Cerb's GTX 970 Performance/Watt Discussion

[Headfoot]

This discussion is being separated from the compensation poll thread. Argue about card noise here instead.
-- stahlhart

but AMD does not offer (yet?) anything remotely comparable in terms of perf/W, and thus noise, especially when not fully loaded (which is the usual case)

Wrong, are you kidding???

Go look up noise charts and find that the Vapor-X and Tri-X are one of the quietest cards on the market. Stop equating one metric to another when you can just read a review directly on that metric.

 

[Cerb]

Wrong, are you kidding???

Go look up noise charts and find that the Vapor-X and Tri-X are one of the quietest cards on the market. Stop equating one metric to another when you can just read a review directly on that metric.

Uh-huh. 41 v. 47dB7? That's not even fair to good AMD cards. AT's tests for noise are not worth using.

Much more accurate results:
http://www.techpowerup.com/reviews/MSI/GTX_970_Gaming/26.html
http://www.techpowerup.com/reviews/Sapphire/R9_290X_Tri-X_OC/23.html
http://www.techpowerup.com/reviews/Sapphire/R9_290_Vapor-X/24.html
P.S. Note that in reality, all of these cards are quieter than AT tests them as!

http://www.techpowerup.com/reviews/MSI/GTX_970_Gaming/25.html
http://www.techpowerup.com/reviews/MSI/GTX_970_Gaming/28.html


A plain old Radio Shack SPL meter, rated for a minimum of 35dBA, will not read anything with my PC under full load, from about 1m away (if the card is only ~30dB, that would make sense, it being by far the loudest component in my system). I have to get a foot more closer to get anything. At idle, it will read nothing even put right up to the fans. Movie playback is enough to cause high enough power use that active cooling will be required with a Radeon, not merely a game that puts a decent amount of stress on the GPU.

I'm not wrongly equating metrics. Watts are Watts. Using fewer Watts is an enabler to lowering noise, which some makers of the new Geforces have implemented reasonably well in their stock coolers. If A uses substantially more atts than B, in a similar form factor, A cannot be made as quiet as B can be. An R9 280X might work as a replacement, but it would cost an added HSF, on top of the card, so wouldn't be too cost-effective. The GTX 9x0 cards use a bit less power under heavy loads, and substantially less under more moderate loads, compared to equally-performing Radeons currently on the market (the latter is also true v. Kepler). None of the Radeons are in the same league with the quieter GTX 970 models (Stix, Gaming, G1, and depending on case, maybe even the BB blower).

 

[Headfoot]

I'm not wrongly equating metrics. Watts are Watts. Using fewer Watts is an enabler to lowering noise

No. Lower noise is lower noise.

All of this nonsense about extra cost and watts enabling noise is rationalizing and sophistry. A 290, with better cooler, that's objectively quieter or equally quiet is CHEAPER than the 970 by $100. See XFX DD 290 @ $220 deal. In reality, aftermarket 290s are 5% slower on average, quieter or equally quiet and $100 cheaper than a 970. I don't care what it costs the manufacturer, I'm not buying shares of stock, I'm buying a GPU. Any need the 970 meets based on gaming performance, heat, power consumption is met by an aftermarket 290 for $100 less. The differences in any tangible performance metric are small. The only big difference between the two is how much cheaper the 290 is. And don't try and tell me 30-40 watts at load between the two actually makes a difference. I wasn't born yesterday. If you want to save 30-40 watts, Menards is selling 60 watt equivalent LED bulbs (7 watts actual power consumption) for $5.99 right now, go buy $100 worth of those with the savings from the 290 and save a hell of a lot more than 30-40 watts.

The only reason to buy a 970 over the 290 at this point in time is either you need nVidia only features (e.g. CUDA) or you're "brand loyal" e.g. buying on emotions and not logic

 

[Cerb]

No. Lower noise is lower noise.

Which, as I already showed, the Radeons do not offer (show me one that can perform video playback with 0dB coming from the fans). The heat generated is why they can't offer it.

All of this nonsense about extra cost and watts enabling noise is rationalizing and sophistry. A 290, with better cooler, that's objectively quieter or equally quiet is CHEAPER than the 970 by $100.

+$70 for an MK-26, +$20-30 for fans, assuming that will be sufficient (that's going to be a bit over $100, shipped). But, that also means losing all my expansion slots. I already did that last time, with Fermi. I like not having to, again. It's not rationalizing. My PC is inaudible, except when gaming. I managed to reach that point nearly a decade ago, and I'm not about to regress from it. Asus and MSI offer Geforces which do not require modifications, to make that happen.

Any need the 970 meets based on gaming performance, heat, power consumption is met by an aftermarket 290 for $100 less.

False. Heat and power consumption (same thing) are going to be substantially higher at anything but a single monitor looking at a blank desktop.

And don't try and tell me 30-40 watts at load between the two actually makes a difference.

The difference is 2-3x that, and the Geforces can use little enough power to not need active cooling when doing basically anything in the desktop (IE, not gaming). If you don't care about that, the Radeons are perfectly good cards. But, some of us do, and it has nothing to do with being faux-green.

 

[garagisti]

Which, as I already showed, the Radeons do not offer. The heat generated is why they can't offer it.

+$70 for an MK-26, +$20-30 for fans, assuming that will be sufficient. But, that also means losing all my expansion slots. I already did that last time, with Fermi. I like not having to, again. It's not rationalizing. My PC is inaudible, except when gaming. I managed to reach that point nearly a decade ago, and I'm not about to regress from it. Asus and MSI offer Geforces which do not require modifications, to make that happen.

The difference is 2-3x that, and the Geforces can use little enough power to not need active cooling when doing basically anything in the desktop (IE, not gaming). If you don't care about that, the Radeons are not bad cards. But, some of us do, and it has nothing to do with being faux-green.

Vapor X coolers on 290s are almost as silent as some 970s... in fact better than some. Sorry if that disappoints.

BTW, did you notice that the link you have provided, they're measuring the performance of a 970 with custom solution against a reference 290/ 290x. Fail!

 

[Cerb]

Vapor X coolers on 290s are almost as silent as some 970s... in fact better than some. Sorry if that disappoints.

It doesn't, because they aren't (and not only being, "as silent," is meaningless ). The attempt at spreading falsehoods is the problem I see, not anything to do with the performance of Radeon cards, which is excellent for the money. The very fact that there are idle noise measurements, above ambient, at that, shows exactly what I'm talking about.

http://www.tweaktown.com/reviews/67...0-4gb-strix-oc-video-card-review/index15.html
The dB are not comparable to other reviews, but the HIS measured quieter than the Sapphire, hence the comparison; and one has no idle measurement, because the fans aren't even on. The 300 series might be able to pull that same thing off, time will tell (it's not like AMD hasn't been working on similar power management tech), but the 200 series cannot, at least not without using 4+ slots, or water.

 

[Paul98]

It doesn't, because they aren't (and not only being, "as silent," is meaningless ). The attempt at spreading falsehoods is the problem I see, not anything to do with the performance of Radeon cards, which is excellent for the money. The very fact that there are idle noise measurements, above ambient, at that, shows exactly what I'm talking about.

http://www.tweaktown.com/reviews/67...0-4gb-strix-oc-video-card-review/index15.html
The dB are not comparable to other reviews, but the HIS measured quieter than the Sapphire, hence the comparison; and one has no idle measurement, because the fans aren't even on. The 300 series might be able to pull that same thing off, time will tell (it's not like AMD hasn't been working on similar power management tech), but the 200 series cannot, at least not without using 4+ slots, or water.

Are they really trying to measure from 2cm in front of the fan? If so that is a joke and can't tell you anything about sound output.

 

[Spanners]

It doesn't, because they aren't (and not only being, "as silent," is meaningless ). The attempt at spreading falsehoods is the problem I see, not anything to do with the performance of Radeon cards, which is excellent for the money. The very fact that there are idle noise measurements, above ambient, at that, shows exactly what I'm talking about.

http://www.tweaktown.com/reviews/67...0-4gb-strix-oc-video-card-review/index15.html
The dB are not comparable to other reviews, but the HIS measured quieter than the Sapphire, hence the comparison; and one has no idle measurement, because the fans aren't even on. The 300 series might be able to pull that same thing off, time will tell (it's not like AMD hasn't been working on similar power management tech), but the 200 series cannot, at least not without using 4+ slots, or water.

The HIS in that review is reference, says it right under the temperature graph.

Seems they are.

 

[Cerb]

Are they really trying to measure from 2cm in front of the fan? If so that is a joke and can't tell you anything about sound output.

AFAIK, TechPowerUp is the only site that comes close to SPCR, in trying to do noise measurements well, but it is another data point, and the relative measurements are, if surprisingly, largely in agreement.

 

[Paul98]

AFAIK, TechPowerUp is the only site that comes close to SPCR, in trying to do noise measurements well, but it is another data point, and the relative measurements are, if surprisingly, largely in agreement.

That test is useless given how mic's and sound work. It's a bad data point and should be thrown out.

 

[Cerb]

Seems they are.

Are, what? Goalpost changes? Yes-sirree (this was actually in the spell-checker's dictionary!). I see no card in those graphs comparable to my own MSI Gaming, the newer GE, nor the Asus Strix. Suitability is limited, today, without modification, to those three series of Geforce GTX 900 series cards. All GTX 970 cards exhibit the performance necessary, but only a handful implement the feature, and not all of those that do implement it well for load noise (IE, the EVGA that turns fans off at low temps with new firmware is still noisy with the fans on).

 

[Spanners]

Are, what? Goalpost changes? Yes-sirree (this was actually in the spell-checker's dictionary!). I see no card in those graphs comparable to my own MSI Gaming, the newer GE, nor the Asus Strix. Suitability is limited, today, without modification, to those three series of Geforce GTX 900 series cards. All GTX 970 cards exhibit the performance necessary, but only a handful implement the feature, and not all of those that do implement it well for load noise (IE, the EVGA that turns fans off at low temps with new firmware is still noisy with the fans on).

"Vapor X coolers on 290s are almost as silent as some 970s... in fact better than some"

Was what you responded to. You said he was spreading falsehoods and that "they aren't".

 

[garagisti]

It doesn't, because they aren't (and not only being, "as silent," is meaningless ). The attempt at spreading falsehoods is the problem I see, not anything to do with the performance of Radeon cards, which is excellent for the money. The very fact that there are idle noise measurements, above ambient, at that, shows exactly what I'm talking about.

http://www.tweaktown.com/reviews/67...0-4gb-strix-oc-video-card-review/index15.html
The dB are not comparable to other reviews, but the HIS measured quieter than the Sapphire, hence the comparison; and one has no idle measurement, because the fans aren't even on. The 300 series might be able to pull that same thing off, time will tell (it's not like AMD hasn't been working on similar power management tech), but the 200 series cannot, at least not without using 4+ slots, or water.

If anything, you are indeed cherrypicking and spreading falsehoods. You're using benchmarks where reference cards are being tested against custom cooled ones. Toms is hardly going to be accused of being an AMD friendly site...
tomshardware.com/reviews/sapphire-vapor-x-r9-290x-8gb,3977-5.html

While methodology involved may vary, and seemingly results are not to your liking (given your reaction and usage of word 'falsehoods'), but it is consistent between the cards and the result is there for you to see. I'm not even copy pasting images/ links, as you may accuse me of spreading more 'falsehoods.'

4+ slots, you sir, really are spreading some falsehoods.

 

[RussianSensation]

Are, what? Goalpost changes? Yes-sirree (this was actually in the spell-checker's dictionary!). I see no card in those graphs comparable to my own MSI Gaming, the newer GE, nor the Asus Strix. .

If you all care about are scientifically measured dBA readings, sure. However, can your human ear actually tell the difference when gaming or watching movies? Surely you have either some good headphones or speakers when you either play games or watch movies with surround sound? I cannot hear my videocards when gaming or watching movies.

I don't know about you but when I hear these R9 290 cards at idle and full load, they are whisper quiet:
http://www.youtube.com/watch?v=ny2xtrT4AiU

^ If you put any of the top after-market R9 290X cards inside a dampened case like the Fractal R4/5, you will not be able to hear them with a custom fan curve.

Seems to be the same things are repeated over and over the internet:

1) R9 290 / 290X are hot and loud at load.

980 G1 Gaming = 43 dBA
http://www.techpowerup.com/reviews/Gigabyte/GeForce_GTX_980_G1_Gaming/26.html

vs.

Sapphire Tri-X 290X = 37 dBA
http://www.techpowerup.com/reviews/Sapphire/R9_290X_Tri-X_OC/23.html

2) R9 290/290X are loud at idle.

I guess PC gamers now have no idea how to set up a custom fan curve in MSI AB to drop fan speeds to levels that are inaudible? 25 dBA at idle. That's not good enough? That's not even close to the best cooler for R9 290/290X cards.

The one consistency seen online since HD7970 review is gamers refuse to be open minded about AMD's open air after-market designs as if they don't even exist. Additionally, the fan curves programmed into the BIOS are fully adjustable to your own liking with 3rd party software. Want to run your R9 290/290X at 50-60C at idle at low % fan speed, you have that option.

When MSI Lightning R9 290X runs at 71C at 1150mhz overclock, you realize just how much headroom you have to drop noise levels and let the card operate at 80-85C?

It's as if since Kepler, all logic about how after-market open air heatsinks function and the context of the maximum ASIC operating temps have went out the window courtesy of close-minded reviewers focusing on AMD reference blower designs and NV's constant PR/marketing wrt GPU's perf/watt and power usage. All we now hear is perf/watt this and that and loud noise levels automatically assumed as a result of higher power usage. Unfortunately this mantra has clouded reason because both concepts do not take into account perf/watt of the total system power usage in games or the readily available option buying an after-market open air heatsink design and setting up a custom fan curve to fit your own specific needs.

This reminds me of the Fermi debacle where I ended up on the opposite end of the spectrum, actually defending the cards. Much to my surprise after re-applying after-market TIM on my 470s and setting up a custom fan curve in AB, I never saw those ludicrous 92-94C loads and 50+ dBA noise levels that nearly every reviewer claimed as 100% true.

Don't forget that with 2-3 cards, AMD has ZeroCore power that basically shuts down the secondary and tertiary cards so that they operate at < 5W.

 

[Cerb]

If you all care about are scientifically measured dBA readings, sure. However, can your human ear actually tell the difference when gaming or watching movies? Surely you have either some good headphones or speakers when you either play games or watch movies with surround sound? I cannot hear my videocards when gaming or watching movies.

Headphones, yes. I don't really have a good option for speakers, without disturbing others. Whisper is louder than it has to be, but I'm willing to put up with it under heavy load, for now. Playing a game, I can tune it out, if it's that low, and then also partially reduced from the headphones I use for gaming being closed (AKG K271 mkII or K340, depending on mood, amped w/ Alpha20s, since none I own need gain from a PC source). When not immersed in a game, whisper quiet is still many times too loud, being used to much quieter.

I've been paying a lot of Tropico 4 lately, for example. With 4xMFAA and FXAA, the video card fans stay off the whole time, and case fans remain under 400 RPM. It doesn't even get close to whisper quiet. I'd love for that to also be the case for all of them, but I'll take what I can get.

The turning off of parts of the card to go super low doesn't help when the GPU is needed to be active. AMD may be able to do just as well in the coming months, with the 300 cards. If so, you can bet Asus and MSI will implement it just as they have with the new Maxwells, an they will be good cards. It's not that they lack the power management technology, but that the power actually used is just too high. It's not like even nVidia has been able to offer this for a long time, either; Kepler's used too much power to just use case air flow, without expensive and large aftermarket coolers, when under low-but-not-idle loads.

 

[Headfoot]

The difference is 2-3x that.

Provide link for this. The difference is not 2-3x ((2x30-40w) = 60-80 to 90-120w) in actual gaming loads between an aftermarket 290 and a 970.

If you're talking about passively cooling a GPU then I can see where this all comes from. Watts matter a lot more on a 100% passive GPU. If you're not talking about a 100% passive GPU, then it simply is not true that the best aftermarket 970 is any quieter than the best aftermarket 290. The Lightning cooler and Tri-X cooler are nearly universally praised as being excellent pieces of engineering. These companies have intentionally differentiated on the noise point in order to show how much better their cards are than the reference one, to sell more of them. It's all very logical.

If you've got links showing the Tri-X or MSI Lightning losing to other fans on noise/temp, I'd love to see them. The only cooler I've seen that is superior to all other is the Inno3d fourty-adjectives cooler: It's not available in the US though.

from http://www.kitguru.net/components/g...980-ichill-herculez-x4-air-boss-ultra-review/.

 

[Headfoot]

The one consistency seen online since HD7970 review is gamers refuse to be open minded about AMD's open air after-market designs as if they don't even exist. Additionally, the fan curves programmed into the BIOS are fully adjustable to your own liking with 3rd party software. Want to run your R9 290/290X at 50-60C at idle at low % fan speed, you have that option.

It's almost as if these readily available facts are being intentionally ignored...

 

[PPB]

The 0db hype has hit hard i think. 0db is useless unless you are in a room where sound intensity is less than 0db. A sound to be appreciated by our ear must be at least 10db more intense than the sound measured in said room. A quiet room ranges between 15 and 20 db, so still at 25db you are whisper quiet. I would rather have my fans run at 20-25db and get better temps than get a passive card at 60c and 0db which are totally useless. And im not even accounting the noise produced by the rest of the fans in your pc. Its utter BS

 

[xthetenth]

What about those of us who want to run their computers in an anechoic room and go insane from hearing their own heartbeat?

Personally I've got the Gigabyte cooler on a 970, and I can't hear the fans under load let alone at idle, and my case is not particularly loud. My loudest fans are 140s moving at about 1k RPM. I'd much rather the fans run at quiet to keep temps down longer.

 

[Headfoot]

What about those of us who want to run their computers in an anechoic room and go insane from hearing their own heartbeat?

If I can hear the fan over my own tinnitus I'm gonna RMA it!!!

 

[96Firebird]

Are people here just not reading posts? Cerb is obviously talking about idle sound and that he likes the features where fans are off at idle.

Either way, I'm not sure why he feels the need to defend his purchase to people who don't even own the cards the topic is about. These posters are having a great time telling other people how they should feel about the 970, like their opinion matters...

 

[RussianSensation]

Are people here just not reading posts? Cerb is obviously talking about idle sound and that he likes the features where fans are off at idle.

Either way, I'm not sure why he feels the need to defend his purchase to people who don't even own the cards the topic is about. These posters are having a great time telling other people how they should feel about the 970, like their opinion matters...

Wrong. This is not what the discussion is about. Re-read everything above your post again. Apparently 25dBA at idle is too loud now and whisper quiet cards are no longer good enough as the card must be 100% silent. You realize that in a well dampened case, there could be absolutely no difference 1 meter away from the gamer between a 20 dB and a 28 dB videocard because the human year has finite frequency spectrum/response?

Also, how come we didn't see NV users talk about 0 dBA being a big deal when it was available way before GTX970 even dropped?

NV's marketing seems to be taking full credit for 0dBA but they basically copied Asus Strix cards. Asus should be given 100% credit first. People make it sound as if prior to 970 you couldn't get a $250+ card with 0dBA fan operation.

http://www.asus.com/Graphics_Cards/STRIXR9280OC3GD5/
http://www.asus.com/uk/Graphics_Cards/STRIXGTX780OC6GD5/

I've owned GPUs for 15+ years. Are you telling me with a straight face that I could hear a 27-28dB GPUs while light gaming with headphones/speakers over keyboard and mouse clicks and overall sound level in an apartment/house? Unless one is gaming in a sound insulated bunker.....

I am basically convinced NV users refuse to accept reality at this point about idle and load noise levels of after-market AMD cards:
http://www.youtube.com/watch?v=4ZlZM-UCXQA

I guess you play your games with no sound whatsoever then right because you can hear those GPUs at load during gaming? Maybe you have the world's best human hearing to be able to hear those fans over music, bass, explosions and gun shots in games? I wish I had superhuman hearing.

But you know what's the most remarkable aspect of all? That Radeon R9 290 Vapor-X had the Intelligent Fan Control for many months prior to any GTX970 but I don't remember even 1 NV user mentioning it as the defining characteristic of buying a high-end graphics card.

I realize people will do everything to justify why 970 is an awesome card post-purchase but it's really grasping for straws when trivial factors such as idle noise levels are used against AMD as if no high-end AMD/NV card ever made before 970 was uber quiet at idle/load.

It's better to simply state you are happy with 970's performance based on the games you play and move on. Once 970 owners start justifying their purchases with some trivial factors that can be easily disproved with scientific data, it's only a matter of time before someone calls them on it.

 

[skipsneeky2]

The 0db hype has hit hard i think. 0db is useless unless you are in a room where sound intensity is less than 0db. A sound to be appreciated by our ear must be at least 10db more intense than the sound measured in said room. A quiet room ranges between 15 and 20 db, so still at 25db you are whisper quiet. I would rather have my fans run at 20-25db and get better temps than get a passive card at 60c and 0db which are totally useless. And im not even accounting the noise produced by the rest of the fans in your pc. Its utter BS

I don't have issues with 0 dba idle temps with my Strix 970,i set up a perfect fan curve of 45% or about 1340rpm when i hit 35cel.Seems aggressive but i only hit that temp the moment a game opens.Never any other time.I haven't really touched 66cel yet in-game either.

I do have a Antec GX500 case with 5 120mm fans on low,one blowing directly on my card.I care more for the longevity of my gpu fans more so then noise but having fans spin for no reason i think is a absolute waste.

My Zotac 770 was spinning about 1050rpm on the lowest setting of 33%,why do i need that when i idled as low as 25cel?Wasteful if you ask me.

 

[Silverforce11]

You guys are discussing fan curves, this is up to individual AIBs and their designs.

Idle power use on modern GPUs are very low with some of that coming from the vram. Most heatpipe designs can passively cool 10-20W no problems.

 

[WhoBeDaPlaya]

Maybe I'm deaf (I do have a little VLF hearing loss), but I can just barely hear the cooler on my XFX 290 DD at 100% with the side of the case open and on the floor (when I was messing around trying to unlock it).

I've seen a lot of people say "ZOMG! 3300rpm! It sound liek a black-label Delta lulz" when it is simply not true. At ~60% (stock 290 DD at load), I don't hear anything with the case open.