NVIDIA Cheating with 3090 and 40 series. Also Intel with 12900ks

[TECHNOMANCER303]

Has anyone noticed the power draw of recent chips? And lack of efficiency? Seems backwards and I'm rooting for AMD. 

 

AMD consistently uses less wattage and performs at or slightly below both Nvidia and Intel in performance. 

 

 

 

Ps: Have been using team Blue and Green for years now.

[tikker]

Performance comes at a cost and it's not linear. I don't see how that would be cheating. A new generation doesn't automatically imply it uses less power and since the 4000 series doesn't exist yet we can only speculate if it's a reasonable increase.

 

If AMD can deliver the same performace at less power consumption, then more power to AMD. That's the great thing about competition: if one does good, it'll incentivise the other side to do better.

[Middcore]

It's not "cheating," it's just a design choice. 

 

I do find it sort of amusing that power efficiency was something people always brought up as one of the reasons to buy an Nvidia card when they had the edge there but nobody brings it up in discussions of what card to get now. 

 

Nvidia made a (probably correct) calculation that nobody really cares about power draw if throwing more juice at the problem results in suitably eye-popping performance numbers to be touted. 

 

The truth is many people secretly or not-so-secretly like the idea of having a power-hungry card that needs a gigantor PSU, as long as they think it's getting them the "best" performance by whatever tiny margin. I've talked to people on subreddits and forums who would insist on buying overkill PSU's purely for e-peen and become actively hostile if you tried to show them power consumption calculators proving it was a waste. 

[RONOTHAN##]

They are efficient, it's just that in order to get the most performance out of them, they need to push the power limits well past where the peak of the efficiency curve is. There's a reason why undervolting Nvidia cards is so popular, you can decrease the power draw by 20%+ by only dropping the performance 2-3%.

 

Intel is the same way, their efficiency is actually basically identical to AMD right now. Look at the 12400, it performs very similar to the similarly positioned 5600X, load power consumption is basically identical, and idle consumption is slightly better than the 5600x. Their top end SKUs just scale with more power consumption, so in order to get the performance crown, they need to raise the power limits in order to do so. 

 

It would be nice to see stuff a little more efficient, especially for some areas, but with power being cheap where I live and I personally preferring large cases with better cooling, it's not that big a deal personally. 

[8tg]

This has gone back and forth for multiple decades. 
How do you increase performance on the same technology when your new technology is taking a while? Add power.

 

The Pentium 4 was technically slower than the Pentium III, because the Athlon XP showed up and shit on everything intel had done.

And still when the Athlon 64 dropped the Pentium D was just two Pentium 4’s stuck together and was a comparative housefire at the time, while that wattage became fairly normal later.

Then came the core2duo and core2quad which owned the entire market while AMD was trying to keep up adding more cores and higher frequency to the Phenom line. 
Intel i-series came out, here comes the AMD FX line, absolute disaster. More cores, more power, suddenly you’ve got the FX 9590 which was known to immediately destroy the VRMs on low end boards and caused multiple actual house fires.

AMD can’t keep up, intel sits on effectively the same tech for 5ish years until Ryzen comes around, suddenly there’s competition, the solution from intel? 8th gen, more cores.

Things stagnate a bit with Ryzen 2, intel goes into 9th and 10th gen with more cores, more power, Ryzen 3 shows it’s face and the 11900k is born, drawing up to 400 watts or more when overclocked just to compete with the 5800x and up.

You’re about to see the same thing happen again from one side or the other. Alder lake is really impressive and another generation of this P+E cores concept might cement intel this next generation, meaning Ryzen 6000/7000 might rely on cores + power again.

 

Nvidia and AMD have been doing the same thing for ages. And with intel showing off a gpu that seems to be on par with an rtx 3060, that’s just the first retail product disclosed, knowing full well they have products for the entire comparative lineup ready to go, there will be rushed launches from both amd and nvidia to compensate for that potential competition.

Intel has the capability to upset that duopoly with massive scale domestic development and then the same thing all 3 companies can do and have done in the past:

-more cores

-more power

 

Because the 4090 is basically that, it’s just more cores and more power but as a gpu instead of a cpu in concept. It’s not like AMD didn’t just do the same thing with the 6950XT, intel can do it just as well and pull a 2nd generation out of thin air drawing 500 watts from the wall and stomping on everyone.

[AnonymousGuy]

29 minutes ago, Middcore said:

The truth is many people secretly or not-so-secretly like the idea of having a power-hungry card that needs a gigantor PSU, as long as they think it's getting them the "best" performance by whatever tiny margin. I've talked to people on subreddits and forums who would insist on buying overkill PSU's purely for e-peen and become actively hostile if you tried to show them power consumption calculators proving it was a waste. 

Why you gotta go write my biography like that?!

 

But seriously in a desktop I couldn't give a shit about power consumption.  The more the better.

[Middcore]

2 minutes ago, AnonymousGuy said:

But seriously in a desktop I couldn't give a shit about power consumption.  The more the better.

 

Well this is a seriously silly way of thinking. There is no benefit to greater power consumption in and of itself, in fact it's a harm (albeit a very very marginal one) to you and everyone else on the planet. 

 

In their own way some PC guys are no different from rednecks in big trucks "rolling coal" to "pwn the libs" when they go by a Prius. 

[Montana One-Six]

1 hour ago, TECHNOMANCER303 said:

Has anyone noticed the power draw of recent chips? And lack of efficiency? Seems backwards and I'm rooting for AMD. 

AMD isn't far off from NVIDIA and in some load scenarios even surpasses NVIDIA in power consumption (talking about top of the line): https://www.techpowerup.com/review/amd-radeon-rx-6950-xt-reference-design/35.html . So I wouldn't really call either of them efficient and depending on the use case one is more efficient than the other.

 

I don't know anything about GPU architecture stuff but I know those don't allow you to break the rules of physics and I know from high school physics that there is no system on earth that is 100% efficient. Since It also seems like the heat output is getting higher with each generation meaning more power that goes in is "wasted". Which makes it impossible at some point to get higher performance without increasing the power to get there. It's a rather crude explanation with basic physics but my point is at some point there is nothing left to do other than increasing power to get more performance.

[Zando_]

2 hours ago, TECHNOMANCER303 said:

AMD consistently uses less wattage and performs at or slightly below both Nvidia and Intel in performance. 

Have.... Have you seen their latest high end cards? Y'know Vega and Vega II were the GPUs that started the trends of high transient spikes that tripped a shitload of PSUs right? I used my Vega FE to heat my room in the winter, it was actually better than my 1080s at shoving heat into the room (and a bit slower to boot). Had the RVII for a bit till AMD fucked the drivers, it wasn't a light sipping card either. From what I've seen the latest cards suck a lot of wattage as well. For even older stuff, friend has a rig with an R9 290X, its a bit faster than my 780s but actually even more power hungry, so AMD cards being power monsters goes back a while.

 

AMD's Ryzen CPUs have been way more efficient than most Intel chips (almost all the past ones and all the current high tier ones, the new i3 1200 can give 6c chips a run for their money though) though, that's for sure.

1 hour ago, Middcore said:

Well this is a seriously silly way of thinking. There is no benefit to greater power consumption in and of itself, in fact it's a harm (albeit a very very marginal one) to you and everyone else on the planet. 

More power usually = more faster if you have the cooling to handle it. Most PC builders want a faster PC, so they don't mind the power draw. I've never shot for more power draw in general use (especially as it's summer rn, my 250-300W of PC - under gaming load - heats my lounge too much already), but I did shove massive power targets through my RVII to try and get the highest clocks possible for benchmarks.

43 minutes ago, Montana One-Six said:

I don't know anything about GPU architecture stuff but I know those don't allow you to break the rules of physics and I know from high school physics that there is no system on earth that is 100% efficient. Since It also seems like the heat output is getting higher with each generation meaning more power that goes in is "wasted". Which makes it impossible at some point to get higher performance without increasing the power to get there. It's a rather crude explanation with basic physics but my point is at some point there is nothing left to do other than increasing power to get more performance.

I think you mean diminishing returns? Have to pump exponentially more power for each step in performance.

[IkeaGnome]

45 minutes ago, Montana One-Six said:

AMD isn't far off from NVIDIA and in some load scenarios even surpasses NVIDIA in power consumption (talking about top of the line): https://www.techpowerup.com/review/amd-radeon-rx-6950-xt-reference-design/35.html . So I wouldn't really call either of them efficient and depending on the use case one is more efficient than the other.

That page actually bring up a very valid point.

2 hours ago, TECHNOMANCER303 said:

And lack of efficiency?

Seems like they are more efficient when locked to 60 fps at the same settings as older cards.

Spoiler

3090 uses less power than a 1650 and 75% of the power of a 1660S. The extra power they can pull is to make more and more frames because no one is happy with enough. They always need more.

[Unedited_Mind]

My RTX 3070 goes from a 220w GPU to a 260w GPU when I allow it to clock it's self to like 2055mhz instead of 1965mhz gaining like 9 FPS or less which for the most part is never noticed in game.

 

If I under-volt it and cap it to 1940mhz I lose around 0.5 FPS from stock and power drops to like 175w.

 

 

Something like this...

 

 

 

 

 

167w vs 220w stock.

[TECHNOMANCER303]

On 6/6/2022 at 3:26 PM, tikker said:

Performance comes at a cost and it's not linear. I don't see how that would be cheating. A new generation doesn't automatically imply it uses less power and since the 4000 series doesn't exist yet we can only speculate if it's a reasonable increase.

 

If AMD can deliver the same performace at less power consumption, then more power to AMD. That's the great thing about competition: if one does good, it'll incentivise the other side to do better.

I guess it goes back to the marketing and the way it is perceived by the end user. I was at MC for the 12th gen launch and people were rushing to get the new chips because of the claimed performance increases. While this was true they conveniently negated the fact that it drew signifyingly more than 11th or 10th gen. You can see LMG video or GN video about that. 

On 6/6/2022 at 3:27 PM, Middcore said:

It's not "cheating," it's just a design choice. 

 

I do find it sort of amusing that power efficiency was something people always brought up as one of the reasons to buy an Nvidia card when they had the edge there but nobody brings it up in discussions of what card to get now. 

 

Nvidia made a (probably correct) calculation that nobody really cares about power draw if throwing more juice at the problem results in suitably eye-popping performance numbers to be touted. 

 

The truth is many people secretly or not-so-secretly like the idea of having a power-hungry card that needs a gigantor PSU, as long as they think it's getting them the "best" performance by whatever tiny margin. I've talked to people on subreddits and forums who would insist on buying overkill PSU's purely for e-peen and become actively hostile if you tried to show them power consumption calculators proving it was a waste. 

As I agree for performance I care about efficiency more. The US grid was not designed for all the new tech. We still use archaic power generation and 120v wiring. 

Also "Giganator" was hilarious. 

On 6/6/2022 at 3:30 PM, RONOTHAN## said:

They are efficient, it's just that in order to get the most performance out of them, they need to push the power limits well past where the peak of the efficiency curve is. There's a reason why undervolting Nvidia cards is so popular, you can decrease the power draw by 20%+ by only dropping the performance 2-3%.

 

Intel is the same way, their efficiency is actually basically identical to AMD right now. Look at the 12400, it performs very similar to the similarly positioned 5600X, load power consumption is basically identical, and idle consumption is slightly better than the 5600x. Their top end SKUs just scale with more power consumption, so in order to get the performance crown, they need to raise the power limits in order to do so. 

 

It would be nice to see stuff a little more efficient, especially for some areas, but with power being cheap where I live and I personally preferring large cases with better cooling, it's not that big a deal personally. 

The best analogy I can think of is a 2004 terminator cobra mustang V8, 410HP, 14MPG. Or 2022 Eco boost V4, 330HP, 32MPG.

I think of it like a cars engine ICE, The industry is going towards 2-4cylider engines with turbos and hybrid designs. When there was a time we would want as many cylinders as possible and slap a supercharger on it.

 

It is the same for GPU/CPU right now we are slapping a supercharger on it and calling it good.

On 6/6/2022 at 3:31 PM, 8tg said:

This has gone back and forth for multiple decades. 
How do you increase performance on the same technology when your new technology is taking a while? Add power.

 

The Pentium 4 was technically slower than the Pentium III, because the Athlon XP showed up and shit on everything intel had done.

And still when the Athlon 64 dropped the Pentium D was just two Pentium 4’s stuck together and was a comparative housefire at the time, while that wattage became fairly normal later.

Then came the core2duo and core2quad which owned the entire market while AMD was trying to keep up adding more cores and higher frequency to the Phenom line. 
Intel i-series came out, here comes the AMD FX line, absolute disaster. More cores, more power, suddenly you’ve got the FX 9590 which was known to immediately destroy the VRMs on low end boards and caused multiple actual house fires.

AMD can’t keep up, intel sits on effectively the same tech for 5ish years until Ryzen comes around, suddenly there’s competition, the solution from intel? 8th gen, more cores.

Things stagnate a bit with Ryzen 2, intel goes into 9th and 10th gen with more cores, more power, Ryzen 3 shows it’s face and the 11900k is born, drawing up to 400 watts or more when overclocked just to compete with the 5800x and up.

You’re about to see the same thing happen again from one side or the other. Alder lake is really impressive and another generation of this P+E cores concept might cement intel this next generation, meaning Ryzen 6000/7000 might rely on cores + power again.

 

Nvidia and AMD have been doing the same thing for ages. And with intel showing off a gpu that seems to be on par with an rtx 3060, that’s just the first retail product disclosed, knowing full well they have products for the entire comparative lineup ready to go, there will be rushed launches from both amd and nvidia to compensate for that potential competition.

Intel has the capability to upset that duopoly with massive scale domestic development and then the same thing all 3 companies can do and have done in the past:

-more cores

-more power

 

Because the 4090 is basically that, it’s just more cores and more power but as a gpu instead of a cpu in concept. It’s not like AMD didn’t just do the same thing with the 6950XT, intel can do it just as well and pull a 2nd generation out of thin air drawing 500 watts from the wall and stomping on everyone.

I agree.

On 6/6/2022 at 3:58 PM, AnonymousGuy said:

Why you gotta go write my biography like that?!

 

But seriously in a desktop I couldn't give a shit about power consumption.  The more the better.

I disagree.

On 6/6/2022 at 4:05 PM, Middcore said:

 

Well this is a seriously silly way of thinking. There is no benefit to greater power consumption in and of itself, in fact it's a harm (albeit a very very marginal one) to you and everyone else on the planet. 

 

In their own way some PC guys are no different from rednecks in big trucks "rolling coal" to "pwn the libs" when they go by a Prius. 

I daily drive a gaming laptop for this reason. I only turn on the PC beast for renders. 

On 6/6/2022 at 4:40 PM, Montana One-Six said:

AMD isn't far off from NVIDIA and in some load scenarios even surpasses NVIDIA in power consumption (talking about top of the line): https://www.techpowerup.com/review/amd-radeon-rx-6950-xt-reference-design/35.html . So I wouldn't really call either of them efficient and depending on the use case one is more efficient than the other.

 

I don't know anything about GPU architecture stuff but I know those don't allow you to break the rules of physics and I know from high school physics that there is no system on earth that is 100% efficient. Since It also seems like the heat output is getting higher with each generation meaning more power that goes in is "wasted". Which makes it impossible at some point to get higher performance without increasing the power to get there. It's a rather crude explanation with basic physics but my point is at some point there is nothing left to do other than increasing power to get more performance.

For AMD its more efficient in CPU rather than GPU.

On 6/6/2022 at 5:24 PM, Zando_ said:

Have.... Have you seen their latest high end cards? Y'know Vega and Vega II were the GPUs that started the trends of high transient spikes that tripped a shitload of PSUs right? I used my Vega FE to heat my room in the winter, it was actually better than my 1080s at shoving heat into the room (and a bit slower to boot). Had the RVII for a bit till AMD fucked the drivers, it wasn't a light sipping card either. From what I've seen the latest cards suck a lot of wattage as well. For even older stuff, friend has a rig with an R9 290X, its a bit faster than my 780s but actually even more power hungry, so AMD cards being power monsters goes back a while.

 

AMD's Ryzen CPUs have been way more efficient than most Intel chips (almost all the past ones and all the current high tier ones, the new i3 1200 can give 6c chips a run for their money though) though, that's for sure.

More power usually = more faster if you have the cooling to handle it. Most PC builders want a faster PC, so they don't mind the power draw. I've never shot for more power draw in general use (especially as it's summer rn, my 250-300W of PC - under gaming load - heats my lounge too much already), but I did shove massive power targets through my RVII to try and get the highest clocks possible for benchmarks.

I think you mean diminishing returns? Have to pump exponentially more power for each step in performance.

I have to move to Antarctica just to make use of the heat coming from my PC.

On 6/6/2022 at 5:25 PM, IkeaGnome said:

That page actually bring up a very valid point.

Seems like they are more efficient when locked to 60 fps at the same settings as older cards.

  Reveal hidden contents

3090 uses less power than a 1650 and 75% of the power of a 1660S. The extra power they can pull is to make more and more frames because no one is happy with enough. They always need more.

Well I do wish for peak performance when everything reaches reality level of gaming. true specs no latency high fps high resolution and Realtime rendering.

On 6/6/2022 at 7:59 PM, Unedited_Mind said:

My RTX 3070 goes from a 220w GPU to a 260w GPU when I allow it to clock it's self to like 2055mhz instead of 1965mhz gaining like 9 FPS or less which for the most part is never noticed in game.

 

If I under-volt it and cap it to 1940mhz I lose around 0.5 FPS from stock and power drops to like 175w.

 

 

Something like this...

 

 

 

 

 

167w vs 220w stock.

I have heard about under volting but I have never tried it. I do know if my ram drops below 1.35v everything stutters.

[xg32]

someone on youtube did a test, the 3090's power draw and performance is linear with the 1080tie so it's actually faster than the previous gens on all metrics.

 

it's more efficient undervolted since it's in the exponential part of its own curve tho*, 350w is fine for me on gpu, wouldn't consider it cheating.

 

for cpu i'm leaning amd+ddr5+16cores.

[dogwitch]

when you have to under volt...  that a issue itself.

the most powerful super computer and many other  that are coming online  are switching  to amd this time...

general that nvidia bread and butter ,(that where consumer gpu come from).

nvidia knew that had heat/power issue right about 1080 era/server gpu.

native rez for games hit a wall with tech around 360 era... hell games current gen are still not using hd as a indrusty standard due to vram cost ,bw and needed size of vram bottom end 24. really 40gb.

hell the new unreal engine. is doing a old trick with a bit of a.i. to render gaming now.

 

[RONOTHAN##]

9 hours ago, TECHNOMANCER303 said:

The best analogy I can think of is a 2004 terminator cobra mustang V8, 410HP, 14MPG. Or 2022 Eco boost V4, 330HP, 32MPG.

I think of it like a cars engine ICE, The industry is going towards 2-4cylider engines with turbos and hybrid designs. When there was a time we would want as many cylinders as possible and slap a supercharger on it.

 

It is the same for GPU/CPU right now we are slapping a supercharger on it and calling it good.

Not exactly for that analogy. Yes, the ol' school V8s with super chargers on them were hilariously inefficient, but that doesn't actually mean that the turbo 4 bangers are actually more efficient. 

 

If you drive an Ecoboost Mustang like you're supposed to drive a mustang (I.E. somewhat aggressively, not babying the throttle) you will actually be getting something more along the lines of 15-20MPG, very similar to the likes of the Mustang GT of the same year if you're driving it the same way. Turbo 4 bangers have a very weird efficiency curve to them, they are very good at RPMs between 1.5k-3k, but above that when the turbo starts to spool the fuel efficiency drops off a cliff (for a variety of reasons, one of which being that in order to prevent the engine from knocking to death they need to run very rich) and becomes about as efficient as a V8 or worse but without the gearing advantage those have, thus getting worse gas mileage. Turbos mean you're either getting power or fuel efficiency, you can't get both. If you're gonna drive a car relatively hard, you'd actually be better off getting a big NA motor if you care about MPG.

 

Plus, thanks to the aforementioned gearing advantages of the V8s, when naturally aspirated they still can get very good gas mileage, the old C5 Corvette for example will get 40MPG highway stock if you aren't dogging on it thanks to the fact that in 6th gear at highway speed the engine will only be turning ~1k RPM thanks to all the torque they've got down low. 

 

If you want a more accurate analogy, it's more like turbo charging a car. The idle efficiency stays roughly the same, though when you hammer it, it's 10-20% faster for ~50% more fuel usage (numbers aren't 100% accurate, but still). It is still a viable strategy, since a 12900K, for example, in games will still be relatively tame power wise (roughly in line with something like a 5900X), it's just that when you're at 100% load they lose any semblance of efficiency and run balls to the wall

[TECHNOMANCER303]

21 hours ago, xg32 said:

someone on youtube did a test, the 3090's power draw is linear with the 1080ti, so it's actually faster than the previous gens on all metrics.

 

it's more efficient undervolted, 350w is fine for me on gpu, wouldn't consider it cheating.

 

for cpu i'm leaning amd+ddr5+16cores.

Link it plz

8 hours ago, RONOTHAN## said:

Not exactly for that analogy. Yes, the ol' school V8s with super chargers on them were hilariously inefficient, but that doesn't actually mean that the turbo 4 bangers are actually more efficient. 

 

If you drive an Ecoboost Mustang like you're supposed to drive a mustang (I.E. somewhat aggressively, not babying the throttle) you will actually be getting something more along the lines of 15-20MPG, very similar to the likes of the Mustang GT of the same year if you're driving it the same way. Turbo 4 bangers have a very weird efficiency curve to them, they are very good at RPMs between 1.5k-3k, but above that when the turbo starts to spool the fuel efficiency drops off a cliff (for a variety of reasons, one of which being that in order to prevent the engine from knocking to death they need to run very rich) and becomes about as efficient as a V8 or worse but without the gearing advantage those have, thus getting worse gas mileage. Turbos mean you're either getting power or fuel efficiency, you can't get both. If you're gonna drive a car relatively hard, you'd actually be better off getting a big NA motor if you care about MPG.

 

Plus, thanks to the aforementioned gearing advantages of the V8s, when naturally aspirated they still can get very good gas mileage, the old C5 Corvette for example will get 40MPG highway stock if you aren't dogging on it thanks to the fact that in 6th gear at highway speed the engine will only be turning ~1k RPM thanks to all the torque they've got down low. 

 

If you want a more accurate analogy, it's more like turbo charging a car. The idle efficiency stays roughly the same, though when you hammer it, it's 10-20% faster for ~50% more fuel usage (numbers aren't 100% accurate, but still). It is still a viable strategy, since a 12900K, for example, in games will still be relatively tame power wise (roughly in line with something like a 5900X), it's just that when you're at 100% load they lose any semblance of efficiency and run balls to the wall

This is hilarious, fellow car guy. Nice.

 

I see your points. 

 

Idle should always be the same. if not better yoy. 

 

 

Id be interested to hook up 2-4 systems all gen over gen and Nvidia VS AMD VS Intel and hook a power meter up the the plug and run the same benchmark test.

 

Maybe an Idea for the LAB @LinusTech 

 

idk how tags work and I couldn't find luke who works in the labs building or Anthony. 

[KaitouX]

https://www.forum-3dcenter.org/vbulletin/showpost.php?p=12852240&postcount=530

https://www.techpowerup.com/review/intel-core-i9-12900k-alder-lake-tested-at-various-power-limits/

You can get some pretty huge power savings with both the 3090 and 12900K by just undervolting or power limiting them, and often the performance penalty is pretty small as long as you don't limit it too much.

 

Nvidia, Intel and AMD all consider that power consumption doesn't really matter when it comes to their flagships, they will gladly increase the power consumption in 30%+ to achieve 5% higher performance in those products. The 5950X doesn't use more power at stock probably because many motherboards wouldn't be able to run it and it would be way too hot for normal air cooling, and they already announced that AM5 will have CPUs with 230W PPT, so AMD is 100% going in the same direction. The 6950XT is basically the same as the 3090, similar performance and similar power consumption. The 3090Ti still is the worst offender though.

Nvidia's Ampere is a bit weird, as even lower end parts are way out of their sweet spot, some GPUs like the 3070/3060Ti are able to achieve similar performance to stock(<5% difference) while using 20% less power. AMD's RDNA2 on the other hand seems to start losing performance earlier in my experience.

[Quackers101]

SLI gone, double the card, watts and price? yes pls.

[RONOTHAN##]

1 hour ago, TECHNOMANCER303 said:

This is hilarious, fellow car guy. Nice.

This is a computer forum, most people who are serious enough to be on here regularly are also pretty into cars, they do share lots of the same draws (customization, crazy performance numbers, etc.). It's rarer to find someone who's not. Computer overclocking IMO is very similar to doing dyno pulls and drag race runs, except a lot more affordable. 

 

1 hour ago, TECHNOMANCER303 said:

Id be interested to hook up 2-4 systems all gen over gen and Nvidia VS AMD VS Intel and hook a power meter up the the plug and run the same benchmark test.

I do have both a 6900XT and a 3080 (long story) that I was planning on drag racing over the week end to see which would stay in my main rig. I can (if you want me to) undervolt them to be running the same general performance (I'll target the same score in something like Time Spy give or take 100 points) and see which card ends up being more efficient. The AMD card will likely be more efficient thanks to it's stupidly aggressive sleep algorithm (the core will actually turn off in between frame renders then turn back on, affectionately referred to by some as "power naps"), but it still might be somewhat interesting to see. 

 

I could theoretically do the same thing with AMD vs. Intel, I do know a guy with 12700K that I could drag race against the similarly performing 5900X, undervolt both and see which would draw less power under load, but at the same time I doubt he'd let me borrow his computer over the weekend. 

[TECHNOMANCER303]

On 6/8/2022 at 11:34 PM, RONOTHAN## said:

This is a computer forum, most people who are serious enough to be on here regularly are also pretty into cars, they do share lots of the same draws (customization, crazy performance numbers, etc.). It's rarer to find someone who's not. Computer overclocking IMO is very similar to doing dyno pulls and drag race runs, except a lot more affordable. 

 

I do have both a 6900XT and a 3080 (long story) that I was planning on drag racing over the week end to see which would stay in my main rig. I can (if you want me to) undervolt them to be running the same general performance (I'll target the same score in something like Time Spy give or take 100 points) and see which card ends up being more efficient. The AMD card will likely be more efficient thanks to it's stupidly aggressive sleep algorithm (the core will actually turn off in between frame renders then turn back on, affectionately referred to by some as "power naps"), but it still might be somewhat interesting to see. 

 

I could theoretically do the same thing with AMD vs. Intel, I do know a guy with 12700K that I could drag race against the similarly performing 5900X, undervolt both and see which would draw less power under load, but at the same time I doubt he'd let me borrow his computer over the weekend. 

That would be a fun experiment. However based on this thread it would be good to have 1 a detailed page of your test ( scientifically) or a lab to test them out.

 

For education on the topic.

 

 

[ewitte]

On 6/6/2022 at 5:58 PM, AnonymousGuy said:

Why you gotta go write my biography like that?!

 

But seriously in a desktop I couldn't give a shit about power consumption.  The more the better.

The problem is even with the 3000 series it is difficult to cool.  I normally pick cards based on cooling performance because I hate fan noise even then I had to undervolt to get to acceptable noise levels.  Granted I didn't get what I usually go after (Strix, etc) due to lack of card availability, but it is not a cheap crappy cooler either.  Funny I didn't think $900 (for the Strix) was acceptable for a 3080 that went right out the window I'm glad I got it early!  I'm a bit worried about 4000 series noise ...

[xAcid9]

On 6/7/2022 at 6:18 AM, TECHNOMANCER303 said:

Has anyone noticed the power draw of recent chips? And lack of efficiency? Seems backwards and I'm rooting for AMD. 

Current gen is more efficient than last gen(majority of them). Techpowerup tested this on their review.

 

 

Just because they're chugging 300-400 watts, doesn't mean they're not efficient. 

[thewelshbrummie]

On 6/8/2022 at 2:14 AM, TECHNOMANCER303 said:

The best analogy I can think of is a 2004 terminator cobra mustang V8, 410HP, 14MPG. Or 2022 Eco boost V4, 330HP, 32MPG.

On 6/8/2022 at 4:05 PM, RONOTHAN## said:

Not exactly for that analogy. Yes, the ol' school V8s with super chargers on them were hilariously inefficient, but that doesn't actually mean that the turbo 4 bangers are actually more efficient.

I'd refer you both to a fairly old Top Gear episode that's relevant. I managed to find my TV recording of the episode (series 11, episode 1 from June 2008), so I can quote the stats. Clarkson took a BMW M3 4 litre V8 around the track for several laps and followed a 1.5L 4 cylinder Prius being driven by the Stig who was driving it as fast as possible (so that Clarkson was effectively driving it at the same speed as the Prius). The BMW did 19.4mpg vs 17.2 for the Prius. So at an equivalent speed, the more powerful car was actually more efficient.

 

That's the point @TECHNOMANCER303 that you've missed by creating this thread. You're looking at headline/max powerdraw and therefore interpreting that as poor efficiency, while ignoring that at a reduced load the parts in question will use a fraction of the maximum power draw that they can use under 100% load. You seem to have forgotten (or intentionally ignored to make a specific point), that components don't use their full power draw at less than 100% utilisation so may be just as efficient, or more efficient, than their predecessors. Just because AMD parts don't go much beyond 170W vs a 241W max for the 12900K at high loads, doesn't necessarily mean that they are more power efficient. As Clarkson put it, it's not what you use but how you use it that matters.

[TECHNOMANCER303]

On 6/12/2022 at 6:10 AM, xAcid9 said:

Current gen is more efficient than last gen(majority of them). Techpowerup tested this on their review.

 

 

Just because they're chugging 300-400 watts, doesn't mean they're not efficient. 

Ill have to check the review

[TECHNOMANCER303]

On 6/12/2022 at 11:52 PM, thewelshbrummie said:

I'd refer you both to a fairly old Top Gear episode that's relevant. I managed to find my TV recording of the episode (series 11, episode 1 from June 2008), so I can quote the stats. Clarkson took a BMW M3 4 litre V8 around the track for several laps and followed a 1.5L 4 cylinder Prius being driven by the Stig who was driving it as fast as possible (so that Clarkson was effectively driving it at the same speed as the Prius). The BMW did 19.4mpg vs 17.2 for the Prius. So at an equivalent speed, the more powerful car was actually more efficient.

 

That's the point @TECHNOMANCER303 that you've missed by creating this thread. You're looking at headline/max powerdraw and therefore interpreting that as poor efficiency, while ignoring that at a reduced load the parts in question will use a fraction of the maximum power draw that they can use under 100% load. You seem to have forgotten (or intentionally ignored to make a specific point), that components don't use their full power draw at less than 100% utilisation so may be just as efficient, or more efficient, than their predecessors. Just because AMD parts don't go much beyond 170W vs a 241W max for the 12900K at high loads, doesn't necessarily mean that they are more power efficient. As Clarkson put it, it's not what you use but how you use it that matters.

Ill go back and watch it. I could also argue that en EV has no efficiency at high speed vs low speed (city/highway) vs ICE that does have efficiency at high speed vs low speed. There are a lot of variables. 

 

And I guess IDC because it is expected to run low at low. I only care about when im using it or pushing it to limits.