[TechSpot] Gaming at 4K: GTX 970 SLI vs. AMD R9 290 Crossfire

[RussianSensation]

About 1 month after 970/980 Maxwells launched, AMD's AIBs promptly adjusted pricing. TechSpot investigated 4K performance with an after-market 970 SLI setup going head-to-head against an after-market R9 290 CF in both stock and overclocked form to see how these setups stack up after prices have adjusted.

Review:

http://www.techspot.com/review/898-geforce-gtx-970-sli-4k-gaming/page7.html

Conclusion:

Stock performance: 970 SLI is 3% faster than R9 290 CF.
OC performance: 970 SLI is 4% faster than R9 290 CF
Power usage: The GTX 970 SLI uses around 10% less power than R9 290 CF
Reviewer's thoughts: "That said, I still prefer the GTX 970s as they are cooler, quieter and use less power"

As I have been saying for a while, when you use an after-market R9 290 card and compare power usage in CF/SLI, you do not double up the power use of each card and the power difference is not 100W per card when comparing an after-market 970 vs. an after-market 290.

What about value?

3 cheapest 970s on Newegg: Zotac $330, MSI $340, Gigabyte $350.
3 cheapest 970 SLI on Newegg: Zotac $660, MSI $680, Gigabyte $700.
Average SLI price: $680

3 cheapest R9 290s on Newegg: HIS $230, Gigabyte $255, MSI $276
3 cheapest R9 290s CF on Newegg: HIS $460, Gigabyte $510, MSI $552
Average CF price: $507

All in all a gamer would be paying 34% more on average for 3-4% more performance at 4K with 970 SLI vs. 290 CF, and if we compare cheapest cards then Zotac 970 SLI is 43% more expensive than HIS R9 290 CF.

1 month after 970 launched, after-market R9 290 CF is the clear value winner for 4K gaming, and that's ignoring the free games with AMD cards and review missing recent titles such as Ryse: Son of Rome and CIV:BE where 290 CF would have been faster too.

Side-note: for those who insist on running reference cards in SLI, the reviewer used tightly packed 970 SLI Gainward/Phantom cards without issues.

 

[KaRLiToS]

Great Post but we have to give credit to the GTX 970 with it's price performance. Without it, the R9 290 would still be priced higher.

Honestly, I don't think there is a clear winner here but two.

 

[bystander36]

I really would have liked to see a noise comparison. I'd definitely be willing to pay a bit more for quieter cards. I can assume the 970's are quieter, but with aftermarket solutions, that may not be a given.

 

[RussianSensation]

Great Post but we have to give credit to the GTX 970 with it's price performance. Without it, the R9 290 would still be priced higher.

Honestly, I don't think there is a clear winner here but two.

Thanks! We have to give credit to R9 290/290X for forcing NV to price 970 so aggressively. It goes both ways. My point is now a gamer can get R9 290 CF setup + "Free" 512GB Crucial MX100 or step up from a Core i5 4690K to a Core i7 4790K and get a new case/CPU cooler considering there is a $170-200 price difference between R9 290 CF and 970 SLI. The choice isn't as clear now in favour of the 970 SLI as it was 1 month ago. Great that competition is bringing this to the gamer. That's why I always say it's best to buy AMD/NV GPUs when they are both locked head-to-head. :thumbsup:

I also forgot to mention that as far as 4K gaming performance goes, we haven't moved much from 1 year ago. We really need those next gen 390X/GM200 cards. For someone like you KaRLiToS there is still no viable upgrade path.

 

[RussianSensation]

I really would have liked to see a noise comparison. I'd definitely be willing to pay a bit more for quieter cards. I can assume the 970's are quieter, but with aftermarket solutions, that may not be a given.

The Gainward 970 is not the best example from NV's line. Their best are MSI Gaming, Zotac Extreme and Asus Strix.

From the AMD side, the HIS R9 290X's cooler is able to drop temperatures to 71C under Gaming and it's decent in noise levels, but far from the quietest R9 290.

As far as noise levels go, I find sound videos very valuable:

Here are a bunch of sound videos of after-market R9 290/290X cards:
http://www.tomshardware.com/reviews/radeon-r9-290-and-290x,3728-8.html

Here are a bunch of sound videos of after-market GTX970 cards:
http://www.computerbase.de/2014-10/nvidia-geforce-gtx-970-test-roundup-vergleich/4/

You can find quiet R9 290s and quiet 970s, as well as loud cards from both camps.

 

[wand3r3r]

The 970 isn't special, it's basically equal to a 290 at 4k (no perceptible difference), the only good thing about it was the price cut it brought (sure it's a slightly refined card with the lower tdp). It came out at a good price, but is no longer worth buying with the premium it has.

XDMA crossfire is also a bonus if you go dual cards.

Interesting the power consumption difference is that minor. It's not what you hear perpetuated across the forums.

 

[KaRLiToS]

I really would have liked to see a noise comparison. I'd definitely be willing to pay a bit more for quieter cards. I can assume the 970's are quieter, but with aftermarket solutions, that may not be a given.

Get a waterblock and Voilà, no noise. (especially if you are willing to pay more). Check my Temps, those are 4 x R9 290x in a sandwhich.

IDLE

UNDER LOAD

 

[RussianSensation]

The 970 isn't special, it's basically equal to a 290 at 4k (no perceptible difference), the only good thing about it was the price cut it brought (sure it's a slightly refined card with the lower tdp). It came out at a good price, but is no longer worth buying with the premium it has.

Gamers just forget that R9 290 was $399 for 1 full year (excluding price gouging due to mining), and when 970 launched, many R9 290 models were already going for $340-360, while R9 290X was selling for $440-460. AMD had a lot of room to cut prices. We even saw that NV had more or less insane mark-up on the 780Ti as prices fell from $630-700 to $350-370 on those cards. That's why I keep reminding gamers of the price/technology curve. Eventually it catches up to all GPUs. When 780Ti was $650 10 months from launch and R9 290X was $500, it was obvious those cards weren't really "worth that much anymore." It was evident their prices would drop like a rock the minute something new came out because the same thing happened to the 280, 480, 580, 7970 and 780, etc.

Hopefully 390X can compete with a cut-down GM200 so that we can see more price wars in 2015.

 

[bystander36]

Well, while temps are interesting, it was noise I was interested in a comparison of. Those links weren't exactly helpful in that regard. The noise one was just on 970's.

And I am not interested in the cost nor effort of a water cooling.

 

[Lonyo]

I really would have liked to see a noise comparison. I'd definitely be willing to pay a bit more for quieter cards. I can assume the 970's are quieter, but with aftermarket solutions, that may not be a given.

34%, to me at least, isn't a "bit more" at those levels of prices. $130 vs $100 is a bit more. $460 vs $660 is "I could buy a different more acoustically effective case and some earplugs or new speakers, or watercool the 290s so they are super quite" sort of difference. Not "a bit".

 

[bystander36]

34%, to me at least, isn't a "bit more" at those levels of prices. $130 vs $100 is a bit more. $460 vs $660 is "I could buy a different more acoustically effective case and some earplugs or new speakers, or watercool the 290s so they are super quite" sort of difference. Not "a bit".

I didn't necessarily say this was the setup I'd buy, but the point remains, noise levels are an important part of the equation. At enough noise, I would not consider a card. I will not wear earplugs or headphones. If a setup is too loud, it just does not enter my machine and I already have a very quiet case for how cool it is. Silverstone cases are great that way. I've experienced loud setups before, never again.

 

[f1sherman]

Power usage: The GTX 970 SLI uses around 10% less power than R9 290 CF

As I have been saying for a while, when you use an after-market R9 290 card and compare power usage in CF/SLI, you do not double up the power use of each card and the power difference is not 100W per card when comparing an after-market 970 vs. an after-market 290.

Especially when you are using some Gainward's GTX 970 abomination that in every single game burns more watts than 7970GHz. D:
Ever seen 7970GHz with the dinky cooler like this?

No?
What does that tell you about Tahiti vs GM204 power consumption and about pwr results from this Gainward card?

 

[Subyman]

So what about 1440p with high refresh? I don't think many people are guying 970's to push 4k right now.

 

[Erenhardt]

I hate reading threads like this. It makes me want to buy - its good value, isn't it? Yeah, yeah, but the longer you wait the more you get for the same $. It comes hard, but I think I will stick to what I have for now.

 

[Termie]

FYI, HardwareCanucks ran a much more complete article on this weeks ago:

http://www.hardwarecanucks.com/foru...52-nvidia-gtx-970-sli-performance-review.html

Not only do they run 1440p in addition to 4K, they also have something closer to a reference 970. TechSpot's Gainward 970 is clearly using too much power. Within 18W of a GTX 780? Yeah, that's not true Maxwell. Frankly, you have to watch out for some of these aftermarket 970s. They're using way, way too much power - sloppy work by the AIB partners, I'd say. That's what you get for no Nvidia reference design.

Overall, 970SLI is better at 1080p/1440p, the 290 catches up at 4K because that's just not Maxwell's strength.

 

[SimianR]

FYI, HardwareCanucks ran a much more complete article on this weeks ago:

http://www.hardwarecanucks.com/foru...52-nvidia-gtx-970-sli-performance-review.html

Not only do they run 1440p in addition to 4K, they also have something closer to a reference 970. TechSpot's Gainward 970 is clearly using too much power. Within 18W of a GTX 780? Yeah, that's not true Maxwell. Frankly, you have to watch out for some of these aftermarket 970s. They're using way, way too much power - sloppy work by the AIB partners, I'd say. That's what you get for no Nvidia reference design.

Overall, 970SLI is better at 1080p/1440p, the 290 catches up at 4K because that's just not Maxwell's strength.

Except they used reference R9 290/290X in that article... not only are those cards going to perform worse but they're also going to use more power and run hotter. So I'm not sure that was a good comparison either.

 

[toyota]

FYI, HardwareCanucks ran a much more complete article on this weeks ago:

http://www.hardwarecanucks.com/foru...52-nvidia-gtx-970-sli-performance-review.html

Not only do they run 1440p in addition to 4K, they also have something closer to a reference 970. TechSpot's Gainward 970 is clearly using too much power. Within 18W of a GTX 780? Yeah, that's not true Maxwell. Frankly, you have to watch out for some of these aftermarket 970s. They're using way, way too much power - sloppy work by the AIB partners, I'd say. That's what you get for no Nvidia reference design.

Overall, 970SLI is better at 1080p/1440p, the 290 catches up at 4K because that's just not Maxwell's strength.

Maxwell cards are NOT as efficient as some people think. on average a typical 970 uses the same power if not more than the reference 980. and the reference 980 is usually between a 680 and 770 in power consumption. the 980 should have been called a 195 watt card if the 680 is a 195 watt card. Nvidia preyed on the ignorant by calling it a 165 watt card so they could have the marketing department call it a 2x perf per watt improvement for some scenarios.


and here on hard they had the 970 gaming even with its mild factory oc exceed the power usage of the 780.


image

 

[f1sherman]

factory OC-ed 970 that is again OC-ed PEAKS above 780

hardly a surprise. also note peak power usage, not average gaming consupmtion.
might just as well be 1ms reading.

 

[SimianR]

Especially when you are using some Gainward's GTX 970 abomination that in every single game burns more watts than 7970GHz. D:
Ever seen 7970GHz with the dinky cooler like this?

No?
What does that tell you about Tahiti vs GM204 power consumption and about pwr results from this Gainward card?

The Gainward GTX 970 is more aggressively clocked than the reference 970's. There is no free lunch, you can't really have it both ways. You can neuter the card and keep the TDP down but then the performance isn't as competitive. Most of the reviews I'm seeing are showing the reference GTX 970 within 10-15 watts of the 7970, what do you think happens when you overclock it even slightly?

 

[toyota]

factory OC-ed 970 that is again OC-ed PEAKS above 780

hardly a surprise. also note peak power usage, not average gaming consupmtion.
might just as well be 1ms reading.

this average power consumption excuse needs to die. every time I test a game the peak is not that much higher than average and the peak and average are BOTH consistently higher on card using more power. by that I mean if a card peaks 15-20 watts higher than another then its usually also pulling about 15-20 more watts for average.

 

[f1sherman]

no free lunch, true that.
or no lunch at all with Gainward

I liked the look of their Phantom cards (not so much nowadays), but consistently they have some of the worst thermals of all NV partners. And now they are topping pwr charts.

Again if you think that 7970GHz has similar pwr consumption to 970, or better like here,
can you explain the lack of small cards with dinky coolers like the one above.

3dcenter 980/970 average across several review websites:
http://www.3dcenter.org/artikel/lau...ch-analyse-nvidia-geforce-gtx-970-980-seite-3

 

[RussianSensation]

Especially when you are using some Gainward's GTX 970 abomination that in every single game burns more watts than 7970GHz. D:

No?
What does that tell you about Tahiti vs GM204 power consumption and about pwr results from this Gainward card?

Why are you surprised? Plenty of awesome GTX970 cards use more power than an after-market 7970Ghz/R9 280X. What makes it worse is that sites like TPU always used reference 7970Ghz edition to also incorrectly overstate the power usage of R9 280X/7970Ghz cards.

Since there are no reference 970 cards, there are a few ones that use less power such as Zotac mini-version or Asus Strix. The point is with the pricing parity so large now, the power advantage of 970 will take a decade to pay off given the $170-200 difference in cost. Statements like "R9 290s run hot and loud" don't apply to after-market versions.

The data on R9 290s has been skewed for a long time now because too many professionals reviews used hot and loud and inefficient reference models.

Of course at 1080p/1440p, 970 SLI is more competitive but even then, it's no longer the go to setup it was 1 month ago. Without even launching R9 390 series AMD is back on the value map imo which is good for the consumers. Going with dual R9 290s and Core i7 + high end after-market cooler/new case/larger SSD vs. 970 SLI + Core i5 stock cooler + mechanical drive for OS for a similar price is a big deal.

Overall, 970SLI is better at 1080p/1440p, the 290 catches up at 4K because that's just not Maxwell's strength.

That's true. As I mentioned already the review didn't even include the 2 new games where R9 290 is superior - Ryse: Son of Rome and especially Civ:BE. Hardware Canucks didn't test those games either. I am looking forward to 2015 game benching as major sites should add these titles as well as Witcher 3, Project CARS, Dragon Age Inquisition/AC U, etc. and finally drop useless GPU benchmarks like WOW and Diablo 3 (looking at you TPU).

 

[f1sherman]

this average power consumption excuse needs to die. every time I test a game the peak is not that much higher than average and the peak and average are BOTH consistently higher on card using more power. by that I mean if a card peaks 15-20 watts higher than another then its usually also pulling about 15-20 more watts for average.

It's not an excuse.
There are already too many systematic errors to allow for yet another one.

I really don't care, and no should, about "peak" power usage that does not even mention duration of this peak. Useless...

 

[toyota]

It's not an excuse.
There are already too many systematic errors to allow for yet another one.

I really don't care, and no should, about "peak" power usage that does not even mention duration of this peak. Useless...

you missed the point...

no card is shooting way up for a peak. their peaks are right in line with their averages. the card peaking 10 watts higher than another is probably using 10 watts more than the other on average too.

 

[f1sherman]

Still waiting for an explanation on why there are no dinky looking 7970GHz or 290.

Or why they are not used in laptops. And why 970 package weighs 1kg, while 290 weighs 2.5 kg. Small things like that...

I mean the first has better pwr characteristics and according to RS 290 is also far from "hot"

@toyota
Oh I understand clearly what you are saying. There is clear and well defined correlation between Peak and Average.
Except there isn't.