaded instantly... Why don't more websites do this?

Because his website is bare? No flash, Jquery or any such thing...

 

[mapesdhs]

Both tasks, gaming and folding.

Again, no useful/relevant detail, so a pointless comment. QED.

And it's pretty obvious I wasn't talking about folding btw, that's an area I haven't
investigated at all because it's not what people on forums are asking about.

Ian.

 

[016ive]

AMD is losing a lot of money allowing Nvidia to rule all categories like this! AMD must release Radeon 8970/8950 NOW!

 

[Mousemonkey]

Both tasks, gaming and folding.

Again, no useful/relevant detail, so a pointless comment. QED.

And it's pretty obvious I wasn't talking about folding btw, that's an area I haven't
investigated at all because it's not what people on forums are asking about.

Ian.

You seem to think that I want or need your help, I don't I just dispute your claims that SLi'd 560Ti's are faster than a 670 and 680 and I'm not the only one it seems.

 

[Killroyjenkins]

Idk this just doesn't impress me much because:

1) The 7950 has 50% more RAM (And you can bet your gonna need it this fall)
2) The 7950 overclocks to give you 20-30% (I have seen 40%) more performance above stock whereas the 760 basically just hits a wall no matter what you do.
3) The 7950 comes with games that can save $100+ or you can sell for $50.

Am I the only one who thinks this about the 760 and 770?

I sort of wanted to see higher settings than just 1080p. All of the results look very close, and it seems the graphics weren't stressed as much as they could be.

On an unrelated note, I Just bought a 7970 for $300. I did get the keys to redeem with AMD for reloaded, but they have yet to give me bioshock infinite. Hopefully they're not out.

 

[mapesdhs]

Ever considered the fact that scaling on older games decrease dramatically as cards get faster?
There is always an FPS wall when using old games to bench new cards, even with a very intense game like Crysis that was used till recently by many reviewers.

Always? Again it depends on the game and definitely on the platform. Crysis exhibits unique behaviour
IMO partly because it wasn't written that well. Crysis2 fares much better on older multiple cards. Saying
there's an FPS wall doesn't mean much when what people actually want to know varies a lot based
on their display res, platform, the game in question, what aspect of their setup they want to upgrade, etc.
Quite often, a decreasing return in improved FPS combines with very high numbers anyway, in which
case the gains are not visible (especially true at low res for older games), but at high res the situation
can be very different. At other times a proposed upgrade won't help at all because of a CPU bottleneck
(not all games are so CPU-bound; CoJ at high res for example, a game which btw seems to favour
AMD cards), but again it depends on the game/task. Each time you guys are stating absolutes when
really there just aren't any.

I'd genuinely like to know exactly what game & system spec/res/setup Mousemonkey is using but he
won't say (why?). If he's seeing two 900MHz 560Tis being beaten by a 660Ti, that really is useful info
and I'd be happy to add some data points to my site if I can, but not unless I have a full description. So
far he hasn't even stated which games he's tested, or given any recorded performance data.

I have to wonder if either of you have even bothered looking at my results...

Ian.

PS. Please don't quite my entire post each time; it's kinda clogging up the thread a bit.