nVidia teasing us with GTX480 maybe a GTX500 in works!

[Elixer]

This. As TSMC's yields go up, we should eventually see a 512-SP GTX 485 part become available, methinks. Or just a 480 "512 edition," with potential optimizations leading to a 485 down the road.

You are assuming TSMC can work out all the issues they have...
TSMC burned both ATI & Nvidia... I bet they can't wait to dump them.

 

[Dark Shroud]

Fermi is Nvidia's first foray into specifically building supercomputers. It's thus more than just a competitor to the 5000 series, which had predictable graphics performance.

Nvidia widened its market focus, so you should widen your perspective.

Like back in the day when nVidia bad mouthed tessellation because other brand could do it & they could not. Or when nVidia essentially talked MS from leaving tessellation out of DX10.

Or the in ability of one nVidia card to power three monitors. The power usage of just powering two monitors is astounding.

GPGPUs are nice and all and nVidia is ahead of the curve with Cuda. But OpenCL is coming and AMD & probably Intel will support that open standard

PhysX isn't used by many games let alone good ones. AMD is actively working with Havoc & Bullet so the other more open & more used physics engines will have very good ATI support. Meanwhile nVidia blocks PhysX hardware accelration if you have an ATI card present.

But all of this means nothing to normal people who just want to play a game on an average level PC.

They won't even be able to run a GTX 400 series without a new PSU & stronger cooling options.

 

[postmortemIA]

now I know why Gates says we'll need personal nuclear reactors at home

 

[Attic]

If there is 485GTX (512sp version) what are it's load temps and power draw going to be?

480GTX looks like a decent upgrade for nVidia fans come from GTX285 or below, i'd just think that water cooling is going to be essential if noise and/or thermals are a concern for the user.

I would not want to stick a 480GTX one of my rigs without getting a handle on the noise it makes at load. I would love to use one on watercooling though, the thing is fast and I like that.

 

[darXoul]

Nah man that's not fair, it's 2 v 1 GPUs You gotta compare it with GTX480 SLI!

That's not fair, either. You're trying to compare a dual GPU single card to a dual card setup costing over 40% more and drawing almost twice as much power, depending on circumstances.

The 5870 was often compared to GTX 295 despite the latter being SLI slapped on one card (or sandwiched) so until nVidia has a dual Fermi card (which, based on 480 experience, is pretty hard to imagine now), the 480 will be compared to the 5970, and for a good reason. Both are single cards, not requiring SLI/XF boards, and both are in the same league of power draw and noise. The 5970 is more expensive and prone to multi-GPU issues but it's also clearly faster than the new GeForce. The 5870 is a single GPU card not much slower than the 480 on average but much cheaper and "docile" in terms of heat/noise/power so if this is a fair comparison, so is the 480 vs. 5970 one.

 

[BFG10K]

There’s no way nVidia can get a 512 SP GTX490 out the door with the current state of the GTX480. Maybe in six months after a respin/refresh, but not now.

As for ATi, I can’t see why they aren’t going with Global Foundries as fast as possible and dropping TSMC. GF is ATi’s spin-off, and they seem to have more advanced process manufacturing too.

 

[Dribble]

Its unlikely NV can fix the power usage of Fermi with "a quick refresh".
It will almost certainly take an architecture change.
I imagine the Catalyst 10.3 drivers ATi pulled out of the magic box also caught NV by surprise.

I agree, don't see a significantly better version till they move to 28nm. That said I expect it to get faster - the 58xx series is now 6 months old and very closely based on the 48xx anyway so drivers are gonna be close to optimal. You've got to expect pretty big improvements for the GTX 4xx cards as they are brand new with a brand new architecture. That's probably significantly more important then unlocking another 6.7% cores.

As for ATi, I can’t see why they aren’t going with Global Foundries as fast as possible and dropping TSMC. GF is ATi’s spin-off, and they seem to have more advanced process manufacturing too.

Do they have 40nm? No, with them you get 45nm and future promises, with TSMC you get 40nm and future promises. Global foundries is a good option, but it's not in front of TSMC when it come to manufacturing. Equally if you think you have better hardware almost better to cover your competitor by using same manufacturing, then risk going somewhere else.

 

[dug777]

I agree, don't see a significantly better version till they move to 28nm. That said I expect it to get faster - the 58xx series is now 6 months old and very closely based on the 48xx anyway so drivers are gonna be close to optimal. You've got to expect pretty big improvements for the GTX 4xx cards as they are brand new with a brand new architecture. That's probably significantly more important then unlocking another 6.7% cores.

While I hope you are right, I believe they have actually had functioning silicon for quite a while now, and I can't imagine that they haven't been beating the living hell out of the driver team for the last couple of months to scrape every possible ounce of fps for launch...

 

[Dribble]

While I hope you are right, I believe they have actually had functioning silicon for quite a while now, and I can't imagine that they haven't been beating the living hell out of the driver team for the last couple of months to scrape every possible ounce of fps for launch...

In that time to just getting it with Sli configurations, physx, and 3d running every game out there at acceptable fps without crashing or artefacts would be a huge win for the driver team.

 

[dug777]

In that time to just getting it with Sli configurations, physx, and 3d running every game out there at acceptable fps without crashing or artefacts would be a huge win for the driver team.

One wonders what would have happened if they had been 'on time' to market then? None of the rumours suggested that software issues lay behind the delays, but I guess anything is possible when you are dealing with rumours

 

[Sylvanas]

I agree, don't see a significantly better version till they move to 28nm. That said I expect it to get faster - the 58xx series is now 6 months old and very closely based on the 48xx anyway so drivers are gonna be close to optimal. You've got to expect pretty big improvements for the GTX 4xx cards as they are brand new with a brand new architecture. That's probably significantly more important then unlocking another 6.7% cores.




Do they have 40nm? No, with them you get 45nm and future promises, with TSMC you get 40nm and future promises. Global foundries is a good option, but it's not in front of TSMC when it come to manufacturing. Equally if you think you have better hardware almost better to cover your competitor by using same manufacturing, then risk going somewhere else.

GF has a more advanced process than TSMC. SOI and power gating with 28nm (or was it 32 I can't remember) outweighs the bulk process offered by TSMC and UMC. Idontcare had an awesome post about this on the CPU forums a while back, I should have bookmarked it.

 

[Genx87]

While I hope you are right, I believe they have actually had functioning silicon for quite a while now, and I can't imagine that they haven't been beating the living hell out of the driver team for the last couple of months to scrape every possible ounce of fps for launch...

I'd guess they have had functioning silicon since last Summer. But not at the speeds and possibly even the configuration that is shipping right now. I expect is the few games Fermi lost significantly to the ATI part. In the coming months drivers will close that gap in performance.

 

[geoffreymac]

There’s no way nVidia can get a 512 SP GTX490 out the door with the current state of the GTX480. Maybe in six months after a respin/refresh, but not now.

As for ATi, I can’t see why they aren’t going with Global Foundries as fast as possible and dropping TSMC. GF is ATi’s spin-off, and they seem to have more advanced process manufacturing too.

I wonder if such a major Fab like TSMC is having issues, maybe there is an issue with the process itself and not the Fabs fault ?, I am assuming the process technology between CPU's and GPU's to be the same (in theory) ? We have a 45nm and 32nm nodes, so why not head towards that (I know currently Intel currently owns the only Fabs that can do it), but for someone like AMD/ATI who use to develop there own in house Fabs, why not keep the same process ? They do overlap on occasion, but seems GPU's have odd numbers. A Core i7 and a Fermi cannot be that diffrent in transistor count ?,

or am I off the mark ?

 

[WicKeD]

They won't even be able to run a GTX 400 series without a new PSU & stronger cooling options.

The user who is buying Fermi most likely has a > 600 watt PSU.

 

[KCfromNC]

If there is 485GTX (512sp version) what are it's load temps and power draw going to be?

Probably similar to a 480 i.e. near the safe limits of the chip - but they'll have to lower the clock speeds to get that to happen.