GTX580 appears on nvidia website

[Zildjian]

This is from Wiki

"It has been reported by users as well as developers that there exists severe performance problems with the nVidia 400-series cards on 3D content-creation applications such as Maya, 3ds Max, Blender, Rhinoceros as well as some OpenGL games, to the extent that video cards two generations older routinely outperform 400-series in such applications and games. The problem, which affects any OpenGL application using textures, accesses framebuffer contents or stores data on the GPU. So far, nVidia has declined to acknowledge or comment on the issue."


Anyone with the 4xx series wanna see how true this is?

as a 3dsmax user i can say that GTX 480 performs better than a GTX 285 in viewport rendering. Sure, it's not "Quadro performance" but still acceptable for normal workloads.

You can also change viewport from OpenGL to DirectX in case there are some errors or performance is slow.

Hopefully i can get a 580 soon so i can bench it against my 285 in 3dsmax.


ps: Unfortunately it's not possible to softquadro anymore :) ( unwinder please :D )

[dragomirc]

I just noticed in this source, which makes me a little iffy on how more rumor like it is
1.) "By activating all 16 SM units GTX 580 will also gain access to more texture units and ROPs."
- There are no more ROPS to gain
2.) "According to sources to NordicHardware it can be as many as 300 million transistors that NVIDIA has been able to cut in this way."
- So they really have no idea, if any transistors have been cut away.

Well how do you explaining lower TDP considering higher clock.

The card should end up clocked at 772MHz for the GPU, 1544MHz for Shaders and 4.0Gbps for 1536MB of GDDR5 memory paired up with a 384-bit interface. Another interesting piece of information is the actual TDP of the new card. Although it has more CUDA Cores and a higher GPU clock, the card has a slightly lower TDP of 244W. We still don't know if Nvidia is still talking about typical load power, as the GTX 480 or GF100 had 250W TDP, which was actualy a bit closer to 300W under full load.

[Godz]

I'm trying to locate on what is rumored as to how they were going to lower TDP. They will be removing some of the abilities, but I just can't remember what they were off hand. The abilities wouldn't effect gameplay or IQ. I read it couple days ago, just don't remember where I saw it.

EDIT: http://vr-zone.com/articles/report-n...ked/10202.html
I'll just toss that in the before heading back to work which basically states that the GTX 580 will have an approximate TDP of 244 watts.

[dragomirc]

I'm trying to locate on what is rumored as to how they were going to lower TDP. They will be removing some of the abilities, but I just can't remember what they were off hand. The abilities wouldn't effect gameplay or IQ. I read it couple days ago, just don't remember where I saw it.

EDIT: http://vr-zone.com/articles/report-n...ked/10202.html
I'll just toss that in the before heading back to work which basically states that the GTX 580 will have an approximate TDP of 244 watts.

I posted before that 300 mil transistors cut would be on the cost of HPC functionality (High Performance Computing).
That might be a answer why 10% higher GPU clock run with lower power use.

[Godz]

I posted before that 300 mil transistors cut would be on the cost of HPC functionality (High Performance Computing).
That might be a answer why 10% higher GPU clock run with lower power use.

Yup that was it, not from your post, but I did see that on another website someplace. Still, depending on what site you read, it's going either way. Some are saying it will use slightly less power, or slightly more. Still with a 15%-20% increase in performance. It's being predicted that AMD 6900 series will hold the crown once released. Time will tell though. I'm sure last minute changes are being made.

[kitch9]

Where's the evidence that AMD are 'weak' in D3D11? If you're referring to the use of tessellation in games and benchmarks, that has almost nothing to do with D3D11 itself. It's almost entirely down to the geometry throughput of the chip and in this area, NVIDIA are definitely well ahead of AMD.

Quad thinks DX11 is just millions of sub-pixel triangles.....

[Noizic]

When it comes to the bus, it's been like around 256-bit for quite some time. Is it because it's not something that's a bottleneck on a GPU so far? I mean, more CUDA/steam cores and higher core clock seems to help the most but I would assume it depends on the situation. Oh and not to mention to avoid hitting the 300W-wall, higher bus equals higher TDP or?

[harlequin]

thats not strictly true

http://www.neeyik.info/3dspecs/3dspe....html#nvidia11

GTX 260 and upwards are on 448 or 512 , whilst the GTX480 is on 348 and 470 is on 320

im very sure neeyik can give a great answer as to the `why` , but 1 thing is chip real estate - 512bit bus costs quite abit of space

[Noizic]

thats not strictly true

http://www.neeyik.info/3dspecs/3dspe....html#nvidia11

GTX 260 and upwards are on 448 or 512 , whilst the GTX480 is on 348 and 470 is on 320

im very sure neeyik can give a great answer as to the `why` , but 1 thing is chip real estate - 512bit bus costs quite abit of space

Considering the HD6870 and HD5870 are both 256-bit I thought it was something it's not really needed. The GTX460 768MB has 192-bit bus but the difference between the 1GB version was even lower than I thought.

[QuadDamage]

thats not strictly true

http://www.neeyik.info/3dspecs/3dspe....html#nvidia11

GTX 260 and upwards are on 448 or 512 , whilst the GTX480 is on 348 and 470 is on 320

im very sure neeyik can give a great answer as to the `why` , but 1 thing is chip real estate - 512bit bus costs quite abit of space

512-bit bus requires additional memory controllers which in turn makes the die bigger and Fermi is big enough as it is. GTX480 is 384-bit by the way.