NVIDIA GeForce GTX 970 SLI 4K and NV Surround Review @ [H]

FrgMstr

Just Plain Mean
Staff member
Joined
May 18, 1997
Messages
55,532
NVIDIA GeForce GTX 970 SLI 4K and NV Surround Review - We take 2-Way NVIDIA GeForce GTX 970 SLI for a spin and compare it to R9 290X CF, R9 290 CF, GTX 780 SLI at 4K resolution as well as NV Surround on a triple-display setup. If you want to see how all these video cards compare in these different display configurations we've got just the thing. Find out what $700 SLI gets you.
 
Not too bad for a 256-bit bus. I'm sure the higher memory clock is helping it keep close to the 290X Crossfire, but frame consistency definitely looks to suffer from the narrower bus width. Good to know I won't take too big of a hit if I decide to go 4k in the near future with my current setup.
 
Simply impressive performance for the power draw. Thanks for another great review, [H]
 
Fixed, thanks. - Kyle

Wow how quickly things change, Drivers and game patches have improved high res gaming since the first GTX 980 SLi review as in many of the game on Nvidia hardware. GTX 980 setup was getting numbers like the 970s in this review against the R290X crossfire. I guess we can expect even better performance on GTX 980 SLi now than we could have just 2 weeks ago when that review was published.

GTX 970s hold up pretty well but it looks like the 290X setup won across the board here and costs less if you factor in the rebate.

Nice review.
 
Last edited by a moderator:
Interesting read.

Looks like the 290x CF cards are still holding their own quite well against a newer generation of cards.
 
"NVIDIA GeForce GTX 970 SLI is faster than the AMD Radeon R9 290 CrossFire setup and the GeForce GTX 780 SLI setup, by a wide margin. AMD Radeon R9 290X CrossFire, however, is faster than GeForce GTX 970 SLI in this game.

GTX 970 SLI is about 9% faster than R9 290X CrossFire and 16% faster than GTX 780 SLI."

I think you have a typo on page two. It would make sense if you remove the "X" from the second part of this (unless 970 SLI is both faster and slower than 290X CF at the same time...).
 
in the BF4 the AMD are supposed to be running in mantle due to this:
Because FRAPS does not support Mantle we have to use BF4's built in frametime capture tool to capture the frame times. From this, we can calculate the average FPS instead of frame rate over time. This process of recording the data can be found here. Because we cannot use FRAPS we have to represent the data to you in bar graph format using average FPS only.

But in the images the API its shown as DX11, also in the apple to apple page are listed as DX11..
 
I just ordered the 4K G-Sync Acer and 2 970s to run it. Glad to know I'll be able to play the games I like.
 
"NVIDIA GeForce GTX 970 SLI is faster than the AMD Radeon R9 290 CrossFire setup and the GeForce GTX 780 SLI setup, by a wide margin. AMD Radeon R9 290X CrossFire, however, is faster than GeForce GTX 970 SLI in this game.

GTX 970 SLI is about 9% faster than R9 290X CrossFire and 16% faster than GTX 780 SLI."

I think you have a typo on page two. It would make sense if you remove the "X" from the second part of this (unless 970 SLI is both faster and slower than 290X CF at the same time...).

Still screwed up. Whats the hold up [H]? This isn't Friday afternoon eh ;)
 
How about throwing some single card numbers for scaling?
 
Still screwed up. Whats the hold up [H]? This isn't Friday afternoon eh ;)


Thanks, it was seen and corrected yesterday, we are having some database/caching issues. Hopefully the live article will be fixed soon. That is the hold up.

Now can we keep the thread on topic please?
 
Good article and interesting numbers... . I think a lot of people forget raw framerate is only one main aspect of performance, the rest being scaling, drivers, extras like DSR, MfAA, physx, and shadowplay/nvenc. Add in low power draw, low heat output, and quiet operation + incredible oc headroom widely lauded and the 970 sli is an insane setup. he
 
Were there any oddities with the 290 (non-X) cards during testing? They're usually right on the heels of the 290x in terms of performance and power consumption, but performance is significantly worse here and power consumption is 150W lower than the 290X cards. The power consumption is really what makes me curious about something not working correctly, much more so than the performance.
 
Thanks for the review Brent and Kyle.
Do you think GTX 970s overclocked to 1450MHz core and 7.8GHz memory could catch the R9 290Xs?
 
Thanks for the review Brent and Kyle.
Do you think GTX 970s overclocked to 1450MHz core and 7.8GHz memory could catch the R9 290Xs?

I believe they would but a overclocked 290X will probably retake the advantage but not by much. They'd end up trading blows.
 
I should admit with Martelol that the performance difference between R9 290 and R9 290X in CF is too big. An investigation is needed...
 
Many thanks for this article and your hard work.

When I read the title, I at first thought you were going to be trying 3x 4K!
 
I should admit with Martelol that the performance difference between R9 290 and R9 290X in CF is too big. An investigation is needed...

With the less power, it looks like the 290s were clock throttling during the tests. Turn the fans to manual and crank it up. Still out of the box at this point in time this is what you get, a few adjustments though can change a lot.
 
With the less power, it looks like the 290s were clock throttling during the tests. Turn the fans to manual and crank it up. Still out of the box at this point in time this is what you get, a few adjustments though can change a lot.
I considered that as a possibility, but it'd take some pretty hefty, extended throttling to create a 150W gap.
 
I wish they acknowledged VRAM usage at 4K as FPS does not tell the whole story. I think suspect much of the stuttering described in their review can be attributed to VRAM running out.

In newer games, I see alot of stuttering despite FPS steady around 60 FPS (i.e. Far Cry 4, Watch Dogs, etc).

Let's be real 4 GB of VRAM does not cut it at 4k and probably won't cut in a year's time even at 1080p. NVIDIA got stingy with this release of Maxwell which is why I'm not upgrading until I see a card with at least 8 GB of VRAM.
 
Let's be real 4 GB of VRAM does not cut it at 4k and probably won't cut in a year's time even at 1080p. NVIDIA got stingy with this release of Maxwell which is why I'm not upgrading until I see a card with at least 8 GB of VRAM.

Looking at how 8gb version of the 290x is gonna carry like 150-200$ premium over the 4gb model, I would call it more of a cost reason. If they did and had to mark cards up cause that ppl would complain a ton about it.
 
Thanks for finally doing a triple monitor Surround/Eyefinity review. This was great. Hope to see more like this in the future,

I am going to disagree with your TR metrics though. I am able to play at max settings w/TressFX @ 6020x1080p in 3D at over 100 FPS with 2x 970s. Now granted I'm running close to the 1500 MHz OC'd, (not sure why you didn't include OC'd cards in your review as well), but I can't imagine that much discrepancy.

I am at work atm, otherwise I would post some screenshots with Afterburner OSD.
 
How is the R9 290 CF 25 - 40% slower than R9 290X CF at 4k in apples to apples comparison. stock for stock R9 290 is 7 - 10% slower than R9 290X. at the same clocks the perf diff is 5% or less. Since both R9 290 and R9 290X use the same XDMA CF they experience similar perf scaling from 1 to 2 GPUs. so its just unbelievable that R9 290X CF is so much more faster than R9 290 CF.

http://www.hardwarecanucks.com/foru...ws/67488-asus-gtx-970-strix-oc-review-11.html

http://www.hardware.fr/articles/928-20/recapitulatif-performances.html

R9 290 Tri-X (1000 Mhz) is 7% slower than R9 290X Tri-X (1040 Mhz). This agrees with the 5% clock for clock perf gain we see according to most reviews.

http://www.hardwarecanucks.com/foru...nvidia-gtx-970-sli-performance-review-15.html

You can see the R9 290 CF is 6% slower than GTX 970 SLI and 7% slower than R9 290X CF at 4k. At 1440p there is a wider gap in favour of GTX 970.

Here is a R9 290 (947 Mhz) CF against R9 295X2 (1018 Mhz) . The diff varies between 15% (max) and 2 - 3% (min) with the avg less than 10%.

http://www.legitreviews.com/xfx-radeon-r9-290-crossfire-video-card-review-at-4k-ultra-hd_139418/5

It looks like the objective of this review was to convey that the R9 290X CF and GTX 970 SLI is a class apart from R9 290 CF which is truly not the case in the real world. :rolleyes: Sapphire R9 290 Tri-X CF provides exceptional 4k performance at USD 270 per card (250 after rebate) . No other GPU beats the price perf of R9 290 in single and multi GPU configs.
 
Last edited:
Hmm i don't see info on what cards were used - maybe 290X is non reference while 290 are reference cards which throttle due to heat ?
 
Hmm i don't see info on what cards were used - maybe 290X is non reference while 290 are reference cards which throttle due to heat ?
They say on the front page right before the test setup.

"We are also using two reference AMD Radeon R9 290 CrossFire video cards. We are also using two XFX Radeon R9 290X Double Dissipation CrossFire video cards so that the stock frequency of 1GHz will be maintained instead of clock throttling."
 
Hmm right I don't know why I was looking at system configuration table to find it instead in the middle of 4th paragraph of our goals....
 
Were there any oddities with the 290 (non-X) cards during testing? They're usually right on the heels of the 290x in terms of performance and power consumption, but performance is significantly worse here and power consumption is 150W lower than the 290X cards. The power consumption is really what makes me curious about something not working correctly, much more so than the performance.

I should admit with Martelol that the performance difference between R9 290 and R9 290X in CF is too big. An investigation is needed...

The answer is the R9 290 cards are throttling badly in CF as they are reference cards. the ref R9 290 sometimes throttles even in single GPU setup.

They say on the front page right before the test setup.

"We are also using two reference AMD Radeon R9 290 CrossFire video cards. We are also using two XFX Radeon R9 290X Double Dissipation CrossFire video cards so that the stock frequency of 1GHz will be maintained instead of clock throttling."

This review is such a joke. What is hardocp trying to prove here ? hardocp wants us to show severely throttled R9 290 CF cards to prove that R9 290 CF is much slower than GTX 970 SLI. the fact is majority of R9 290 cards available now have custom coolers and have much better cooling/acoustics. In fact ref R9 290 cards are a minority. People buying ref R9 290 do so for watercooling. nobody is buying ref cards especially for CF because they realize the noise, cooling and throttling problems. Even if some do they just care a damn about acoustics and increase fanspeed to avoid throttling while playing with headphones.

This is why I think the conclusion for the review was first decided and then a R9 290 CF setup which throttles and performs poorly was chosen to prove their conclusion. :rolleyes:

If hardocp was fair they would pick the xfx r9 290 and max out the fans to avoid throttling.

www.amazon.com/XFX-Double-947MHz-Graphics-R9290AEDFD/dp/B00HHIPM5Q/

Surprisingly they praised the low noise on the XFX R9 290X DD even at max fan speed.

http://www.hardocp.com/article/2014...tion_edition_crossfire_review/10#.VHCFhWes98E

"The XFX Double Dissipation fans are the quietest fans we’ve ever encountered at high RPM speeds. These fans are definitely quieter than the ASUS DirectCU II fans when the RPMs and fan speed are increased. We experienced the XFX fans running at 93% fan speed, but, we did not hear them at that level. Whereas, 93% fan speed on the ASUS DirectCU II video cards is extremely loud. On the ASUS cards we wouldn’t go over 75% fan speed because the noise is too loud. However, on the XFX Double Dissipation Edition video cards the fans were literally quiet, even at near maximum fan speed. While playing games, we never noticed fan noise at all. There was no ramping up of fan speed heard while starting to game and leaving the system on all day. "

The worst part is they mention they used XFX R9 290X DD to avoid throttling while for R9 290 they used reference cards. So how does this make an objective and fair comparison. In fact why not use ref R9 290X CF with severe throttling to drive an even more ridiculous conclusion that GTX 970 SLI is faster than R9 290X CF at 4k. The fact is this same review with a XFX R9 290 DD in CF without throttling (selling for USD 270) would mean that the R9 290 CF is very close (or on par) to GTX 970 SLI for a much lower price and maybe hardocp did not want to provide such a conclusion.
 
Last edited:
As an Amazon Associate, HardForum may earn from qualifying purchases.
Everything is clear now. Actually it is a little bit unfair from reviewers side as there no single world in conclusion about reasons why R9 290 in CF have such a low performance. Unexperienced readers could think that GTX 970 in SLI are going to be superior over all models of R9 290.
I feel that retesting is needed as this review had lost it's value.
 
Everything is clear now. Actually it is a little bit unfair from reviewers side as there no single world in conclusion about reasons why R9 290 in CF have such a low performance. Unexperienced readers could think that GTX 970 in SLI are going to be superior over all models of R9 290.
I feel that retesting is needed as this review had lost it's value.

They used reference R9 290's, which are what come in 99% of computers sold with that card and is available on most retail shelves except online stores where someone specifically looks for a non-reference model. The R9 290 cooler is well known to be extremely poor, being hot (poor thermal capacity and dissipation), causing throttling of clockspeeds (particularly in crossfire but even in single-card mode on its default fan profile out of the box), and have a noisy/whiny fan that is very audible and annoying.

Combine the high power draw and hot-running Hawaii chip with that and you get a poorly-performing, feature-lacking (no dsr, mfaa, shadowplay, physx, anything?? hello, amd!!!) monstrosity that costs just as much to boot! Small wonder in just part of a quarter nvidia's been killing it in sales: http://www.jonpeddie.com/publications/add-in-board-report/ Mystery solved!

9Y5JD6I.jpg


Market trends toward better products in tech, nothing new ;). I really hope AMD can put out something more competitive inside of this next quarter or the numbers are going to be really bad for us as consumers... competition helps everyone buying :(.

Total AIB shipments increased this quarter to 12.4 million units from last quarter.
AMD’s quarter-to-quarter total desktop AIB unit shipments decreased 9.5%.
Nvidia’s quarter-to-quarter unit shipments increased 24%.
Nvidia continues to hold a dominant market share position at 72%.
Figures for the other suppliers were flat to declining.
 
Last edited:
They used reference R9 290's, which are what come in 99% of computers sold with that card and is available on most retail shelves except online stores where someone specifically looks for a non-reference model. The R9 290 cooler is well known to be extremely poor, being hot (poor thermal capacity and dissipation), causing throttling of clockspeeds (particularly in crossfire but even in single-card mode on its default fan profile out of the box), and have a noisy/whiny fan that is very audible and annoying.

Combine the high power draw and hot-running Hawaii chip with that and you get a poorly-performing, feature-lacking (no dsr, mfaa, shadowplay, physx, anything?? hello, amd!!!) monstrosity that costs just as much to boot! Small wonder in just part of a quarter nvidia's been killing it in sales: http://www.jonpeddie.com/publications/add-in-board-report/ Mystery solved!

9Y5JD6I.jpg


Market trends toward better products in tech, nothing new ;). I really hope AMD can put out something more competitive inside of this next quarter or the numbers are going to be really bad for us as consumers... competition helps everyone buying :(.

All they would of had to do was use 2 Tri-X 290's from sapphire which you can buy or $269 now (which is a great deal) to prevent throttling.

They used XFX 290x's which werent reference. I honestly think [H] should of left out the 290's in this review since they are in a totally different price bracket then the 970/290x.

Just my 0.02c
 
All they would of had to do was use 2 Tri-X 290's from sapphire which you can buy or $269 now (which is a great deal) to prevent throttling.

They used XFX 290x's which werent reference. I honestly think [H] should of left out the 290's in this review since they are in a totally different price bracket then the 970/290x.

Just my 0.02c

They probably would have throttled anyway due to the high heat output, from what I've seen in other reviews, just not as much. I can't say I disagree though, R9 290's are about as expensive as a basic 970 is if you count in the free game (AC U, FC4, or The Crew) or a little cheaper if not, at this point.

EDIT: Oh, I thought they had used reference 290's, they used non-ref ones already? Yeah, the throttling is just from the high heat output in the case then at that point, the TRI-X would just push the heat into the case faster and recirculate it :(.
 
They probably would have throttled anyway due to the high heat output, from what I've seen in other reviews, just not as much. I can't say I disagree though, R9 290's are about as expensive as a basic 970 is if you count in the free game (AC U, FC4, or The Crew) or a little cheaper if not, at this point.

EDIT: Oh, I thought they had used reference 290's, they used non-ref ones already? Yeah, the throttling is just from the high heat output in the case then at that point, the TRI-X would just push the heat into the case faster and recirculate it :(.

Also remember with AMD you get a shitload of free games too.

Right now a 290 TRI-X is $269 with free games.

290x TRI-X is $319 with free games

and a 970 is $329 with free games.

Which is why I stated, the 290 is in another price bracket. Its $60 cheaper then a 970.
 
Also remember with AMD you get a shitload of free games too.

Right now a 290 TRI-X is $269 with free games.

290x TRI-X is $319 with free games

and a 970 is $329 with free games.

Which is why I stated, the 290 is in another price bracket. Its $60 cheaper then a 970.

Oh wow, they did drop their prices a lot then... good to see, they really needed to in order to remain competitive at all :).
 
Oh wow, they did drop their prices a lot then... good to see, they really needed to in order to remain competitive at all :).

exactly. Since the 970/290x are even in performance mostly. the 290's had to drop in price.

and at $269 they really are a great deal right now, which is why we are seeing 960 GTX rumors starting to float around.
 
They probably would have throttled anyway due to the high heat output, from what I've seen in other reviews, just not as much. I can't say I disagree though, R9 290's are about as expensive as a basic 970 is if you count in the free game (AC U, FC4, or The Crew) or a little cheaper if not, at this point.

EDIT: Oh, I thought they had used reference 290's, they used non-ref ones already? Yeah, the throttling is just from the high heat output in the case then at that point, the TRI-X would just push the heat into the case faster and recirculate it :(.

http://hardocp.com/article/2014/11/19/nvidia_geforce_gtx_970_sli_4k_nv_surround_review#.VHK5tnutHhA

"For comparison we are using two reference GeForce GTX 780 SLI video cards. We are also using two reference AMD Radeon R9 290 CrossFire video cards. We are also using two XFX Radeon R9 290X Double Dissipation CrossFire video cards so that the stock frequency of 1GHz will be maintained instead of clock throttling. In this way we are comparing GTX 970 SLI to GTX 780 SLI, R9 290 CrossFire and R9 290X CrossFire. We have to include R9 290X CrossFire because of the online price drops making the R9 290X such as the exact one we are using here in this review price competitive with GTX 970 now at $350. "

http://www.hardocp.com/article/2014...tion_edition_crossfire_review/11#.VHK7b3utHhA

In an earlier review here hardocp says that the XFX R9 290X DD in CF is much faster than reference R9 290X at uber mode and it consistently maintains 1 Ghz.

"XFX Radeon R9 290X Double Dissipation CrossFire decimated reference AMD Radeon R9 290X CrossFire cards in "Uber" mode. In every gaming situation the XFX R9 290X DD CrossFire cards were faster, in some cases a lot faster. The reason for this is simple, the reference AMD R9 290X video cards are throttling below 1GHz most of the time, which we have found to be much more the case in a CrossFire application. In some instances to great degrees, hundreds of MHz. This negatively impacts performance, especially in CrossFire. Uber mode is not a fast enough fan speed to keep the reference cards from throttling. (Interesting though, if R9 290X CrossFire is installed in a case that will focus on GPU airflow, such as the setup that Kyle used in his R9 290X usage article, this can be overcome.)

The custom cooling on the XFX R9 290X Double Dissipation is robust enough to allow the GPU to maintain 1GHz at all times. In every game we played, in every scenario, and for extended lengths of time, 1GHz was the consistent frequency. This means performance is better simply because the XFX custom video cards are able to deliver the intended clock speed. It is sad that it takes a custom card though to provide what AMD intended in the first place. These custom XFX cards fix AMD’s R9 problems."

hardocp used XFX R9 290X DD to avoid throttling while using reference R9 290s which throttle badly in CF. The worst part is newegg or amazon or any other online retailer has more than 10 custom coolers R9 290 / R9 290X SKUs and barely 1 or 2 reference models. With the vast majority of models being custom coolers I do not understand why hardocp picked a reference R9 290 when they know its going to throttle. This is a misleading review which does not reflect the real performance of R9 290 CF at 4k. The truth is R9 290 CF competes with GTX 970 SLI at 4k while selling at lower prices. The tradeoff is heat, power consumption and noise.
 
in the BF4 the AMD are supposed to be running in mantle due to this:

But in the images the API its shown as DX11, also in the apple to apple page are listed as DX11..

I agree completely, what's going on here?

Was mantle used or not?
 
Back
Top