-
Geforce 9600GT Some Close Pics. & SLI Results
-
Looks like a great card, if priced well. Im wondering why is that card so damn long? Though, I doubt that the retail version will remain the same.
-
Quote:
Originally Posted by
Calmatory
Looks like a great card, if priced well. Im wondering why is that card so damn long? Though, I doubt that the retail version will remain the same.
Yes it's as long as the 8800GT.Strange for a midrange card.(it also uses aux. power)
The 8600GT was much smaller.
-
such a card already exists for some time.... ati 3850 now lets see what price range and power consumption this will be
-
I thought it was on a 128bit bandwidth...
-
this is a situation where tri-sli becomes interesting. whats 3x9600gt gonna cost, a little under $600?
-
Except it does not heve 2 SLI connectors, so max cards in SLI = 2
And so far SLI scales terribly bad, 10k vs 13k in 3dm06? thats like 30%? I guess its a driver problem.
-
Quote:
Originally Posted by
xoqolatl
Except it does not heve 2 SLI connectors, so max cards in SLI = 2
And so far SLI scales terribly bad, 10k vs 13k in 3dm06? thats like 30%? I guess its a driver problem.
I guess CPU bottleneck, early drivers and overall bad design of SLI. :shrug:
-
This is looking good!
Cheers Absolut.
-
Quote:
Originally Posted by
Calmatory
I guess CPU bottleneck, early drivers and overall bad design of SLI. :shrug:
Nope,the CPU is not a bottleneck,it's the way how the final mar06 score is calculated.SLI scales pretty good(but i think the CF with RV670 is still a little better in that department).
-
That single 06 score looks high considering it has 64sp rather than 128 of the 8800U. Maybe it is a misprint, hence why the SLi is 13k?
Regards
Andy
-
Quote:
Originally Posted by
informal
Nope,the CPU is not a bottleneck,it's the way how the final mar06 score is calculated.SLI scales pretty good(but i think the CF with RV670 is still a little better in that department).
The CPU at 3Ghz is most certainly a bottleneck.
-
not for the 9600GT. If this Nvidia's new Architecture each shader might give better performance vs the old 8 series. The 10K in 06 does look a bit fishy. Core and shader clocks could be super high thus giving it a better scores at lower resolutions. Crank the res up and watch the score fall off quite a bit as raw shader power can not save the card.
-
Quote:
Originally Posted by
..::ABSOLUT::..
Yes it's as long as the 8800GT.Strange for a midrange card.(it also uses aux. power)
The 8600GT was much smaller.
I guess PCI-e2 can deliver enough power for this card without extra PCI-E connector. But that the PCI-E connector is purely there to make shure that there is OC room and to make shure its Compatible with PCI-e boards.
And i guess the SLI scores suck cause of driver issues?
-
Quote:
Originally Posted by
vengance_01
not for the 9600GT. If this Nvidia's new Architecture each shader might give better performance vs the old 8 series. The 10K in 06 does look a bit fishy. Core and shader clocks could be super high thus giving it a better scores at lower resolutions. Crank the res up and watch the score fall off quite a bit as raw shader power can not save the card.
Having 2 9600GT in SLI, a 3Ghz C2D is most certainly a bottleneck :rolleyes:
-
I hope the difference between the 9800s and the 9600s is the same as from 8800 to 8600
-
Quote:
Originally Posted by
CraptacularOne
Having 2 9600GT in SLI, a 3Ghz C2D is most certainly a bottleneck :rolleyes:
prove it.
-
Quote:
Originally Posted by
xlink
I hope the difference between the 9800s and the 9600s is the same as from 8800 to 8600
Ha no way!! Then there GF9800 must be about 70-90% faster than GF8800:) Is it possible?? I don`t think so. IMO probably G100 will bring us such a big performance boost over G80.
-
Quote:
Originally Posted by
halo112358
prove it.
*sigh*....as if the numbers for the SLI run weren't enough...... Seriously, if you think a 3Ghz C2D is enough to push the cards in SLI to their limit in 3Dmark...you have a LOT to learn about benchmarking.
-
Quote:
Originally Posted by
CraptacularOne
*sigh*....as if the numbers for the SLI run weren't enough...... Seriously, if you think a 3Ghz C2D is enough to push the cards in SLI to their limit in 3Dmark...you have a LOT to learn about benchmarking.
I'm not asking for you to flame me, I'm asking for you to prove your statement.
Here's the deal, a single c2d running at 2.66ghz is fast enough that an 8800GTX is GPU bound. A single 9600GT is similar in speed to an 8800GTX, which means that a 9600GT should also be GPU bound.
I'm not as familiar with SLI setups and that's why I asked you to prove your statement, I want to know also.
-
Quote:
Originally Posted by
halo112358
I'm not asking for you to flame me, I'm asking for you to prove your statement.
Here's the deal, a single c2d running at 2.66ghz is fast enough that an 8800GTX is GPU bound. A single 9600GT is similar in speed to an 8800GTX, which means that a 9600GT should also be GPU bound.
I'm not as familiar with SLI setups and that's why I asked you to prove your statement, I want to know also.
I'm not flaming you, I'm educating you. And no a 2.66Ghz C2D is NOT enough to remove the CPU bottleneck from a 8800GTX. In the resolution that 3Dmark06 benches in (1280x1024) you will need a lot more CPU speed to realize the potential of a single 8800GTX.
Here is a simple test I did for you. One run at 2.66Ghz and another at 3.6Ghz. The 8800GTX remained at the same speed for both tests. Look at the difference in SM2 and SM3 scores. In the test with the higher speed CPU you'll notice that there is quite a difference. Why? because @ 2.66Ghz the CPU cannot feed the GPU enough data to have it run at full potential.
2.66Ghz
http://xs223.xs.to/xs223/08053/2.66ghz420.jpg.xs.jpg
3.6Ghz
http://xs223.xs.to/xs223/08053/3.6ghz630.jpg.xs.jpg
-
having a good cpu helps in 3dmark06, cause that is like a cpu bench nowadays, it cant really compare graphics cards anymore
-
VERY nice card, but it better be priced better than the 3850
-
do u guys think its faster than the 8800gts 320?
i think it just might be
-
nice benches craptacular one. Now I wonder if you have crysis installed, can you do the same tests using fraps or maybe the built in benchmark? :p:
-
Quote:
Originally Posted by
CraptacularOne
Having 2 9600GT in SLI, a 3Ghz C2D is most certainly a bottleneck :rolleyes:
of course not. Why would you think these cards are that powerful, there actually not.
-
Quote:
Originally Posted by
ownage
of course not. Why would you think these cards are that powerful, there actually not.
A card that scores 10k in 06 is not powerful?...*sigh* are you blind or just cannot read? 3Ghz C2D is a bottleneck for them it's plainly obvious from the score being only 13K in 06 with SLI.
-
Quote:
Originally Posted by
CraptacularOne
A card that scores 10k in 06 is not powerful?...*sigh* are you blind or just cannot read? 3Ghz C2D is a bottleneck for them it's plainly obvious from the score being only 13K in 06 with SLI.
Flaming :down:
Or they could have SLI issues with this card, the benchmark is inconclusive.
-
Quote:
Originally Posted by
halo112358
Flaming :down:
Or they could have SLI issues with this card, the benchmark is inconclusive.
I beg to differ. It explains why 3DMark06 gets owned by the newer quads with the same cards as 6 months ago. People were saying that the 8800's were CPU limited since they came out. Well, at least in 3DM06. And he wasn't flaming, he was pointing out valid points.
-
Thanks for the test CraptacularOne. I just learned that I need to run my cpu faster playing crysis. :up:
-
Quote:
Originally Posted by
halo112358
Flaming :down:
Or they could have SLI issues with this card, the benchmark is inconclusive.
Again, I wasn't flaming. What is your obsession with thinking that I am? Do you want me to flame you?
I am trying to point out that at 3Ghz a C2D is not pushing the new 9600GT to it's limit and at that speed it has absolutely no hope what so ever for pushing SLI 9600GT's even close to their limit.
-
The 13k SLI score gets bottlenecked by the CPU score. If you look at the singlecard SM2 and SM3 scores and compare it to the SLI SM2 and SM3 scores, you see much less bottleneck. 3Dmark06 is just stupid, its to cpu depended.
SM2 and SM3 scores are good, but if both scores are over 5000, and the cpu score is 2600, then the scores goes down.
Look at this again, http://img151.imageshack.us/img151/8...22a7fe7sp2.jpg
No bottleneck in other benchmarks.
-
Quote:
Originally Posted by
grimREEFER
do u guys think its faster than the 8800gts 320?
i think it just might be
nope, and i would choose everytime a G80 core based 8800GTS then 64su 9600GT :]
-
Quote:
Originally Posted by
grimREEFER
do u guys think its faster than the 8800gts 320?
i think it just might be
It might not be faster at stock speeds, but no doubt that this thing (with it's 65nm process) will oc up the ass.
-
Quote:
Originally Posted by
GAR
VERY nice card, but it better be priced better than the 3850
I agree, looks like the best midrange card ever made by nv, very different from the jump from 7600GT to 8600GT/S crap :yepp:.My next card probably
-
To me, it seems the 9600GT is right between the 3870 and 8800gt
Releasing it, would kill the sale of 8x00 series (especially all the pos 8x00 256MB cards)
Question:
Why the hell is the GPU shimless?!?!
-
Quote:
Originally Posted by
CraptacularOne
The CPU at 3Ghz is most certainly a bottleneck.
dif between conroe CPU @3.0Ghz and 3.6Ghz is like 100-200 points when using an 8800GTS 640.
so I am thinking ti's the otherway around.
-
Quote:
Originally Posted by
xlink
dif between conroe CPU @3.0Ghz and 3.6Ghz is like 100-200 points when using an 8800GTS 640.
so I am thinking ti's the otherway around.
If you looked at CraptacularOne's post on page 1 you'd have noticed that simply raising the CPU from 2.66 to 3.6GHz increases the SM2 & SM3 scores. It's either the CPU or the FSB holding the card back by not passing the data to it fast enough.
-
Quote:
Originally Posted by
grimREEFER
having a good cpu helps in 3dmark06, cause that is like a cpu bench nowadays, it cant really compare graphics cards anymore
Thanks for CPUs being way too slow compared to GPUs. Besides, 3DMark was never meant to be a "GPU bench".
Well... I started to think about this "CPU bottleneck"-issue more deeply. This could be proven by someone who has similar card with similar CPU speed. Raise the CPU speed by 20%, then see the difference, then do the sme to GPU, Vmem and shaders and see which gives the biggest boost. Of course the CPU speed affects SM2/SM3 scores, it always does. The other thing is how MUCH it affects.
More and more I suspect bad drivers being the problem. Though, 100% more GPU power would require 100% more CPU power, that would show the efficiency of SLI, without major bottlenecks.
-
Quote:
Originally Posted by
fragmasterMax
It might not be faster at stock speeds, but no doubt that this thing (with it's 65nm process) will oc up the ass.
Given it's high stock clocks (650) and weak cooling, I really doubt that.
The 8800GT (600) and even more so the GS (575?) and the old G80-GTS (510?)
have more oc'ing headroom; the old G80 also has better cooling. G92 is also
65nm, and with their lower clocks I think they'll clock much higher in %-s.
-
Quote:
I am trying to point out that at 3Ghz a C2D is not pushing the new 9600GT to it's limit and at that speed it has absolutely no hope what so ever for pushing SLI 9600GT's even close to their limit.
yeah cpu is a limiting factor for sli in SOME games....or is that most games?
im assuming graphics power limits crysis performance (but that is a guess, but is it also cpu bound? i dunno)
sli cards are useful when AA and AF are on and higher res just to make a generalisation....
the higher the grfx settings the less cpu bound you become.
My medieval 2 needs more cpu power :( AND graphics power...so in this case i'd say that it is grfx AND cpu bound
9600gt looks like an appealing cheapy dual card idea;any results around for tri sli with 3x 9600gt?
yes 9600gt with decent cooler will appear surely from some vendor or other i'd have thought.
...all these cards are quite capable of running majority of games at decent settings; the difference only comes in at res above 16x12 and or with higher aa and af.
2x9600gt costs a mere 100 ish aussie bucks less than an 8800gtX, and i believe, as has been proven in various threads, that the GTX is the better performer still :)
so i say false economy sli-wise(but im forgetting the 24-30inch res dudes, they probly need the gfx powa), but a nice little perfomer in a single card setup to be sure.
so, it's all just more of the same.
not even tri sli will dig you out of the crysis hole in my opinion, but i havent seen figures of that yet....well cos 3 x 9600gt's are probly roughly equal in price to 1 single ultra; someone prolly already knows which one will win in a 9600gt x 3 (...780i boardy...) vs 1 x ultra...?
IT IS NEVER ENUF....so you spend a fortune on multigpu crapola :lol:
OR IT IS TOO MUCH...and you tweak your settings and live with a cheaper grfx solution, get great fps and pretty good grfx quality;) :)
