Overclocking, Power, Noise and Conclusions
Though the benchmarks on the previous pages are important to see, the real differentiating features of retail cards from the same GPU class come in the form of power, noise, and overclocking. In these areas, the MSI GTX 970 Gaming is superb.
The GeForce GTX 970 GPU, based on the GM204 GPU, is a fantastic card when it comes to power consumption and power efficiency. With a total system power draw of 292 watts, the MSI GTX 970 Gaming is just 4 watts higher than the reference card (which is well within a margin of error). However, the AMD options, in the form of the Radeon R9 290X and the R9 290, use significantly more power. The AMD flagship R9 290X uses 130 watts more than the MSI GTX 970 Gaming; the R9 290 uses 76 watts more. Those are not small numbers in the grand scheme of things and is something we harped on in our initial GTX 980 and GTX 970 review. If you need a card that is going to run as cool and as quiet as possible, the GTX 970 is a better starting position than either of the AMD options shown here.
Sound testing is important as well when it comes to selecting a graphics card. Each of the cards tested here are retail models. The MSI GTX 970 Gaming is obviously our review target, the "reference" results come from an EVGA GTX 970 ACX 1.0 card, and the R9 290X is an ASUS DirectCU II unit. Even with the high quality cooling solution provided by ASUS on the AMD Radeon R9 290X, the MSI GTX 970 Gaming is a full 2 dBA quieter under full load and nearly 2 dBA quieter at idle. The MSI card was essentially silent during use, even with the machine sitting on our desk during testing.
Okay, now let's dive into overclocking. As always, our results are not 100% indicative of what you will get and they instead should be used a starting point of reference. All overclocking includes variability from chip to chip and GPUs are no different than CPUs in that regard.
Using the latest beta version of the MSI Afterburner overclocking software, I was able to push the MSI GTX 970 Gaming card from the reference base clock speed of 1114 MHz up to an impressive 1364 MHz; that is a 250 MHz offset. That also pushes the Boost clock to 1503 MHz though in our testing and looping of Metro: Last Light, we were able to maintain a clock speed of 1578 MHz pretty consistently over 30 minutes of gaming. At stock settings, that clock speed was 1303 MHz – we were able to maintain a 21% higher clock speed with some very basic and easy overclocking on the MSI GTX 970 Gaming.
And just as important, GPU temperature never went over 70C and the noise levels when overclocked were only modestly increased. (The fan only went up about 120 RPM or so.)
This overclocking was done with the default options and the standard maximum of +87 mV of additional voltage applied to the GPU. Nothing fancy, just moving a few sliders around to get this added performance.
What does that mean for real world performance? The graphs above show that in Metro: Last Light, the overclocked settings result in a 12% faster average frame rate when compared to the stock settings that were already overclocked. This also pushes the performance of the MSI GTX 970 Gaming ahead of a stock Radeon R9 290X.
Closing Thoughts
Now that the GTX 970 cards are showing consistently in stock at online retailers like Amazon and Newegg, I think you'll find more and more PC gamers looking towards this option for upgrading a gaming rig this holiday (and into 2015). As of this writing, there are seven total GTX 970 SKUs in stock – a big increase when we compare it to the first 60 days or so of NVIDIA's GM204 release. It is possible that all the gamers excited about the GTX 970/980 release back in September have moved on or purchased competing products, but I think that is unlikely. With the gluttony of gift cards and checks being passed around in the month of December, I think you'll find upgrades are going to be quite frequent.
With all of that in consideration, is the MSI GTX 970 Gaming the right card for you? It's hard to to see why it wouldn't be – it offers pretty much everything you can ask for. You have a custom cooler that is able to keep the GPU incredibly cool, even when overclocked. The fans are quiet in both load and idle conditions. Our card overclocked very well, getting another 20% boost in clock speed over the out of box settings. And current pricing has the card for just $349, only $20 over the base price. Oh, and you can get a copy of Assassin's Creed Unity, Far Cry 4 or The Crew included as well. (As of this publication the GTX 970 Gaming is out of stock at Amazon.com – but keep an eye out as we are told stock should return very shortly.)
UPDATE: You can find the MSI GTX 970 Gaming at BHPhoto.com in stock as well.
The GTX 970 is definitely a better value than the GTX 980, but AMD has strong competition in the form of the Radeon R9 290X and R9 290. As you saw in our benchmarks on the previous pages, both Hawaii GPUs are offering comparable performance – faster in a couple of instances, slower in a couple others, and essentially tied in a couple more. But pricing definitely favors the AMD cards. You can find R9 290X cards for $349 and R9 290 cards for little as $289 – and considering the performance deltas of these solutions and the GTX 970, AMD has the performance-per-dollar edge. But if that were all that mattered, then AMD would have the market sewn up and that obviously isn't the case.
There are certainly a lot of options out there for gamers looking at the GTX 970 for their next GPU upgrade and many of them are going to be high quality solutions that meet your needs. For users that like the styling, features, and price of the MSI GeForce GTX 970 Gaming, this purchase is an easy decision.
zzzzzzzzzzzzzzzzzzzzzzzzzzzzz
zzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzz
So this is probably a dumb
So this is probably a dumb question, but why is all the testing at 1440p? At this level of GPU, do they not show a lot of difference at 1080p? If so, does this mean that if you are gaming at 1080p still that you can’t go wrong with any of these cards?
My opinion – yes – a single
My opinion – yes – a single 970 or 980 pretty much conquers everything (with a fast enough CPU) at 1080p. The only exception being the Oculus rift DK2 (1920×1080) and the 75 FPS *minimum* you want to maintain. Even there, it’ll do it for 98% of 3D games, but a few like Elite Dangerous will require low settings to perform consistently.
On the flip side, 4K gaming is a weird spot for these cards as some games are already running into VRAM issues @ 4K with 4 GB cards. Star Citizen will even use 3.5 GB at 2560×1600 (my Radeon 7970 3 GB fell flat on it’s face due to VRAM usage at x1600). SC is CryEngine based..
My msi 970 doesn’t conquer
My msi 970 doesn’t conquer anything to say this is absolutely not true can you honestly say no matter what game or bench you throw at a 970 with whatever settings its going to be a solid 60fps well believe me its not… how about euro truck sim 2 @ 80p dropping to 40fps although this is an actually driver problem dont get carried away saying it will destoy anything at 80p it wont..
Euro Truck Sim 2?!@? Cmon
Euro Truck Sim 2?!@? Cmon man…
yeah its called a driver
yeah its called a driver problem, and i have just been playing planet side 2 with the exact same overclock as what ryan achieved and its not on max settings because it will drop below 60fps don’t try and pretend 80p is delt with..
Yeah, sorry, I should have
Yeah, sorry, I should have spelled that out more. At 1080p, this card is MORE than powerful enough for just about any scenario and could even be considered overkill for many users. We tested 2560×1440 (and sometimes at 4K) for hardware that really needs that kind of resolution to show differences.
Even Ubisoft crap ports ?
Even Ubisoft crap ports ?
I hope you must have realized
I hope you must have realized that how much money you can save by subscribing to any such website.
They will be having their wedding and of course
they will have a sweet honeymoon where Edward will say,
“Last night was the best night of life. I also like the scene where Cheryl confesses to Gracie that she was attacked, which prompts Gracie to teach Cheryl self-defense.
Hey Ryan. Big fan
Hey Ryan. Big fan here!
Wasn’t the whole “$400+ GPU is too much for 1080p” put to bed with the introduction of high refresh monitors?
I don’t think any GPU is overkill at any resolution if your intention is too keep the GPU(s) for a while. I’d rather have 100fps, than just enough.
Because 1080p is not an issue
Because 1080p is not an issue anymore… its done, conquered for all future releases too. Midrange cards are fully capable to rape todays games on 1080p. Entry level cards even push solid 60 for all new games on med/high.
1080p is now, what 720p was to you before you read this.
If you think its not conquered, just because some idiot turned up 32xAA in crysis3, or enabled all the broken sun ray effects in stalker clear sky – then screw that guy, he clearly is an imbecile.
Rape? Seriously?
Rape? Seriously?
Seriously? Seriously?
Seriously? Seriously?
Ryan, I have a request: A
Ryan, I have a request: A full review dedicated to comparing overclocked flagship cards. GTX 970 vs GTX 980 vs R9 290x vs R9 290. OC them all as high as possible and then go through all the same testing. Realistically I’d prefer to see this all done with results of both single card and SLI/CF so we can see temps on the top card. Realistically i’d like to see these results with both 2560×1440 & 4k. I realize that with both single and CF that would be a lot of data…but yes a lot of data would be great to see!
I think you also used to do a bang/buck type comparison on your reviews. Basically how much do you get for what you paid for. A review like what i’m requesting seems to be a great place to have bang/buck comparisons.
Thanks.
The biggest problem with
The biggest problem with showing temps is that everyone house is different, an people live in different places, different PC setups, etc. So the only way Ryan can show off true temps of card if he was using a temp controlled room an that cost alot of money.
For me, i just skip all the BS about card temps for this very reason.
Well could get away with just
Well could get away with just a small caption says “Using open air test bench and ambient air temp of xx degree’s”. that would pretty much eliminate that problem.
Are my eyes playing tricks on
Are my eyes playing tricks on me? Or are you slipping with no anti-AMD drivel?
All temp readings needs to
All temp readings needs to have that kind disclaimer on them.
The reason we haven't included temps (which I probably still should) is because temps are regulated by the cooler to not go over a set level. For example, even when overclocked, the GTX 970 Gaming tested here runs at 80C. To compensate, the fan just spins as fast as necessary to keep it cool.
All temp readings needs to
All temp readings needs to have that kind disclaimer on them.
The reason we haven't included temps (which I probably still should) is because temps are regulated by the cooler to not go over a set level. For example, even when overclocked, the GTX 970 Gaming tested here runs at 80C. To compensate, the fan just spins as fast as necessary to keep it cool.
i still think a 1080P
i still think a 1080P comparison should be done anyways. not everyone out there doesn’t have 1140p screens. i know it takes more time but for us the consumer would like to know what the majority of people are using for gaming and thats 1080P really.
I’m confused by the sound
I’m confused by the sound levels of the MSI GeForce GTX 970 4GB Gaming. You say it is 30.9 dbA at idle, but it is supposed to be semi passive cooling as long as the card is below 60c and the fans don’t run. How are you recording 30.9 dbA if the card is idling? Is your case or room especially warm?
I’d like to know because I want to buy this card for a HTPC and have it silent when not gaming and only using XBMC.
Thanks for writing up this
Thanks for writing up this review 🙂 I really think 1080p results should be included for us mainstream gamer/users. EVGA is kinda pricey and the cooler is just ok if you ask me. I got the MSI gtx 970 gamer oc version and love it! I’ll have to say the GTX 970 is nice now when it comes to maximum details/performance for 1080p.
Hi, I’ve just installed a msi
Hi, I’ve just installed a msi gtx970 to replace my gtx670sli. here are the benchmark comparatives made in 3dmark. Interesting to see on many points actually.
but there’s a mystery that remains to be solved. If you take a look to the cloud gate test result for the gtx970, you’ll notice a massive drop in the physics score which is not consistent with all the other results – by so far its clear there is something going on there.
http://www.3dmark.com/3dm/5572125
I’ve tried so far:
uninstalling then cleaning with clean sweep and reinstalling graphic drivers
reinstalling 3dmark
playing with NVidia 3dsettings in config panel
on 3dmark FAQ:
seems that ‘cloud gate’ is testing directx10 functions -through directx11
on the web: I cant find anybody who reported this so far…
here are the two pages for
here are the two pages for comparatives:
gtx970: http://www.3dmark.com/3dm/5572125
gtx 670sli: http://www.3dmark.com/3dm/2635825
I have switched RAM for
I have switched RAM for TridentX ddr3-2400 16gig and it didn’t change a thing. Same results.
I am trying to find out how
I am trying to find out how much power i am pulling with 2 gtx 970 at a 5760×1080 trying to keep a min of 75 fps and are they able to do it