GeForce 680 GTX

[baron75mk2]

Just curious to know what do you mean by that.

Just that a lot of folks ( not necessarily on PCS forum ) were always saying that SLI had tons of issues & was not worth bothering with , but im getting none of them as i did with crossfire (id say they were definitely right with that)

& some folks said that a single GTX680 would outpace 2 GTX580s simply because of the higher clock speeds on the GTX680 & that it would be newer tech ( this was before release of the 680 of course)

But i just don't always get something because its newer tech or because everyone says it great , i just calculate & make up my own mind on all things after taking others opinions.

Was just saying i was glad i went my own way with the GTX 580s SLI instead of Single GTX 680 as the benchmarks all over the net are confirming what i originally thought to begin with & this little bench session has also confirmed it as well.

Was not trying to offend anyone with a GTX680 in any way at all (hope it didn't come across that way , if it did i apologize).

but benches are benches after all :yes:

 

[keynes]

Nah, never intended to imply I was offended by your comment. I'd have expected sli gtx 580 to outperform a single gtx 680.

 

[baron75mk2]

Cheers Baron, glad you had some fun benchmarking yours too!

it certainly does seem that SLI 580s was a good option to go with, definitely higher scores. I'd considered doing SLI with the 680s but having gone with the 4GB version I've no idea if SLI would be possible due to the size of the jetstream...

Are the jetstreams wider than a standard card then ? , my gainward phantom 3 is a 2 & a half slot card , but i just put that on the bottom & got a reference GTX580 for the top card , worked out better for airflow in the end than two Gainwards as the top reference blows the air right out of the back of the case & spacing is better than what it would have been with 2 gainward phantom 3s

 

[baron75mk2]

Nah, never intended to imply I was offended by your comment. I'd have expected sli gtx 580 to outperform a single gtx 680.

what i was planning to do was to keep up the trend & when the newer nvidias come out ( i suppose they would be 7XX series) , then gettng two GTX680s cheaper & just continuing this trend.

But from what i understand unless i upgrade to the ivy bridge to enable the pci express 3 then it might not be worth getting the 2 680s as they perform better on pcie 3 & could possibly be bottle-necked by running them on pcie 2.
( not so sure about this as Ive seen plenty of folks running 680s on sandys & getting excellent results from them)

I think i will do some research on it closer to the time i suppose , but it would be a shame if i couldn't do it as that was the plan all along really - wouldnt want to go off upgrading my i5 as i wanted it to last a few years before doing that & im very pleased with the performance of it.

 

[blindhamster]

hmm I wonder which PCI slot my 680 is in...

 

[baron75mk2]

hmm I wonder which PCI slot my 680 is in...

what MOBO & processor do you have ?

 

[blindhamster]

3.5ghz i7 3770k overclocked to 4.6 by PCS and the z77 Sabertooth, which has 1 PCI2, 1 PCI1 adn 2 PCI3

 

[Buzz]

Use the slot that would be closest to your CPU that the card fits in. Is usually the one.

 

[baron75mk2]

Ive got 2X PCIe 3 slots but no ivy bridge , suppose i could always throw the 2X GTX680s in there & do the processor upgrade later , even in PCIe 2 they have to be faster than 2X GTx580s surely ?

( come on buzz help me out here , tell me what i want to hear -lol )

 

[blindhamster]

just checked the z77 Sabertooth manual, looks like PCS installed my 680 in the PCIE 1 slot. Will confirm when I get home before doing the rest of the list!

 

[baron75mk2]

just checked the z77 Sabertooth manual, looks like PCS installed my 680 in the PCIE 1 slot. Will confirm when I get home before doing the rest of the list!

No cant be , it requires at leased a PCIe 2.0 slot to work surely ? - it is for the GTX580 just looked at my box.

On that board , the GPU should be in either the white slot or the brown slot (makes no difference they are both PCIe 3.0 /2.0 16X) (8X in dual GPU mode , negligible difference between 16X & 8x )

But the black slot is PCIe 2.0 but it only runs in 4x mode (Don't want it in that one)

There isn't a PCIe 1 Slot on that mobo , just 3 PCIe 2.0 (1x) slots , small inch long ones

 

[blindhamster]

as Baron correctly stated, it's in the right PCI slot...

Buzz:
http://3dmark.com/3dm11/3485462;jsessionid=1wewe493f7n3c1kx4c7ub2a8bd
Thats with overclocking turned off and the system on default settings. My CPU apparently only clocked up to 3.1Ghz in that test which is more than a bit off in and of itself

3DMark Score
P9459

and a comparison with my previous score.

Notice the only major difference is my physics score dropped by 2k. Which makes sense as the test specifically states its a CPU only test...

It still seem to me like the card isn't working as well as it should.

 

[tom_gr7]

ssorry if ya done this already, but have ya contacted palit/nvidia n checked nvidia forums? bet some guru's on there will know.

 

[blindhamster]

I checked with Palit, they weren't especially helpful, simply saying that there is often some variation between the same cards, some working better and some working worse than standard... the fact that resulted in a 2k point difference is somewhat silly IMO, but it was all they had to say on the matter.

New drivers released today, I installed them, set physX to be CPU based only, my score was nearly identical to with the previous drivers (actually the physics score was a little higher and the graphics score a little lower - i.e. the graphics score got even worse)

 

[tom_gr7]

Sorry I aint too experienced with physx, but mines on "auto select"

again sorry if its been mentioned before, (not read the whole thread) Nut ya could use GPU-Z, and click "log to file", whilst benchmarking or gaming, then you can see how high the clock speed + everything else went?

 

[blindhamster]

question to everyone, when the GPU is working flat out (e.g. in benchmarking - what sort of numbers would I expect to see for GPU load, power consumption etc?) (I'm looking at my logs - tom reminded me that I had infact set logging on earlier today! thanks to )

 

[blindhamster]

so, I installed virtua mvp - which came with my motherboard, i then installed latest drivers from Nvidia. I also tried swapping the graphics card from the first PCI-Express slot to the 2nd - that didn't do anything so I then put it back.

I also ran Prime95 and FurMark to test both my GPU and CPU.
Results:
Prime95 I ran with 8 of the working sets for about half hour, temperature remained between 70 and 77 throughout. Dropped back to 30s within a minute or so of finishing.
FurMark scored 2808 in the preset 1920x1080 benchmark GPU was at about 73 degrees for that with an average fps of 47 - pretty sure that Virtua MVP didn't have any affect in this app.


I then ran 3dmark11 again as well as Heaven.
Results:

3dmark11

P13329
http://3dmark.com/3dm11/3487732;jsessionid=1kpcu77620a6w1bp64c9prratl

heaven

FPS:
49.2
Scores:
1239
Min FPS:
28.8
Max FPS:
242.8

Settings
Render: direct3d11
Mode: 1920x1080 8xAA fullscreen
Shaders: high
Textures: high
Filter: trilinear
Anisotropy: 16x
Occlusion: enabled
Refraction: enabled
Volumetric: enabled
Tessellation: extreme

I then uninstalled MVP as 3dmark complains that its basically 'cheating' and honestly I'd rather see what the card is actually doing...
Results:

3dmark11 with no MVP

P10038
http://3dmark.com/3dm11/3487968;jsessionid=12awt2g4dphiqsu10vawnwrla

heaven with no MVP

FPS: 47.6
Scores: 1198
Min FPS: 30.0
Max FPS: 122.3

So... heaven performs now just like it did before I installed MVP, however 3dmark11 seems to have improved by a fraction (300 points), I can only assume due to installing the latest drivers today

I also spoke with a chap called Sen at Nvidia, that told me if the GPU isn't performing then I should have it replaced, but I need to speak to PCS about that... Though while using Virtua MVP my scores are similar or higher than those that were not I guess, simple fact remains that the score should be higher without Virtua MVP however. So I think I'm still not satisfied with the gtx 680 4gb jetstream...

The other thing I noticed, the 2GB jetstreams clock etc are set to higher settings... anyone in the know know what I would need to change to set them the same without breaking anything? seem odd the 4GB didn't have the same clock settings set.

 

[PaulH]

http://www.guru3d.com/article/palit-geforce-gtx-680-4gb-jetstream-review/22

You are getting more than the review score for the 4gb version on that site? I honestly dont see a problem unless you are comparing against other highly overclocked systems.

The 4gb versions are also lower clocked than the 2gb versions by Palit themselves and they are ALL done like this, have a look at page one of the review.

http://www.guru3d.com/article/palit-geforce-gtx-680-4gb-jetstream-review/1

As it has already been explained the 705mhz is incorrectly being read - My opinion is theres nothing at all wrong with the card and for some reason your running benchmarks instead of playing games!!!

Also, word of warning dont start playing with the manufacturers default clocks, it will affect your warranty.

 

[blindhamster]

I guess I'll have to suck it up then! I guess it just comes down to seeing other peoples scores with regular gtx 680s, which do not note them as overclocked so my comparison (and expectations) are then raised by this. Also note that their 3dmark11 score is using an older and lower scoring CPU as well, whereas my score was mostly based off CPU lol, where most peoples scores of about 10000 that I've seen were mostly from the GPU (Graphics scores of 11-12k and Physics scores of about 8k, using older cpus, my cpu when overclocked is scoring about what it is for everyone else - 11.5k where my gpu just barely scraped in at 10008 with the new nvidia drivers which upon re evaluation is probably not that bad for a stock model...).

I did an experiment after my last post yesterday, and set the memory frequency and clock frequency and boost frequency to match the 2GB Jetstream, I also set the power to use up to 110% and it upped the score by a further 300 points. which does sit in line with the review you linked - I did these settings using 'ThunderMaster' the application that comes from Palit as part of the 4GB Jetstream (though I saved those settings as a profile and set them back to default afterward)

I did order some extra RAM as some suggested my RAM could be the bottleneck (doubt it but it doesnt hurt I guess). Now... I'd quite like to SLI my GPU, but that seems unlikely with the Jetstream, thoughts?

 

[PaulH]

I genuinely do not think anything is bottlenecking you, it wont say that there cards are overclocked, you could tweak it and tweak it to your hearts content but from what you have shown here it is getting the same as the reviewers scores everywhere I have looked - which to be honest are the ones i would trust.