NVIDIA GeForce GTX 970 Offers Great Linux Performance

[schmidtbag]

I wonder why in a lot of the benchmarks, the 780ti is shown to be a little bit faster than the 980 by a few FPS. Strange ... wasn't Maxwell supposed to be faster?

In terms of pure numbers, the 780Ti is a superior GPU in almost every way. But the design, drivers, and fabrication of the 980 gives it superior performance in many cases.

[tajjada]

In terms of pure numbers, the 780Ti is a superior GPU in almost every way. But the design, drivers, and fabrication of the 980 gives it superior performance in many cases.

But in the end of the day, how is design or any of that other stuff relevant if it can't deliver numbers in performance? What use is a card with "superior design, drivers, and fabrication" if it can't deliver FPS in games? Isn't that the whole reason why we buy graphics cards in the first place?

The only advantage I really see is power consumption. But then, could one overclock the 980 to the point where it has the same power consumption as a 780ti but obviously much superior performance? It would produce equivalent heat and power consumption in that case. (sorry if I am being stupid)

[artivision]

In terms of pure numbers, the 780Ti is a superior GPU in almost every way. But the design, drivers, and fabrication of the 980 gives it superior performance in many cases.

Simple: "780ti=2800cores*900mhz" wile "980=2000cores*1200mhz". "ti" is 5% stronger but has 5% less FPS because of RAM and temperature difference since "980" consumes half the watts. In comparison with AMD cards, because GTX from Fermi to Maxwell has half the Cuda cores 64bit length, 2000 NV shaders are like 3000 AMD shaders or 3500 if you add the Mhx difference to. In comparison with non CNG AMD GPUs that they didn't have FMAC, its like 4500 shaders at the same frequency.

[schmidtbag]

But in the end of the day, how is design or any of that other stuff relevant if it can't deliver numbers in performance? What use is a card with "superior design, drivers, and fabrication" if it can't deliver FPS in games? Isn't that the whole reason why we buy graphics cards in the first place?

I don't get what you're asking - the 980 outperformed the 780Ti in many cases, so it does deliver the performance. But much like everything hardware related in the PC world, there is no such thing as "one size fits all". This is why it took over 20 years for intel to make a decent mobile platform, or why AMD's CPUs are very good at a small handful of things and utter crap at everything else, or why ARM scales up horrendously.
Anyway, once there is a 980Ti, I'm sure nothing will defeat it. The 980 was just the introduction to the new architecture.

The only advantage I really see is power consumption. But then, could one overclock the 980 to the point where it has the same power consumption as a 780ti but obviously much superior performance? It would produce equivalent heat and power consumption in that case. (sorry if I am being stupid)

If you were to overclock the 980 to match the power consumption of the 780Ti, I think the 980 would often exceed the 780Ti where it otherwise lags but I'm sure there are some situations where the 780Ti will still outperform. The 780Ti has significantly more texture mapping units, shading units, and a wider memory bus. These can have a big role depending on what is being tested.

[b15hop]

Wow, roughly $100 less than the 780Ti and it outperforms that in many cases. I don't remember if that holds true in Windows, but at least in linux it's a clear choice that the 900 series is good to go for. Kind of funny how ALL 970s are completely sold out on newegg.

I'm concerned that the only difference is some clever driver updates. On paper the older GPU has more grunt...

[b15hop]

I wonder why in a lot of the benchmarks, the 780ti is shown to be a little bit faster than the 980 by a few FPS. Strange ... wasn't Maxwell supposed to be faster?

In terms of pure numbers, the 780Ti is a superior GPU in almost every way. But the design, drivers, and fabrication of the 980 gives it superior performance in many cases.

More reason to believe that it's just some driver and firmware optimisation fixes with a less powerful GPU. The new GPU won't ever be overclocked. Nvidia has to make money so all they will do is release a new GPU that is clocked higher that supersedes the 980gtx.

Simple: "780ti=2800cores*900mhz" wile "980=2000cores*1200mhz". "ti" is 5% stronger but has 5% less FPS because of RAM and temperature difference since "980" consumes half the watts. In comparison with AMD cards, because GTX from Fermi to Maxwell has half the Cuda cores 64bit length, 2000 NV shaders are like 3000 AMD shaders or 3500 if you add the Mhx difference to. In comparison with non CNG AMD GPUs that they didn't have FMAC, its like 4500 shaders at the same frequency.

It's possible that the newer GPU is cooler because it is on a smaller micron... This allows the newer GPU to use half the Wattage. But cutting down the amount of overall power was also done by having less cuda cores... Which makes me think this will also be very game specific. Some games like to have more cores while others like to have faster GPU RAM and core speed..

[schmidtbag]

I'm concerned that the only difference is some clever driver updates. On paper the older GPU has more grunt...

The GPU is a brand new architecture. Of course drivers are part of it, but they're not the only thing.

[hooly]

n00b alert!

pls excuse the somewhat OT question but ...

can anyone recommend installation procedures for an Asus Strix GTX970 on a Mint17 (w i7-4770) system?

trying to install the NVidia driver (from download) without the card in place fails from the get-go ("no NVidia GPU on your system ..."). as soon as i plug the card in i get no video output whatsover, mobo or GPU. without a boot menu can't do much.

seems like a bit of a Catch-22 situation.

[Kano]

Seems you forgot to connect the PCI-E power connectors.

[hooly]

Seems you forgot to connect the PCI-E power connectors.

no, the power _was_ connected.

turned out that the trick was to tell the driver installation to proceed in spite of the failed pre-conditions (no card present).

did that then installed the card and voila! up and running!

ran some of the Phoronix benchmarks, results came in pretty close to Michael's, +/- 1-2%. pretty satisfied with that especially since it was straight out of the box.