The first "proper" Kepler news Fri 17th Feb?

TheRealDeal · 19 Mar 2012 at 14:09

N19h7m4r3 said:
True, although I'm not a fan of synthetic benchmarks. Pity, it's something at least.

I completely agree, that'll at least give a comparison. Although who runs a 7970 with an i3 2100?

Cpu's have never really affected scores on heaven so using an i7 probably would not have affected the score.

psychodil · 19 Mar 2012 at 14:10

Just 700mhz at idle thats all.

N19h7m4r3 · 19 Mar 2012 at 14:10

TheRealDeal said:
Cpu's have never really affected scores on heaven so using an i7 probably would not have affected the score.

Surely the low resolution would cause a CPU bottleneck like it does in gaming.

EDIT:
Here's what a user at overclock.net posted using his 7970

kitch9 · 19 Mar 2012 at 14:11

N19h7m4r3 said:
Surely the low resolution would cause a CPU bottleneck like it does in gaming.

Its not very heavy on CPU's at all as its designed to be a GPU test primarily.

TheRealDeal · 19 Mar 2012 at 14:12

N19h7m4r3 said:
Surely the low resolution would cause a CPU bottleneck like it does in gaming.

Heaven is gpu bound as far as i know even at pretty low resolution. His cpu speed is over 3ghz so its not so slow.

1kiwi · 19 Mar 2012 at 14:13

straxusii said:
First benchies from the guy in the shop with one.

(For comparison I get 48 at these same settings, my 480 runs ~580 speed, his CPU is an I3 but even so I am underwhelmed)

snip

I'm assuming that an overclocked 580 would be quicker then? Anyone care to bench theirs?

N19h7m4r3 · 19 Mar 2012 at 14:14

kitch9 said:
Its not very heavy on CPU's at all as its designed to be a GPU test primarily.

TheRealDeal said:
Heaven is gpu bound as far as i know even at pretty low resolution. His cpu speed is over 3ghz so its not so slow.

Ah alright. I never really use synthetic benchmarks anymore so never looked into it.

Hopefully some gaming benchmarks are out soon.

ALXAndy · 19 Mar 2012 at 14:16

kiwi1 said:
I'm assuming that an overclocked 580 would be quicker then? Anyone care to bench theirs?

Yes an overclocked 580 would be faster than a 480 overclocked.

Havana_UK · 19 Mar 2012 at 14:17

It's hitting good max FPS but very low min FPS. That sounds like a CPU bottleneck to me.

ALXAndy · 19 Mar 2012 at 14:19

Havana_UK said:
It's hitting good max FPS but very low min FPS. That sounds like a CPU bottleneck to me.

Heaven, IMO, is a wildly inaccurate benchmark.

For example if I run the benchmark straight away I get stuttering on the first run, which brings down my min FPS score.

If I allow it to run for a while before starting the benchmark it is far better.

However, I could run it ten times and whilst the overall score and FPS will be pretty much the same the min FPS is a different story.

It always tends to hiccup at the same place, dragging down my min FPS.

xsistor · 19 Mar 2012 at 14:25

that is using extreme tessellation at a crappy res. that whole bench leaks of fail. It's non standard in so many ways

TheRealDeal · 19 Mar 2012 at 14:31

Could it be that kepler just aint as good at tesselation as the fermi cards were. The cards look way different in specs. Also mid range 4-5 series were not tesselation power houses like the high end were.

CAT-THE-FIFTH · 19 Mar 2012 at 14:36

straxusii said:
Also he says clock speed as reported in the Nvidia CP and GPUz both say 700mhz

That would very high if it were the 2D clocks which I doubt they are. Also GPU-Z reports the default clockspeed as the maximum stock clockspeed of the GPU. The 2D clockspeeds would be listed in the GPU clockspeed pane not the default clockspeed pane.

So,is the 1006MHZ clockspeed,the boost clockspeed or are the review models pre-overclocked??

Havana_UK · 19 Mar 2012 at 14:37

TheRealDeal said:
Could it be that kepler just aint as good at tesselation as the fermi cards were. The cards look way different in specs. Also mid range 4-5 series were not tesselation power houses like the high end were.

Could be valid, as we all believe that GK104 was originally destined to be a GTX660 series GPU...

CAT-THE-FIFTH · 19 Mar 2012 at 14:40

Havana_UK said:
Could be valid, as we all believe that GK104 was originally destined to be a GTX660 series GPU...

What would be quite funny is if indeed the cards were originally badged as a GTX660TI or a GTX670TI,and at the last minute Nvidia asked partners to reflash them as the GTX680,and some partners forgot to do so before sending out the cards!!

stulid · 19 Mar 2012 at 14:41

Just run the Heaven3.0 at exactly the same settings with a stock 2500K+GTX570 = 1049.

Havana_UK · 19 Mar 2012 at 14:42

CAT-THE-FIFTH said:
What would be quite funny is if indeed the cards were originally badged as a GTX660TI or a GTX670TI,and at the last minute Nvidia asked partners to reflash them as the GTX680,and some partners forgot to do so before sending out the cards!!

Unfortunately it would only be REALLY funny if they also came with a GTX660 pricetag

Havana_UK · 19 Mar 2012 at 14:44

stulid said:
Just run the Heaven3.0 at exactly the same settings with a stock 2500K+GTX570 = 1049.

So approximately a 30% improvement....not three bad (considering it's supposed to be a GTX660).

Unfortunately with a GTX680 name tag on the card it's not the sort of performance improvement I'd expect at all, and I blame AMD for not making the 7970 good enough

CAT-THE-FIFTH · 19 Mar 2012 at 14:46

N19h7m4r3 said:
Surely the low resolution would cause a CPU bottleneck like it does in gaming.

EDIT:
Here's what a user at overclock.net posted using his 7970

The score is at 1920X1080 - the GK104 score is at 1600X900.

Rroff · 19 Mar 2012 at 14:47

straxusii said:
First benchies from the guy in the shop with one.

(For comparison I get 48 at these same settings, my 480 runs ~580 speed, his CPU is an I3 but even so I am underwhelmed)

Who can spot the 2 things wrong with this image?