NVIDIA GeForce GTX 1660 Linux Benchmarks

[phoronix]

Phoronix: NVIDIA GeForce GTX 1660 Linux Benchmarks

Last week NVIDIA announced the GeForce GTX 1660 as the newest RTX-less Turing GPU but costing only $219+ USD. The GTX 1660 is a further trimmed down version of the GeForce GTX 1660 Ti that launched several weeks prior. After picking up an ASUS GeForce GTX 1660 Phoenix Edition, here are Linux OpenGL/Vulkan gaming benchmarks compared to a wide assortment of AMD Radeon and NVIDIA GeForce graphics cards under Ubuntu.

http://www.phoronix.com/vr.php?view=27693

[vipor29]

not sure what is going on with hitman but i know for a fact a 1660 does not beat a 1080.the 1080 mops the floor with a 1660 lol.weird driver issues there.

[parityboy]

Michael

Many thanks for this; nice to see the Vega 56 is in fact becoming better and and better value for consumers as the price comes down.

A couple of questions:
1) What happened to the Vega 64? It's listed on the system setup page in the output box for PTS.
2) On page 3, why is the chart for F1 2018 1440p duplicated? Was the second one supposed to be 4K?
3) Why are some games benched at 1080p but not 1440p, and vice versa? Is this a PTS thing?

[Michael]

Michael

Many thanks for this; nice to see the Vega 56 is in fact becoming better and and better value for consumers as the price comes down.

A couple of questions:
1) What happened to the Vega 64? It's listed on the system setup page in the output box for PTS.
2) On page 3, why is the chart for F1 2018 1440p duplicated? Was the second one supposed to be 4K?
3) Why are some games benched at 1080p but not 1440p, and vice versa? Is this a PTS thing?

Only Vega 56 was tested, wasn't going all the way to high-end comparison in this article. In the system table it shows as RX Vega 56/64 based upon the driver entries / PCI ID database.

Whoops, will get that F1 2018 chart fixed.

Resolutions/settings were chosen for trying to get around the maximum playable on GTX 1660 class hardware. On some games 1440p runs nicely while other more demanding games 1080p is more appropriate.

[chui2ch]

Only Vega 56 was tested, wasn't going all the way to high-end comparison in this article. In the system table it shows as RX Vega 56/64 based upon the driver entries / PCI ID database.

Whoops, will get that F1 2018 chart fixed.

Resolutions/settings were chosen for trying to get around the maximum playable on GTX 1660 class hardware. On some games 1440p runs nicely while other more demanding games 1080p is more appropriate.

Do you communicate out the extremely poor results for the 1080/1070 in Hitman to Feral?

[skeevy420]

Interesting results. I'm just glad about AMD's fine wine drivers and the 580 being able to compete with a new Nvidia card.

I wonder how much performance would change and what the performance per dollar would be if Michael could undervolt the GPUs and test them.

With my RX 580, simply lowering the last 3 states by 25mv was enough to make it run around 15C lower under a full load which stopped thermal throttling, lowered the fan noise level, and improved its gaming performance.

With almost maxed out settings, at 1080p Metro 2033 Redux went from 60 fps smooth for one minute of game play when throttling would kick in at 75C causing dips 40-45 fps to playing it with 60 fps with dips down to 50-55 fps in the more GPU intense areas while only hitting 65C without needing the fans to kick in full blast.

[parityboy]

Only Vega 56 was tested, wasn't going all the way to high-end comparison in this article. In the system table it shows as RX Vega 56/64 based upon the driver entries / PCI ID database.

Is that in terms of performance, price point or positioning in the relative product stacks? I ask because I notice that the GTX1080 and RTX2070 are in there and they are in the same performance ballpark as the Vega 64. Since the Radeon VII is now AMD's top end GPU and competes with the RTX2080 in terms of price and performance, the inclusion of the RTX2070 and GTX1080 makes the V64 a curious omission.

Resolutions/settings were chosen for trying to get around the maximum playable on GTX 1660 class hardware. On some games 1440p runs nicely while other more demanding games 1080p is more appropriate.

Gotcha.

[aufkrawall]

Michael Would it be possible to always include results of maximum/ultra details at least for 1080p resolution?

[xorbe]

Did I miss a Phoronix post about the new nVidia 418.56 driver dated March 18th? Maybe it's kernel 5.1 related to fix a build issue.

http://download.nvidia.com/XFree86/Linux-x86_64/418.56/

[Hans Bull]

Did I miss the darktable OpenCL benchmark?