VOGONS

Common searches


First post, by clueless1

User metadata
Rank l33t
Rank
l33t

I just noticed that TechPowerUp has incorporated GPU power consumption results with VSYNC enabled:
https://www.techpowerup.com/review/msi-radeon … -x-trio/35.html

This really shines a light on a new metric and it's something I've personally played with in my own benchmarking for about a year now. When you have a system with a limited PSU (mine is not upgradeable), it changes how you shop for graphics cards. Even if you have a good PSU but prefer a lower energy footprint, these results are interesting.

Last year when I was shopping for a GPU that would work well in my system (325W PSU), I settled on the 1650 Super as it was the fastest card available with a 100W TDP. These new results now show which more powerful GPUs will work well in my system with a 60 fps cap, including several RTX 20xx and RTX 30xx GPUs.

I'm just glad someone is now doing this, it's great information that as far as I know, no other GPU review site has ever utilized on a mass scale. Bravo to TechPowerUp!

edit: really impressed with the RTX 3070 taking the top spot with only 73W used during gaming at 60 fps (compared to 220W with VSYNC off). None of the AMD cards come close to this kind of efficiency and it even beats the crap out of my 1650 Super, which drops from 106W to 104W when turning VSYNC on.

The more I learn, the more I realize how much I don't know.
OPL3 FM vs. Roland MT-32 vs. General MIDI DOS Game Comparison
Let's benchmark our systems with cache disabled
DOS PCI Graphics Card Benchmarks

Reply 1 of 13, by badmojo

User metadata
Rank l33t
Rank
l33t

That is a great metric - some big differences there. I have a 144Hz monitor so have been turning vsync off and going for the highest frames possible but I capped a game at 70FPS the other day because the GPU was sitting at 100% and the fan noise was distracting, which I guess means I was wasting power too!

Life? Don't talk to me about life.

Reply 2 of 13, by clueless1

User metadata
Rank l33t
Rank
l33t
badmojo wrote on 2020-12-30, 07:06:

That is a great metric - some big differences there. I have a 144Hz monitor so have been turning vsync off and going for the highest frames possible but I capped a game at 70FPS the other day because the GPU was sitting at 100% and the fan noise was distracting, which I guess means I was wasting power too!

Yeah, that's another use case - reducing noise and/or heat. You could cap your framerate at 144 so you only lose frames beyond your refresh rate, but depending on the graphics card, it may not matter (if your GPU seldom approaches 144 fps in your games). The sweetest spot is when your GPU can output a consistent framerate equal to your refresh rate.

The more I learn, the more I realize how much I don't know.
OPL3 FM vs. Roland MT-32 vs. General MIDI DOS Game Comparison
Let's benchmark our systems with cache disabled
DOS PCI Graphics Card Benchmarks

Reply 3 of 13, by canthearu

User metadata
Rank Oldbie
Rank
Oldbie
clueless1 wrote on 2020-12-29, 14:49:

Last year when I was shopping for a GPU that would work well in my system (325W PSU), I settled on the 1650 Super as it was the fastest card available with a 100W TDP. These new results now show which more powerful GPUs will work well in my system with a 60 fps cap, including several RTX 20xx and RTX 30xx GPUs.

Does not work this way unfortunately,

With vsync on, the card will go from idle state to full power and then back to idle each time a frame is rendered. So your power supply has to still be able to deliver the full amount of power the video card would otherwise require regardless of vsync being on or off.

Average power consumption is definitely reduced, but peak is largely unaffected and it would be a real risk to significantly under supply the card based on running with vsync.

Reply 4 of 13, by clueless1

User metadata
Rank l33t
Rank
l33t
canthearu wrote on 2020-12-30, 14:10:
Does not work this way unfortunately, […]
Show full quote
clueless1 wrote on 2020-12-29, 14:49:

Last year when I was shopping for a GPU that would work well in my system (325W PSU), I settled on the 1650 Super as it was the fastest card available with a 100W TDP. These new results now show which more powerful GPUs will work well in my system with a 60 fps cap, including several RTX 20xx and RTX 30xx GPUs.

Does not work this way unfortunately,

With vsync on, the card will go from idle state to full power and then back to idle each time a frame is rendered. So your power supply has to still be able to deliver the full amount of power the video card would otherwise require regardless of vsync being on or off.

Average power consumption is definitely reduced, but peak is largely unaffected and it would be a real risk to significantly under supply the card based on running with vsync.

You may be right, but I do remember having my system connected to a watt meter and watching it while using my system with VSYNC locked to 60 fps and I do not recall the watt meter showing any power spikes, but that doesn't mean they aren't happening.

The more I learn, the more I realize how much I don't know.
OPL3 FM vs. Roland MT-32 vs. General MIDI DOS Game Comparison
Let's benchmark our systems with cache disabled
DOS PCI Graphics Card Benchmarks

Reply 5 of 13, by canthearu

User metadata
Rank Oldbie
Rank
Oldbie
clueless1 wrote on 2020-12-30, 15:07:

You may be right, but I do remember having my system connected to a watt meter and watching it while using my system with VSYNC locked to 60 fps and I do not recall the watt meter showing any power spikes, but that doesn't mean they aren't happening.

Mains side power meters do not update at 60hz or higher, and the power draw on the mains side is somewhat smoothed out by all the capacitors on the way through the power supply, and on the video cards. But this does not mean that the cards will not suffer too much voltage drop during the usage spikes if the power supply isn't able to respond with enough output. You would need to measure the 12V lines going directly to the video card to be able to measure power usage properly.

Then you also get frame spikes, where rendering a particular frame is a lot slower than typical, and you would suffer even greater voltage drop on these frames if the power supply cannot keep up. This could result, on the user side, as a system lockup, or even just a crash to desktop from game (As video cards have gotten a lot better at resetting themselves when things go wrong, rather than hard crashing)

Reply 6 of 13, by Silanda

User metadata
Rank Member
Rank
Member

I think this is an interesting thing to measure purely as a reference for building a generally power efficient system, but I wouldn't use it as a guide for choosing GPUs for a lower power PSU. In addition to what canthearu said, the average power draw is going to vary wildly from game to game. I have an RTX 3070, which they rate as drawing 73 W vsynced at 60Hz. Play something like Cyberpunk 2077 or Quake II RTX though and it can pull ~220W sustained while vsynced at 60Hz.

Reply 7 of 13, by clueless1

User metadata
Rank l33t
Rank
l33t
Silanda wrote on 2020-12-30, 19:42:

I think this is an interesting thing to measure purely as a reference for building a generally power efficient system, but I wouldn't use it as a guide for choosing GPUs for a lower power PSU. In addition to what canthearu said, the average power draw is going to vary wildly from game to game. I have an RTX 3070, which they rate as drawing 73 W vsynced at 60Hz. Play something like Cyberpunk 2077 or Quake II RTX though and it can pull ~220W sustained while vsynced at 60Hz.

Interesting. Cyberpunk is what they use in this test:

V-Sync: If you don't need the highest framerate and want to conserve power, running at 60 FPS is a good option. In this test, we run Cyberpunk 2077 at 1920x1080, capped to 60 FPS. This test is also useful in testing a graphic card's ability to react to situations with only low power requirements. For graphics card that can't reach 60 FPS at 1080p, we report the power draw at the highest frame rate they can achieve.

The more I learn, the more I realize how much I don't know.
OPL3 FM vs. Roland MT-32 vs. General MIDI DOS Game Comparison
Let's benchmark our systems with cache disabled
DOS PCI Graphics Card Benchmarks

Reply 8 of 13, by cyclone3d

User metadata
Rank l33t++
Rank
l33t++

Even if you cap a game at 60fps, what happens when you try to run a game that is more demanding and thus the GPU usage is more and requires more power?

All these types of benches are doing is showing what the power draw is with specific games.

Yamaha modified setupds and drivers
Yamaha XG repository
YMF7x4 Guide
Aopen AW744L II SB-LINK

Reply 10 of 13, by ZellSF

User metadata
Rank l33t
Rank
l33t

Lots of review sites do power efficiency. This is just doing it while targeting a very arbitrary performance point. What if you run at a different resolution? refresh rate? game settings? different sync technology? etc...

I would be very surprised if this actually will end up being genuinely more useful to anyone than existing power efficiency metrics. Existing power efficiency metrics aren't better, they're equally useful so what TechPowerUp is doing here isn't stupid, just weird.

Reply 11 of 13, by Silanda

User metadata
Rank Member
Rank
Member
clueless1 wrote on 2020-12-30, 23:19:
Silanda wrote on 2020-12-30, 19:42:

I think this is an interesting thing to measure purely as a reference for building a generally power efficient system, but I wouldn't use it as a guide for choosing GPUs for a lower power PSU. In addition to what canthearu said, the average power draw is going to vary wildly from game to game. I have an RTX 3070, which they rate as drawing 73 W vsynced at 60Hz. Play something like Cyberpunk 2077 or Quake II RTX though and it can pull ~220W sustained while vsynced at 60Hz.

Interesting. Cyberpunk is what they use in this test:

V-Sync: If you don't need the highest framerate and want to conserve power, running at 60 FPS is a good option. In this test, we run Cyberpunk 2077 at 1920x1080, capped to 60 FPS. This test is also useful in testing a graphic card's ability to react to situations with only low power requirements. For graphics card that can't reach 60 FPS at 1080p, we report the power draw at the highest frame rate they can achieve.

Given the variety of cards they are testing, and that some don't support it, I assume that ray tracing was turned off.

Reply 12 of 13, by clueless1

User metadata
Rank l33t
Rank
l33t
Silanda wrote on 2020-12-31, 14:26:
clueless1 wrote on 2020-12-30, 23:19:
Silanda wrote on 2020-12-30, 19:42:

I think this is an interesting thing to measure purely as a reference for building a generally power efficient system, but I wouldn't use it as a guide for choosing GPUs for a lower power PSU. In addition to what canthearu said, the average power draw is going to vary wildly from game to game. I have an RTX 3070, which they rate as drawing 73 W vsynced at 60Hz. Play something like Cyberpunk 2077 or Quake II RTX though and it can pull ~220W sustained while vsynced at 60Hz.

Interesting. Cyberpunk is what they use in this test:

V-Sync: If you don't need the highest framerate and want to conserve power, running at 60 FPS is a good option. In this test, we run Cyberpunk 2077 at 1920x1080, capped to 60 FPS. This test is also useful in testing a graphic card's ability to react to situations with only low power requirements. For graphics card that can't reach 60 FPS at 1080p, we report the power draw at the highest frame rate they can achieve.

Given the variety of cards they are testing, and that some don't support it, I assume that ray tracing was turned off.

I agree with your assumption. Otherwise they couldn't have results for the cards that don't support RT.

In the review comments, there was this exchange regarding the choice of CP for this metric:

Xuper said: W1zaard , can you test Vsync 60hz on any game like Doom/Gear5/Battle V for ref. 6800XT vs 3080? Cybepunk is high […]
Show full quote

Xuper said:
W1zaard , can you test Vsync 60hz on any game like Doom/Gear5/Battle V for ref. 6800XT vs 3080? Cybepunk is highly optimized for nvidia cards.

Cyberpunk is the most popular game by far. AMD has optimized their driver for CP, too. 6800 XT is just as fast as 3080.

These 3 games run WAY too high FPS for 60 FPS V-Sync to give useful results. Any other game, people will ask "what about Cyberpunk?"

The more I learn, the more I realize how much I don't know.
OPL3 FM vs. Roland MT-32 vs. General MIDI DOS Game Comparison
Let's benchmark our systems with cache disabled
DOS PCI Graphics Card Benchmarks

Reply 13 of 13, by clueless1

User metadata
Rank l33t
Rank
l33t
canthearu wrote on 2020-12-30, 15:39:
clueless1 wrote on 2020-12-30, 15:07:

You may be right, but I do remember having my system connected to a watt meter and watching it while using my system with VSYNC locked to 60 fps and I do not recall the watt meter showing any power spikes, but that doesn't mean they aren't happening.

Mains side power meters do not update at 60hz or higher, and the power draw on the mains side is somewhat smoothed out by all the capacitors on the way through the power supply, and on the video cards. But this does not mean that the cards will not suffer too much voltage drop during the usage spikes if the power supply isn't able to respond with enough output. You would need to measure the 12V lines going directly to the video card to be able to measure power usage properly.

Then you also get frame spikes, where rendering a particular frame is a lot slower than typical, and you would suffer even greater voltage drop on these frames if the power supply cannot keep up. This could result, on the user side, as a system lockup, or even just a crash to desktop from game (As video cards have gotten a lot better at resetting themselves when things go wrong, rather than hard crashing)

Good points, and I just paid more attention to the graph of the 6800XT at the top of the linked page. Look at the VSYNC graph on the far right...141W average is right down the middle of that thick band that goes from ~100W to ~180W. There's a little spike that goes up to about 220W and down to about 70W. So your point is made and I stand corrected. But I still think it is a useful metric. If you can put less stress on your GPU and PSU just by enabling VSYNC, then why not? Of course, if you have a high refresh rate monitor, then that's another story. But for those of us who have 60Hz displays, it does apply.

The more I learn, the more I realize how much I don't know.
OPL3 FM vs. Roland MT-32 vs. General MIDI DOS Game Comparison
Let's benchmark our systems with cache disabled
DOS PCI Graphics Card Benchmarks