Turing Review: GTX 1660 Super and GTX 1650

[Paragon]

Hi everyone

I just finished my latest set of benchmarks on the entry level Turing architecture cards. In summary, on Windows 10:

GTX 1660 Super: 660K PPD (seeing 750K in Linux but that is only informally tested so far)
GTX 1650: 310K PPD

Not too bad for these little cards. Anyone else seeing similar numbers on these?

Full article is below, complete with power consumption and efficiency plots.

https://greenfoldingathome.com/2020/05/ ... ed-review/

[bikeaddict]

My GTX 1650 Super (MSI Gaming X GDDR6) on Linux has averaged 403K PPD over the last week with a daily high of 567K. Monitoring the power use in the GreenWithEnvy app shows a draw between 75-82W much of the time with rare peaks over 90W.

[JimboPalmer]

I run a MSI Low Profile GTX 1650 without a power plug. 280k to 290k PPD on Core_22. I expect that it will be slower when Core_21 comes back.

Low Profile is never the fastest, but they fit the PCs I have.

[Mxyzptlk]

Thanks for the write up! I see that you are using MSI Afterburner to throttle the power to the GPU. What is the best way to go about finding the best PPD for a card that you have found. I have a 2060, 2060 Super, and a 2070 Super I would like to find there sweet spots. Thanks.

[Paragon]

The best PPD is usually found by overclocking or pushing the power slider above 100 percent. For efficiency, you go the other way. Usually I step the power down in increments of ten percent and let it do a bunch of work units (a weeks worth). You need to get a large sample size of work units to understand what your settings are doing, since the PPD varies a lot between work units. Collect a weeks worth of data, adjust settings, collect data, average, repeat...

For PPD/ Watt optimization, I usually find the lowest power setting to be the best

[Mxyzptlk]

I have been monitoring and recording my projects and the speeds. Do the card manufacturers post a 'lowest' power setting? or is that found by trial and error? What happens when the card power is to low?

[VegaZhree3]

Hey that's my card!

I bought a GTX 1660 Super last month, just to start folding. It's my first GPU btw.

I'm a newbie folder, so i didn't monitor it properly via an app, i'm just writing what i saw from the client. and just saw the HFM.net thread. But when on full power the PPD varies between* 620-860K. Of course it depends on the project.

The PC is in the room that i sleep, so i draw the power slider and temp limit to the lowest for the least heat and noise output before going to bed. The GPU draws around 70W and the PPD varies between* 550-720K.

Using Windows 10 and the GPU runs at 1950 MHz. (+75 MHz on MSI Afterburner). I remove it when i'm going to sleep.

The model i use is this: https://www.msi.com/Graphics-card/GeFor ... cification

Has really good cooling and quiet. My case doesn't have a good airflow but it never exceeds 70C and fans hover around %45. It's better when the power is limited, it runs around 60C and fans run at %25 which is the lowest possible.

[JimF]

I run a GTX 1660 Ti on Linux. It gets about the same PPD as a GTX 1070; sometimes slightly more (base models, not overclocked).

[BobWilliams757]

Great review Paragon. I'm strongly considering picking up a 1650 Super myself, as it should fall right between both of these cards fairly well. All great bang for the buck on return and power consumption. And it's really cool to see that people such as yourself are looking at contributing in a green way when possible.


Nuclear Wessels.... love it. I might have to join!

[Endgame124]

Hi everyone

I just finished my latest set of benchmarks on the entry level Turing architecture cards. In summary, on Windows 10:

GTX 1660 Super: 660K PPD (seeing 750K in Linux but that is only informally tested so far)
GTX 1650: 310K PPD

Not too bad for these little cards. Anyone else seeing similar numbers on these?

Full article is below, complete with power consumption and efficiency plots.

https://greenfoldingathome.com/2020/05/ ... ed-review/

Quick heads up - your table at the start of the article says the 1660 super has gddr 5, but it’s actually gddr 6.

Edit:
EVGA precision X allows for a wider range of clock control than MSI afterburner - you can take the memory clock 2x lower on precision X. Given the 1660 super has tons of bandwidth due to the gddr6, my non scientific results show that you can clock down the ram to -1000, and leave the GPU clock higher and get the best mix of ppd per watt.

[MeeLee]

Great review Paragon. I'm strongly considering picking up a 1650 Super myself, as it should fall right between both of these cards fairly well. All great bang for the buck on return and power consumption. And it's really cool to see that people such as yourself are looking at contributing in a green way when possible.


Nuclear Wessels.... love it. I might have to join!

At 90W it's not really a good card, if you know you can get around 970k PPD on most RTX 2060 GPUs, when capping their power to 125-130W. That's 50% more power, for 3-4x more PPD.

Personally I wouldn't look at anything below a 1660 Super / 1660Ti when buying new.

[Paragon]

I have been monitoring and recording my projects and the speeds. Do the card manufacturers post a 'lowest' power setting? or is that found by trial and error? What happens when the card power is to low?

It seems to vary between cards. 60% is a typical minimum setting I've seen, but some can go a bit lower, some not so much. I'm not sure why this is.

[Paragon]

Hi everyone

I just finished my latest set of benchmarks on the entry level Turing architecture cards. In summary, on Windows 10:

GTX 1660 Super: 660K PPD (seeing 750K in Linux but that is only informally tested so far)
GTX 1650: 310K PPD

Not too bad for these little cards. Anyone else seeing similar numbers on these?

Full article is below, complete with power consumption and efficiency plots.

https://greenfoldingathome.com/2020/05/ ... ed-review/

Quick heads up - your table at the start of the article says the 1660 super has gddr 5, but it’s actually gddr 6.

Edit:
EVGA precision X allows for a wider range of clock control than MSI afterburner - you can take the memory clock 2x lower on precision X. Given the 1660 super has tons of bandwidth due to the gddr6, my non scientific results show that you can clock down the ram to -1000, and leave the GPU clock higher and get the best mix of ppd per watt.

Good catch, I'll fix that. And thanks, I think I used EVGA precision once way back and it worked fine. Will have to check it out again.

[VegaZhree3]

Edit:
EVGA precision X allows for a wider range of clock control than MSI afterburner - you can take the memory clock 2x lower on precision X. Given the 1660 super has tons of bandwidth due to the gddr6, my non scientific results show that you can clock down the ram to -1000, and leave the GPU clock higher and get the best mix of ppd per watt.

Can you share a stable overclock setting for 1660 Super on MSI Afterburner or EVGA Precision X? I don't want to break WUs for it. Also how much PPD you get with the overclock?

[HaloJones]

Can you share a stable overclock setting for 1660 Super on MSI Afterburner or EVGA Precision X? I don't want to break WUs for it. Also how much PPD you get with the overclock?

my stable overclock is very likely different to yours. there really isn't a reusable overclock guaranteed to work for every card and setting and software demands