![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
Video card (and CPU) comparison websites that include efficiency?
- Thread starter .oO
- Start date
- Joined
- Jun 15, 2005
- Messages
- 2,789
Short answer is no. You can compile the data yourself but it will take a great deal of time and effort. I do a lot of that sort of testing personally so feel free to PM me with your target power budget and resolution/refresh rate. I'm assuming this post is an inquiry for gaming and general computing.
- Joined
- Dec 24, 2008
- Messages
- 2,295
Not sure it's quite what you want,
but techpowerup usually has GPU vsync 60hz locked power usage, as well as efficiency power/frame numbers for their GPU reviews.
https://www.techpowerup.com/review/galax-geforce-rtx-5080-1-click-oc/41.html
https://www.techpowerup.com/review/galax-geforce-rtx-5080-1-click-oc/42.html
but techpowerup usually has GPU vsync 60hz locked power usage, as well as efficiency power/frame numbers for their GPU reviews.
https://www.techpowerup.com/review/galax-geforce-rtx-5080-1-click-oc/41.html
https://www.techpowerup.com/review/galax-geforce-rtx-5080-1-click-oc/42.html
- Joined
- Aug 15, 2005
- Messages
- 9,272
Unfortunately, they eventually stopped doing the watts-per-frame calculation.
And the gaming at 60hz power chart could be a little misleading. If you click the power consumption testing details button, at the top of the page:
It reveals that test is based on Cyberpunk at 1080p/60. Its an interesting, but sort of nebulous data point, in terms of specific usefulness.
serpretetsky,
Thanks, those pages are helpful (I didn't know of those info pages specifically)...But I mean the side-by-side comparisons like HWCompare, CPU-monkey, etc... does. Where you can choose and two or more cards or CPU's to compare in detail. That is particularly useful.
I find it strange that efficiency, which seems to be one of the top most important attributes, is not included in those spec comparisons.
[edited the words in parenthesis above for clarity]
Last edited:
I've seen stuff like this but I don't recall where. I suppose I could try to Google it for you but nah, don't feel like it.
Some bits I remember -
Intel ARC GPUs suck at idle power draw if you have more than one screen attached. Also less efficient under load, at least for gaming.
NV vs. AMD GPUs? I dunno.
Raptor Lake beats Arrow Lake and pretty much anything AMD based on CCDs at idle power draw. Arrow Lake is only slightly worse. That actually makes sense if you think about it except for Raptor Lake having a little lower idle draw than Arrow Lake. Intel desktop procs are based on laptop designs, while AMD's faster stuff is based on server designs. The situation reverses under load. Arrow Lake is more efficient than Raptor Lake, and AM5 cpus beat both.
Some bits I remember -
Intel ARC GPUs suck at idle power draw if you have more than one screen attached. Also less efficient under load, at least for gaming.
NV vs. AMD GPUs? I dunno.
Raptor Lake beats Arrow Lake and pretty much anything AMD based on CCDs at idle power draw. Arrow Lake is only slightly worse. That actually makes sense if you think about it except for Raptor Lake having a little lower idle draw than Arrow Lake. Intel desktop procs are based on laptop designs, while AMD's faster stuff is based on server designs. The situation reverses under load. Arrow Lake is more efficient than Raptor Lake, and AM5 cpus beat both.
They still do it I think (that one of the most recent review, 9070 gre), they only present it as a ratio without giving the raw number (which will be a bit meaningless by itself to people):
https://www.techpowerup.com/review/sapphire-radeon-rx-9070-gre-pulse/39.html
Why would you think that?
Most people don't see it that way. For desktops, efficiency has only been something people generally care about in small form factor or other situations where they can only have smaller power supplies, or have to content with massive amounts of heat without being able to effectively dissipate it. In such cases, you might place efficiency over performance once the performance reaches the threshold you need for the application. For most gaming desktops, this is a non-issue. You can still air cool or use AIO's for the most powerful CPU's and graphics cards when it comes to gaming. Most modern cases are adequate for air flow, etc. and thus, it's not really a big deal. Your microwave, AC and other home appliances typically impact your power bill far more than your computer ever could.
Not only that, but it's a "pay to play" type scenario. Generally, people want the most performance they can afford. Efficiency doesn't really factor into that as the more powerful cards take more power. For the most part, this doesn't make much of a difference from brand to brand or model to model. What matters most is the GPU type. Sure, there are exceptions where a given AIB might have an overclocked card that pulls more watts due to the way its BIOS is set but those are pretty rare in the grand scheme of things. In short, I think people would like more efficiency all things being equal. But unfortunately, more powerful cards generate more heat and use more power. That's just the way it is.
Past that, there is testing methodology which differs from site to site. Most reviewers don't really have the equipment to test this in isolation. That is, you are only getting a ballpark view of it from a review. Unless you are matching the configuration and settings of the review system exactly, your mileage will vary a lot. The chart posted above me shows more powerful cards as being more efficient, meaning they are factoring performance with power consumption. Naturally, a 3090 is still a decent performer, but its inefficient due to its age and architecture. Meanwhile, a 5080 consumes more power than a 5060, but you get a lot more performance for that increase.
Most people believe that it matters more in a mobile scenario when using laptops. I'd understand why you'd think that, and when operating on battery power alone, that is true. Efficiency does equate to battery life, but gaming (or doing any performance intensive task) on a laptop is going to kill your battery or your performance (or both) to a point where you won't be gaming or doing CPU, memory or GPU intensive tasks for very long like that. You plug your laptop in when you need performance for any length of time. Once you do that, the efficiency is basically a moot point just like it is on the desktop. All laptops use far less power than desktops, meaning that these have even less of an impact on your electric bill.
To put it simply, for desktops, you shop for performance first based on your budget. Keep in mind, outside of factory overclocked models with power limits raised in their firmware, all cards with the same GPU will be roughly the same when it comes to efficiency. That is, a Sapphire 9070 XT and an ASUS 9070 XT will basically pull the same power unless factory overclocking is involved. All 5090's have a 600w TDP, and thus generally have the same or similar efficiency unless you are talking about some wildly overclocked factory model with extra power connectors. The only place it's probably worth considering is when AMD and NVIDIA have GPUs at the same performance level that are around the same price. Basically, outside of the budget end of the spectrum (which I don't look at to be honest), they generally don't. That table above me shows a 9070 as being the most efficient card outside of the 5080, but it gets out performed by a lot of other cards. If it doesn't hit your electric bill that hard, then what does it matter? Few people would buy a 9070 over a 5080 because of some raw efficiency calculation. They either want the performance and features of the 5080, or they are looking at a lower performing card that fits their budget better.
Sure, there are some people that might place efficiency calculations over raw performance, but those people probably aren't gaming at 4K or higher resolutions and trying to max out the settings on the latest AAA games.
AC yes, but I a microwave should consume significantly less than a ~12 hours a week of gaming desktop computer, they do not run that many minutes a day.
Counter intuively, I think it could be the other way around, most in the desktop space never cared about energy efficiancy at all, once it fit on the power supply they have that its, what matter is performance per dollar and noise (noise is were efficiancy can indirectly matter).
The 4k-AAA max setting are the only people to ever reach the 3090ti, 5090, scenario that power efficiancy for power efficiancy sake (because of cooling or just not able to reach noise target even with fancy solution) can become important for the user, the under-volter, the putting a max frame to keep power in check would tend to be them I would imagine, the I have good money, could sale the 4090 for good money and could have had a 5090 for a reasonable cost, but keep the 4090 for power efficacy reason.
I doubt the 9060 or a 5060ti crowd would think about efficiacy difference like that more than those that lived throught a 600 watt GPU system.
Interesting info, thanks.
True. I don't really pay much attention to how much power my gaming rig draws while gaming, but I do notice the heat it produces.
My main rig probably pulls a bit under 1kW while gaming. 5090 + all the other bits. Don't forget the monitor. That's like $0.10 an hour. I'm not interested in gaming slower to save a few pennies an hour. The spot where I'd limit it's draw is in lighter games that don't need a lot of FPS, especially in summer. That's not about saving $, it's about not having to run a fan or the window unit in my office. I have central AC, but a 5090 rig is a space heater and the central can't keep my office cool without freezing the rest of the house. Box fans and window units are noisy. That's really what it's about - noise.
The spot where this gets to be a big deal is machines that are on all the time. Servers of whatever type. People doing datacenter buildouts pay attention to power efficiency.
The box I'll actually pay a little attention to power consumption on is my little home server/NAS box. It runs 24/7/365, and has modest performance requirements. There a 10W difference in idle draw is probably ~$10/year. In other words I'll check the idle power draw on whatever cheap CPUs I'm looking at whenever I upgrade that box if I don't just recycle a board and proc.
Dan_D and LukeTbk,
Your the points about it not being the main interest of the core crowd makes sense, but techpowerup etc do comparison and charts of efficiency, and system monitors always include % of processing power used (not the same but seems related). Thats why it seems lacking or weird that spec is left out on the handy side-by-side comparison sites.
Your the points about it not being the main interest of the core crowd makes sense, but techpowerup etc do comparison and charts of efficiency, and system monitors always include % of processing power used (not the same but seems related). Thats why it seems lacking or weird that spec is left out on the handy side-by-side comparison sites.
- Joined
- Aug 15, 2005
- Messages
- 9,272
This isn't really a replacement for a strict "watts per frame" chart. As this chart is relative to the card being reviewed. Its still useful. But....you have to do more work to extrapolate what this actually means. And then see how that fits into your goals and budget.
I.E. this chart says the 9060 XT is 105% efficient, compared to the 9070 GRE. Which.....isn't really useful. As both the performance gap and power use gap, are larger than "105%" seems to imply. What I mean here is that it can be difficult for an everyday human to truly understand what percentages mean, when expressed as one hundred and X percent.
TPU's Average FPS performance chart, has the 9070 GRE about 31% more performant (about 30 more frames on average), relative to the 9060 XT. And the power use chart shows about 25% more power use (about 61 watts more), for that 31% extra performance.
IMO, a casual glance at a 9060 XT being "105%" more efficient, would make me think two things:
1. its pretty close in peformance (which is not correct)
2. The 9070 GRE doesn't use a lot more power (which is not correct. 60+ watts is a fairly notable amount of extra power, regarding GPUs)
In the end, I think its better to just give us charts with direct comparisons. Too many "relative" comparisons, makes things a little too foggy, unless we are accountants or whatever.
-----------------------------------------------------------------
Last edited:
- Joined
- Dec 24, 2008
- Messages
- 2,295
I dont understand either of these 2 quotes. Isn't the percentage directly based off of frames/watt?
edit: i guess I should get my units right: fps / watt . which becomes frames/joule
yes I think, by wanting a frame per watt graph, they do not mean a graph that only show frames per watt, it would not be more telling for the users at all and would have 100% of the same problems, but those graph we sometime see where you have power-fps and frame per watt ratio all at the same time, a bit like those frame per dollar graph that also do it like that.
say something like this:
But with power, because a graph that say 3 frames per watt per for 9070 versus 2 frame per watt for a 3090TI, versus 150%-100%, that exactly the same for the reader.
- Joined
- Dec 24, 2008
- Messages
- 2,295
yeah i guess. But i dont know why you would assume 100% vs 105% efficiency means 2 gpus have similar FPS or similar power consumption.
- Joined
- Aug 15, 2005
- Messages
- 9,272
I'm not really sure how to re-phrase it. But.....let's re-read my post, and then try to re-phrase it, anyway:
I.E. this chart says the 9060 XT is 105% efficient, compared to the 9070 GRE. Which.....isn't really useful. As both the performance gap and power use gap, are larger than "105%" seems to imply. What I mean here is that it can be difficult for an everyday human to truly understand what percentages mean, when expressed as one hundred and X percent.
TPU's Average FPS performance chart, has the 9070 GRE about 31% more performant (about 30 more frames on average), relative to the 9060 XT. And the power use chart shows about 25% more power use (about 61 watts more), for that 31% extra performance.
The 9070 GRE averages about 30 frames more at 1080p and 1440p. And it uses about 25% more power. (compared to the 9060 XT)
But, when they say the 9060 XT is 105% efficient Vs. the 9070 GRE------its not easy to understand that means it uses 25% less power, but also gets 30 less frames/31% less frames.
Expressing differences as "one hundred and X percent" is not an easy way to understand what it actually means for something which is actually explained by multiple GPU metrics, which are individually represented by very different values.
I.E. this chart says the 9060 XT is 105% efficient, compared to the 9070 GRE. Which.....isn't really useful. As both the performance gap and power use gap, are larger than "105%" seems to imply. What I mean here is that it can be difficult for an everyday human to truly understand what percentages mean, when expressed as one hundred and X percent.
TPU's Average FPS performance chart, has the 9070 GRE about 31% more performant (about 30 more frames on average), relative to the 9060 XT. And the power use chart shows about 25% more power use (about 61 watts more), for that 31% extra performance.
The 9070 GRE averages about 30 frames more at 1080p and 1440p. And it uses about 25% more power. (compared to the 9060 XT)
But, when they say the 9060 XT is 105% efficient Vs. the 9070 GRE------its not easy to understand that means it uses 25% less power, but also gets 30 less frames/31% less frames.
Expressing differences as "one hundred and X percent" is not an easy way to understand what it actually means for something which is actually explained by multiple GPU metrics, which are individually represented by very different values.
Last edited:
- Joined
- Dec 24, 2008
- Messages
- 2,295
i think youre saying fps/watt is just a difficult to understand metric in general? because im not sure how 100 fps/watt vs 105 fps/watt would make it any clearer than a %. you still cant deduce fps or power from it.
Raw result would be meaningless, it all depend on the game setting anyway, 70 fps mean nothing by itself, 3.2 frame per watt even less.
The importance of having actual power-and frame in the graph is rooted in the fact that frame per watt is not an actual relevant metric to anyone pretty much, would it fit in my PSU, will it be noisy will tend to be what people care about, after that it is frame per $, if it fit in their PSU about everyone in the mid tier at same noise level people will take the more performant cards even if it is less power efficiant.
The importance of having actual power-and frame in the graph is rooted in the fact that frame per watt is not an actual relevant metric to anyone pretty much, would it fit in my PSU, will it be noisy will tend to be what people care about, after that it is frame per $, if it fit in their PSU about everyone in the mid tier at same noise level people will take the more performant cards even if it is less power efficiant.