[H] got over 1500mhz on their MSI 970 Gaming Card on Stock Volts

[Phynaz]

We use a power meter to look at total system wattage at the wall. The numbers are live, and there is no way to capture all the power numbers from one game to calculate an average.

A $25 Kill-O-Watt will tell you power consumed over time.

 

[3DVagabond]

I'm not dismissing your results or saying they're wrong. There are always going to be reviews of hardware that skew from the normal results most other review sites get. It happens all the time. This time, it just happens to be your review's power numbers. Your numbers were probably accurate for the exact samples you tested but, because they are skewed far from the normal results most reviews are showing, are not indicative of most gtx970 samples.

[H] also tests differently. They play games instead of running canned benchmarks.

 

[3DVagabond]

http://beyond3d.com/showpost.php?p=1876415&postcount=2363

Very interesting.

 

[Termie]

Remember, Brent J. was kind enough to come in here and answer your questions, folks. To the extent possible, just pass along helpful critiques. He's put lots of time into bringing you articles that you want to read.

 

[BrentJ]

Understand our power results are total system wattage, not the video card isolated. Again, total system wattage garnered from playing real games, I look at the power being drawn in every game we test for every card.

 

[Fastx]

FWIW

Gigabyte GTX 970 total system power usage overclocked.
https://www.youtube.com/watch?v=1cxCn_7MZZE&list=UU9Tn-atYOt8qZP-oqui7bhw


Gigabyte GTX 970 stock speed system power usage.
http://www.techspot.com/review/885-nvidia-geforce-gtx-970-gtx-980/page7.html

Brenj numbers look close to me on this overclocked GTX 970 G1.

 

[Subyman]

Yup, I can confirm. I got 2x MSI 970's in the mail today. BOTH do 1554mhz on stock voltage. I can brush up against 1600mhz, but, like HardOCP, I need a bios mod to get more power to them. I did +500 on the memory too. Both cards. These are killer in SLI. I can't increase voltage without touching the power limit.

Also, running Firestrike Extreme, I have an idle load of 170W (monitors are plugged in too.) Max load was 573W, GPUs only. 573-170 = 403W / 2 = 201.5W - Idle Wattage (~6W) = 195.5W under load overclocked. Very rough estimates. My 660W Seasonic runs 2x970's easily.

 

[amenx]

Negative side, another 970 OC card benched/reviewed against bone stock competition..... wtf guys? This makes no sense, i'd like to know if this is an nVidia preferred reviewing method or if reviewers are making this call on their own.

I would call it laziness of reviewers to add in more test cards. Just stick to reviews that have ref cards alongside in the charts for comparison, ie, TPU and Guru3d. H typically has a very meager selection of cards in their reviews.

 

[tviceman]

[H] also tests differently. They play games instead of running canned benchmarks.

I am fairly positive that all the links I provided on the previous page aren't using built in benchmarks.

 

[Rvenger]

Remember, Brent J. was kind enough to come in here and answer your questions, folks. To the extent possible, just pass along helpful critiques. He's put lots of time into bringing you articles that you want to read.

^^^This, and for the others, if I see anymore harsh criticism, I will be handing out infractions based on a personal attack.

-Moderator Rvenger

 

[f1sherman]

It looks like Maxwell's efficiency is part hard/soft-ware as suggested by some review sites, since depending on the load, it could be very efficient or the same as Kepler.

Interesting that Anno loads it like crazy, it must be the DX11 compute they use.

Non-reference cards only.

And they are still not Kepler. Ofc if you clock them high enough you will eventually lose all Maxwell perf/W advantage.
I see nothing hard/soft-ware here. Even if certain workloads punish OC-ed cards more. That's still Tom's only.

 

[Attic]

Yup, I can confirm. I got 2x MSI 970's in the mail today. BOTH do 1554mhz on stock voltage. I can brush up against 1600mhz, but, like HardOCP, I need a bios mod to get more power to them. I did +500 on the memory too. Both cards. These are killer in SLI. I can't increase voltage without touching the power limit.

Also, running Firestrike Extreme, I have an idle load of 170W (monitors are plugged in too.) Max load was 573W, GPUs only. 573-170 = 403W / 2 = 201.5W - Idle Wattage (~6W) = 195.5W under load overclocked. Very rough estimates. My 660W Seasonic runs 2x970's easily.

:ninja: That's a serious getup running on 660w PSU. Man, those clocks 1.5ghz. Monitors plugged in are going to muck up power breakdown IMO.

I test peak power using firestrike extreme as well as Valley benchmark (maxed). Firestrike extreme produced roughly the highest I saw, though close up tree section in Valley (near end of run) produced roughly same.


Thanks to BrentJ for stopping in and giving clarity. You guys and Ryan here at Anand the best in the business. Though [H] did once get me into an 8600GTS (BFGTech! Model) that I was very unhappy with once in hand :sneaky:

Personally I like peak power numbers (as long as not furmark driven), also look for average even if an eyeball over 2-3mins of running to see what's generally going on with power. I look to [H] for gameplay evaluation, not in depth power analysis TBH. I generally care much more about gameplay evaluation and how the cards deliver a game at appropriate settings.

 

[f1sherman]

@Silverforce11

You can see reference cards up there, 580, 290, with a similar Anno-pwr increase pattern.

Are they hard/soft-ware too? And yet ref. 970/980 remain perf/W rock solid.

 

[tviceman]

Basically Gigabyte's Maxwell cards have abnormally high power consumption based on several reviews across the web. Only [H] is showing the MSI Gaming 900 series to have abnormally high power consumption, while other MSI gaming reviews show power consumption to only have moderate increases.

As I have said several times in this thread, occasional anomalies are normal and expected. Not all reviews are going to show the exact same results.

 

[BrentJ]

All testing scenarios are also different, I can't tell you how others test power, only how we do it, and that we use games to look at the wattage from.

 

[BrentJ]

I would call it laziness of reviewers to add in more test cards. Just stick to reviews that have ref cards alongside in the charts for comparison, ie, TPU and Guru3d. H typically has a very meager selection of cards in their reviews.

I had less than a week to do this review, again, without rest and sleepless nights. I squeezed in everything I could.

There will be follow-up articles that cover things we did not in the original reviews. There will also be future retail card reviews that will have comparisons by price, as usual. This is not the only 970 review we are ever going to do.

To get these reviews done, is very much the opposite of lazy.

 

[BrentJ]

BTW, about the power numbers, last thing I will say. We have an ASUS 970 STRIX to evaluate, I will make sure we have both the MSI power numbers and ASUS power numbers in that review to compare side-by-side the cards and see which is more power efficient, along with the 780.

 

[Fastx]

Basically Gigabyte's Maxwell cards have abnormally high power consumption based on several reviews across the web. Only [H] is showing the MSI Gaming 900 series to have abnormally high power consumption, while other MSI gaming reviews show power consumption to only have moderate increases.

As I have said several times in this thread, occasional anomalies are normal and expected. Not all reviews are going to show the exact same results.

A few questions for you TV and/or anyone else.

1.Do you have any other links showing a MSI GTX 970 gaming card oc to 1500 in actual gaming showing total system wattage measured at the wall?
I'd be interested to see them. ^

2.What exactly in your opinion would cause the Gigabyte GTX 970 to use that much more wattage (both @PL 110%/+87Mv) than a MSI 970 Gaming card?

3. Do you think Toyota's overclocking throttling possible problem could be due to his card (non gaming 970 card) having 2-2x6 connectors and not a 1x8 and 1x6 power connector?

 

[tviceman]

A few questions for you TV and/or anyone else.

1.Do you have any links showing a MSI GTX 970 gaming card oc to 1500 in actual gaming showing total system wattage measured at the wall?
I'd be interested to see them. ^

2.What exactly in your opinion would cause the Gigabyte GTX 970 to use that much more wattage (both @PL 110%/+87Mv) than a MSI 970 Gaming card?

3. Do think Toyota's overclocking throttling possible problem could be due to his card (non gaming 970 card) having 2-2x6 connectors and not a 1x8 and 1x6 power connector?

1. I did a quick search on the web and did not see any reviews showing power consumption of the MSI gaming gtx 970 when overclocked. So long as the voltage is unchanged, though, power consumption will linearly increase (percentage wise) with respect to clock speeds. So if the MSI gaming 970 defaults to 1300mhz boost, and it is overclocked to 1500mhz boost without adjusting the voltage, the power consumption increase would be 15% vs. out of the box power consumption before overclocking. Overclocking the vram will increase power consumption as well but not my very much.

2. The gigabyte apparently has a higher TDP limit than any other 970/980 currently. This may account for why.

3. Toyota's problem could be a combination of having a faulty card and/or the card's bios limiting how much TDP headroom the card had. He reported that it was actually crashing at stock settings, which makes me believe that either his card was faulty, or MSI needs to fine-tune the TDP allowance of this particular model.

 

[thilanliyan]

So long as the voltage is unchanged, though, power consumption will linearly increase (percentage wise) with respect to clock speeds.

Are you sure that's correct? I thought once you OCed (even without voltage changes), the temps would go up, which increases power consumption further, making it a non-linear increase in the process?

 

[Abwx]

The Gygabyte cards have power management reduced to minimal quantity, thoses cards can sustain full throughput while the regular cards are limited to very short peaks that allow to get max throughput when needed, wich is a rare occurance in games, this explain why THG got 280W peaks while power comsumptions numbers were 150-170W or so in games and 240-280W once they fed thee card with what they call uncompressed datas, that is, enough datas to get the max throughput sustained.

In principle Nvidia s power management should had detected the overload in respect of defined specs and throttle the card accordingly, but looks like the bios wasnt set within Nvidia specifications.

 

[tviceman]

Are you sure that's correct? I thought once you OCed (even without voltage changes), the temps would go up, which increases power consumption further, making it a non-linear increase in the process?

Temperature might affect power consumption to a small degree, but not that much. Nvidia tunes the fans to keep temps from exceeding 80-81 degrees anyways.

Glaze over each one of these r9 290x reviews, and notice how little variance in power consumption there is despite all of them boasting 15-20+ degrees cooler running vs. the reference AMD card.

http://www.techpowerup.com/reviews/MSI/R9_290X_Lightning/
http://www.techpowerup.com/reviews/Powercolor/R9_290X_PCS_Plus/
http://www.techpowerup.com/reviews/MSI/R9_290X_Gaming/
http://www.techpowerup.com/reviews/Sapphire/R9_290X_Tri-X_OC/
http://www.techpowerup.com/reviews/ASUS/R9_290X_Direct_Cu_II_OC/

The MSI lightning has the largest out of box overclock (8%), is 22-23 degrees cooler than the reference r9 290x, and consumes 6% more power than a stock r9 290x. The sapphire is able to reduce power consumption even with a small overclock, but we're not talking anything massive.

 

[Abwx]

Glaze over each one of these r9 290x reviews, and notice how little variance in power consumption there is despite all of them boasting 15-20+ degrees cooler running vs. the reference AMD card.

Power comsumption is one thing, perfs are another one..
Close to same power comsumption but 10-17% better perfs for the coolest cards since they dont throttle.

 

[thilanliyan]

The MSI lightning has the largest out of box overclock (8%), is 22-23 degrees cooler than the reference r9 290x, and consumes 6% more power than a stock r9 290x. The sapphire is able to reduce power consumption even with a small overclock, but we're not talking anything massive.

I don't think you can compare different cards and come to that conclusion. The core binning could be different, stock voltage could be different, components used on the PCB could be different, etc, which can all lead to different power consumption numbers relative to the reference card. I think the only way to tell is by using the SAME card at 2 different temps and check the power consumption isn't it?

According to that, power consumption increases fairly linearly with temp. Couple that with OCing (which will increase consumption IIRC), and you would have a more than just linear increase wouldn't you?

 

[tviceman]

Do you have the full link to that image?