But in 4x0 bench, CPU is at 4.2Ghz and we know 3D Mark 06 is quite sensitive to CPU speed... though if true, idle temps are good and nowhere near what Charlie claimed.
NVIDIA GF100 & Friends speculation
- Thread starter Arty
- Start date
But in 4x0 bench, CPU is at 4.2Ghz and we know 3D Mark 06 is quite sensitive to CPU speed... though if true, idle temps are good and nowhere near what Charlie claimed.
Did you notice fan speed?
yep, 79% but still at that speed 34 degrees seems pretty good. Considering that Charlie claimed GTX480 is going to idle at 70 degrees.
I must say I am not too familiar with GPU cooling, but I assume that if these temps are true, lowering fan speed to 30% should not raise temp by more than 15 degrees.
For a chip like Fermi with 3billion transistors, 51 degrees at idle with 40% fan is ok IMHO.
I found the original picture, no need to zoom...
No it's not.
H: 2.30
L: 2.30
Pixel for pixel its not a perfect square and the width is a little bit smaller 5 to 6 pixels, so hard to say what the size is without a clearer picture.
79% at idle is too high, whichever way you slice it.Did you notice fan speed?
Unless it was coming down after just having run a benchmark for a while...
Did you notice fan speed?
Yeah, and why do you set your fan to almost 80% when your idle temp is 34 deg? Overclock? Also GPU-Z seems to indicate its 512 cores part so maybe GTX 480? (or fermi chips are still not properly recognized). The test seems to be run at 1920x1200 so the score seems to be very, very good (i am not familiar with 3dmark06 results but aren't all the records done at 1280x1024? - hence the cpu dependency?) .
Also CPU-Z screen shows i7@4,2GHz yet 3DM06 shows its only @4GHz.
Also if you see at the everest screen there is GPU name in Chinese. Maybe someone can translate that?
If the fan is set to auto it wont spin @ 80% when chip is at 34degs. Unless its a bug/bad temp reading from software or fan is set manually to 80%.
Heh, HD4890 class theoretical texture throughput. It's really too bad they had to disable half of those texture units eh neliz?
I didnt quite understand if you were being sarcastic or not.. But that score is really bad.. a GTX285 gets 50k on multi texture fill rate..
And a HD5870 60k! I didnt quite understand if you were being sarcastic or not.. But that score is really bad.. a GTX285 gets 50k on multi texture fill rate..
I thought it was clear that it was a terrible result. No sarcasm in sight
Heh, HD4890 class theoretical texture throughput. It's really too bad they had to disable half of those texture units eh neliz?
yeah, about 2/3rd of a GTX285, whodathunkit?
Silent_Buddha
Legend
Isn't the fan also twice as powerful (amp draw) and smaller than the 5870 fan? Not sure how it compares to the GTX 28x fans.
I'd hate to try to imagine what 80% fan speeds sounds like on that sucker. Has to be manually set for overclocking or to give the impression to casual observers that temperature isn't that bad. I can't imagine an IHV now days would spec their cards to run the fans that high.
Regards,
SB
I'd hate to try to imagine what 80% fan speeds sounds like on that sucker. Has to be manually set for overclocking or to give the impression to casual observers that temperature isn't that bad. I can't imagine an IHV now days would spec their cards to run the fans that high.
Regards,
SB
I thought it was clear that it was a terrible result. No sarcasm in sight
Ok thanks for the clarification. I wonder if the rumoured B1 will reactivate the other 64 TMUs, if they exist.. That could give a pretty nice performance boost, if the GPU is indeed limited by them
yeah, as long as you ignore that Nvidia needed +6 months, and probably + ~50% die area and + ~50% power consumption to match something that AMD - according to neliz - could easily release as regular 5890, if they chose to.
It's nothing without the room temperature as well.
Similar threads
- Deleted member 2197
- Architecture and Products
