All guess work at the moment and no solid data from anywhere as yet.
Not long to wait
Not long to wait
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
Geforce Titan rumours.
- Thread starter Gregster
- Start date
Soldato
http://www.guru3d.com/news_story/nvidia_geforce_titan_780_3dmark_score.html
Sorry if it's a repost...
Gordon Bennett if that's true then holy sheet !
Sorry if it's a repost...
Gordon Bennett if that's true then holy sheet !
Permabanned
My humble opinion is its going to be a 680 refresh. Now, I am probs way off here but with no solid evidence it got me to thinking. Remember how the 580 was to the 480? Remember guys, this is just my opinion!
Permabanned
There isnt any solid evidence yet is there?
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 33,188
I haven't run 3dmark in like 12 years but, does 11 use physics score to add to the total score, in other words, is that score not likely going to be in huge part because the 680gtx has crappy compute performance, and the GK110 is a compute beast by design? If it does, the difference in compute performance between a GK110/Gk104 would likely make a monumental difference there.
not entirely sure why people are talking about ground breaking or new, or different, 300mm^2 core that costs £300 and has X performance, 600mm^2 core that costs £700-800 that has X*1.8 score shouldn't be even remotely surprising. This is bog standard for the past 10 years, the only difference is worsening processes, harder to make bigger cores, worse power scaling with new processes, increased cost of wafers and production on every new process mean 500-550mm^2 cores, can't be made in the £400-500 bracket from Nvidia anymore
14nm and below could be serious issues because scaling is getting worse and worse. Intel's 14nm uses 14nm finfets, but other parts will still be 20nm, the main issue is its taking more passes to complete the various stages of cpu production, increasing time each wafer takes effectively increases costs a lot. GPU's need to badly fix the crossfire/sli issues, probably done by making smaller cores and sticking them on an interposer rather than via a off die slow ass link like current dual gpu cards are done, huge bandwidth and tiny latency vs current setups.
It's an interesting card if they made them in volume, cheap enough to push 300-350mm^2 cores from AMD and Nvidia down into the sub £200 range again... but they won't, so meh.
Gonna stick my neck out, from what I can make out a 690 is about 45% faster than a 680, allowing for clocks a Titan should be about 45% faster than than a 680 so it should be around 690 performance !
This is not even considering the advantages the wider bus will have etc, and I've always been sure the 680 was running past it's optimum efficiency re clocks and volts.
Lets wait and see
This is not even considering the advantages the wider bus will have etc, and I've always been sure the 680 was running past it's optimum efficiency re clocks and volts.
Lets wait and see
Wow, better than I thought, Love the will it play crysis...that has been my target, now I fear It will be to expensive
Man of Honour
The important question is can the Titan play Minesweeper @more than 30fps.
Soldato
Gonna stick my neck out, from what I can make out a 690 is about 45% faster than a 680, allowing for clocks a Titan should be about 45% faster than than a 680 so it should be around 690 performance !
This is not even considering the advantages the wider bus will have etc, and I've always been sure the 680 was running past it's optimum efficiency re clocks and volts.
Lets wait and see
Eh? The 690 is around 95% faster than a 680.
Soldato
It's different to the one that was reported as fake and was actually 2x680. But this could also be another fake.
Man of Honour
Thats an old one but at least they had the sense to not be connected to 3dmark11 for it to register.
If you look at the individual test results they are very similar to 2 x gk104s either on a GTX 690 or two single GTX 680s. If it was a GK110 used it would show a different variety of results in the individual tests (some higher some lower).
Last edited:
Sorry Rusty, but you surprise me saying something like that
How can a 690 be 95% faster than a 680
you are virtually saying crossfire is almost 100% efficient then ?Actually, now you say that, I think I recall the same crossed out card details from a picture in the other thread....could be wrong
Soldato
Sorry Rusty, but you surprise me saying something like that
How can a 690 be 95% faster than a 680
Firstly, it's not crossfire.
Secondly, in benchmarks I got near perfect scaling going from one GPU to two and in games like BF3 the same was observed.
Thirdly, it just is around 95% faster - 90% at worst - so if you're surprised this says more about you than it does my percentages.
Firstly, it's not crossfire.
Secondly, in benchmarks I got near perfect scaling going from one GPU to two and in games like BF3 the same was observed.
Thirdly, it just is around 95% faster - 90% at worst - so if you're surprised this says more about you than it does my percentages.
Well I just assumed a 690 is 2 GK104 chips with a bridge chip running AFR or whatever, I admit to never actually reading a review
I based my % on some numbers for Crysis benchies, but here is a more comprehensive comparison....
http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_690/28.html
Soldato
Well I just assumed a 690 is 2 GK104 chips with a bridge chip running AFR or whatever, I admit to never actually reading a review
I based my % on some numbers for Crysis benchies, but here is a more comprehensive comparison....
http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_690/28.html
You can't use relative performance graphs really because it includes games where:
a) there's no benefit in running multi GPU
b) CPU limited games
c) other reasons which I can't be bothered to go through
The result is the graph gets skewed.
The 690 is indeed two full fat (slightly down clocked) GK104 680 GPUs.
When discussing how fast the card is relative to its younger brother it's disingenuous to look at things where the extra grunt makes no difference. You could say that in certain games it's not that much faster than a 680 but that is a fault of the software itself rather than talking about the hardware in isolation.
Last edited:
Firstly, it's not crossfire.
Secondly, in benchmarks I got near perfect scaling going from one GPU to two and in games like BF3 the same was observed.
Thirdly, it just is around 95% faster - 90% at worst - so if you're surprised this says more about you than it does my percentages.
Just had a quick scan..."The GTX 690 has two GK104 Kepler GPUs arranged in an internal SLI configuration"
, Think You could have worked out I called SLI, "Crossfire", slight freudian slip there...my error 
Soldato
Just had a quick scan..."The GTX 690 has two GK104 Kepler GPUs arranged in an internal SLI configuration"
Of course I knew what you meant but doesn't mean you weren't wrong
.