SLI and CrossFire
We got our hands a set of three AMD Fury X cards and three GeForce GTX 980 Ti cards. How does these beasts scale?
Last week I sat down with a set of three AMD Radeon R9 Fury X cards, our sampled review card as well as two retail cards purchased from Newegg, to see how the reports of the pump whine noise from the cards was shaping up. I'm not going to dive into that debate again here in this story as I think we have covered it pretty well thus far in that story as well as on our various podcasts, but rest assured we are continuing to look into the revisions of the Fury X to see if AMD and Cooler Master were actually able to fix the issue.
What we have to cover today is something very different, and likely much more interesting for a wider range of users. When you have three AMD Fury X cards in your hands, you of course have to do some multi-GPU testing with them. With our set I was able to run both 2-Way and 3-Way CrossFire with the new AMD flagship card and compare them directly to the comparable NVIDIA offering, the GeForce GTX 980 Ti.
There isn't much else I need to do to build up this story, is there? If you are curious how well the new AMD Fury X scales in CrossFire with two and even three GPUs, this is where you'll find your answers.
If you are already familiar with our testing setup and how Frame Rating works when comparing GPUs, you can jump start yourself to the Grand Theft Auto V results.
Testing Configuration
The specifications for our testing system haven't changed for this review as we are still basing it on the Sandy Bridge-E system.
Test System Setup | |
CPU | Intel Core i7-3960X Sandy Bridge-E |
Motherboard | ASUS P9X79 Deluxe |
Memory | Corsair Dominator DDR3-1600 16GB |
Hard Drive | OCZ Agility 4 256GB SSD |
Sound Card | On-board |
Graphics Card | AMD Radeon R9 Fury X 4GB NVIDIA GeForce GTX 980 Ti 6GB |
Graphics Drivers | AMD: 15.15 beta NVIDIA: 353.30 |
Power Supply | Corsair AX1200i |
Operating System | Windows 8.1 Pro x64 |
What you should be watching for
- AMD Fury X vs AMD Fury X CrossFire – How well does the new Fiji GPU with the 15.15 beta driver scale? Does it do better than NVIDIA's GTX 980 Ti cards in SLI?
- AMD Fury X vs AMD Fury X 3-Way CrossFire – Adding a third GPU is always kind of a question mark; do you really see enough performance benefit to justify the added cost?
- AMD Fury X CrossFire vs GTX 980 Ti SLI – Probably the most important comparison to be seen here – can the AMD Fury X in CrossFire out-scale and surpass the performance of the GTX 980 Ti?
Frame Rating: Our Testing Process
If you aren't familiar with it, you should probably do a little research into our testing methodology as it is quite different than others you may see online. Rather than using FRAPS to measure frame rates or frame times, we are using an secondary PC to capture the output from the tested graphics card directly and then use post processing on the resulting video to determine frame rates, frame times, frame variance and much more.
This amount of data can be pretty confusing if you attempting to read it without proper background, but I strongly believe that the results we present paint a much more thorough picture of performance than other options. So please, read up on the full discussion about our Frame Rating methods before moving forward!!
While there are literally dozens of file created for each “run” of benchmarks, there are several resulting graphs that FCAT produces, as well as several more that we are generating with additional code of our own.
If you don't need the example graphs and explanations below, you can jump straight to the benchmark results now!!
The PCPER FRAPS File
While the graphs above are produced by the default version of the scripts from NVIDIA, I have modified and added to them in a few ways to produce additional data for our readers. The first file shows a sub-set of the data from the RUN file above, the average frame rate over time as defined by FRAPS, though we are combining all of the GPUs we are comparing into a single graph. This will basically emulate the data we have been showing you for the past several years.
The PCPER Observed FPS File
This graph takes a different subset of data points and plots them similarly to the FRAPS file above, but this time we are look at the “observed” average frame rates, shown previously as the blue bars in the RUN file above. This takes out the dropped and runts frames, giving you the performance metrics that actually matter – how many frames are being shown to the gamer to improve the animation sequences.
As you’ll see in our full results on the coming pages, seeing a big difference between the FRAPS FPS graphic and the Observed FPS will indicate cases where it is likely the gamer is not getting the full benefit of the hardware investment in their PC.
The PLOT File
The primary file that is generated from the extracted data is a plot of calculated frame times including runts. The numbers here represent the amount of time that frames appear on the screen for the user, a “thinner” line across the time span represents frame times that are consistent and thus should produce the smoothest animation to the gamer. A “wider” line or one with a lot of peaks and valleys indicates a lot more variance and is likely caused by a lot of runts being displayed.
The RUN File
While the two graphs above show combined results for a set of cards being compared, the RUN file will show you the results from a single card on that particular result. It is in this graph that you can see interesting data about runts, drops, average frame rate and the actual frame rate of your gaming experience.
For tests that show no runts or drops, the data is pretty clean. This is the standard frame rate per second over a span of time graph that has become the standard for performance evaluation on graphics cards.
A test that does have runts and drops will look much different. The black bar labeled FRAPS indicates the average frame rate over time that traditional testing would show if you counted the drops and runts in the equation – as FRAPS FPS measurement does. Any area in red is a dropped frame – the wider the amount of red you see, the more colored bars from our overlay were missing in the captured video file, indicating the gamer never saw those frames in any form.
The wide yellow area is the representation of runts, the thin bands of color in our captured video, that we have determined do not add to the animation of the image on the screen. The larger the area of yellow the more often those runts are appearing.
Finally, the blue line is the measured FPS over each second after removing the runts and drops. We are going to be calling this metric the “observed frame rate” as it measures the actual speed of the animation that the gamer experiences.
The PERcentile File
Scott introduced the idea of frame time percentiles months ago but now that we have some different data using direct capture as opposed to FRAPS, the results might be even more telling. In this case, FCAT is showing percentiles not by frame time but instead by instantaneous FPS. This will tell you the minimum frame rate that will appear on the screen at any given percent of time during our benchmark run. The 50th percentile should be very close to the average total frame rate of the benchmark but as we creep closer to the 100% we see how the frame rate will be affected.
The closer this line is to being perfectly flat the better as that would mean we are running at a constant frame rate the entire time. A steep decline on the right hand side tells us that frame times are varying more and more frequently and might indicate potential stutter in the animation.
The PCPER Frame Time Variance File
Of all the data we are presenting, this is probably the one that needs the most discussion. In an attempt to create a new metric for gaming and graphics performance, I wanted to try to find a way to define stutter based on the data sets we had collected. As I mentioned earlier, we can define a single stutter as a variance level between t_game and t_display. This variance can be introduced in t_game, t_display, or on both levels. Since we can currently only reliably test the t_display rate, how can we create a definition of stutter that makes sense and that can be applied across multiple games and platforms?
We define a single frame variance as the difference between the current frame time and the previous frame time – how consistent the two frames presented to the gamer. However, as I found in my testing plotting the value of this frame variance is nearly a perfect match to the data presented by the minimum FPS (PER) file created by FCAT. To be more specific, stutter is only perceived when there is a break from the previous animation frame rates.
Our current running theory for a stutter evaluation is this: find the current frame time variance by comparing the current frame time to the running average of the frame times of the previous 20 frames. Then, by sorting these frame times and plotting them in a percentile form we can get an interesting look at potential stutter. Comparing the frame times to a running average rather than just to the previous frame should prevent potential problems from legitimate performance peaks or valleys found when moving from a highly compute intensive scene to a lower one.
While we are still trying to figure out if this is the best way to visualize stutter in a game, we have seen enough evidence in our game play testing and by comparing the above graphic to other data generated through our Frame rating system to be reasonably confident in our assertions. So much in fact that I am going to going this data the PCPER ISU, which beer fans will appreciate the acronym of International Stutter Units.
To compare these results you want to see a line that is as close the 0ms mark as possible indicating very little frame rate variance when compared to a running average of previous frames. There will be some inevitable incline as we reach the 90+ percentile but that is expected with any game play sequence that varies from scene to scene. What we do not want to see is a sharper line up that would indicate higher frame variance (ISU) and could be an indication that the game sees microstuttering and hitching problems.
just in time for new amd
just in time for new amd drivers to come out, and important ones at that.
great article
As a Crossfire 7970 user the
As a Crossfire 7970 user the first page is so informative and frustrating 🙂
Are those results using the most recently updated drivers, or do they reflect past results with old drivers?
Thanks 🙂
That’s just old, example
That's just old, example data.
Hey Ryan, can you feel that
Hey Ryan, can you feel that big frametime variance on 3-way sli in BF4. Is it easily noticable even tho the average frame time is pretty ok?
I’ve mentioned before that
I’ve mentioned before that you should no longer use the old examples. It just confuses people and gives a false impression of current crossfire performance to less informed readers. What is the purpose of showing these outdated graphs?
Because it shouldn’t matter,
Because it shouldn't matter, and that's what the word "EXAMPLE" indicates. But thanks for the advice.
I hear you, but it would be
I hear you, but it would be nice to get a follow up on this and see what the current state of it is.
I know you guys did some follow up last summer, but there have been a few driver releases subsequently and it would be interesting to see the detail.
I think that rather than saying ‘example’ up at the top of the section in text, it would be better to put a date on the graphs such as:
Spring 2014 – Catalyst Drivers X.YY
That way we could easily see when they were done.
Thanks for all the good articles.
I agree about the confusion.
I agree about the confusion. When I realized it was referring to an older article, I skipped those graphs.
wow man. Thanks Ryan. Nice
wow man. Thanks Ryan. Nice review as always
It just goes to show that in
It just goes to show that in 4k where 2 cards are a necessity at present the Fury Xs are performing extremely well in CFX. Kudos to AMD.
Sure we would all love to have 60 FPS solid with one card with high detail settings but when you have scaling this good .. 95 % !! it doesn’t get much better.
Thanks for the review, very informative.
Problem is both Fury X and
Problem is both Fury X and 980ti are overkill for 4K, you need 2 of them for 60fps and 2 980’s or Fury Pro’s will likely get over 60fps anyway.
Where is the Skyrim chart? I
Where is the Skyrim chart? I thought Ryan in the Youtube video said there is testing for Skyrim.
Crap, I left that page
Crap, I left that page out…how did I do that? It honestly didn't show nearly any scaling past two cards at all, and the AMD results were poor thanks to it being a DX9 title.
That’s too bad, because
That’s too bad, because Skyrim is like the best game ever. 🙂
Somebody told me there were some workarounds to get Crossfire/SLI working on Skyrim. But I think even then we have the frame consistency issue with Crossfire + Skyrim. (And by Crossfire/SLI I am talking just about 2 way, let alone 3 way.)
BTW, how many ms of frame variance is it until our eyes begin to see that something’s wrong?
And finally as usual, great work on the GPU reviews. 🙂
Good to know for when I play
Good to know for when I play DX9 only titles 🙂
Dear PCper,
Can you please
Dear PCper,
Can you please only put one card/solution on each frametime graph. My eyes would greatly appreciate that.
Sincerely,
Kennai
Well, good exercise for your
Well, good exercise for your eyes is crucial.
That’s the excuse I’ll use
That’s the excuse I’ll use the next time I play too much video games.
Agreed, there is too much
Agreed, there is too much information on the graph to decipher it.
I’d suggest you use some sort of dynamic graph where each item can be displayed or hidden.
Yeah, the mess they are in
Yeah, the mess they are in now makes it hard to read right at times.
Fury X vs 980 ti idle diff is
Fury X vs 980 ti idle diff is 8.4w
Fury X vs 980 to load diff is 18w
Your numbers.
Yet, an extra 100w is ‘Normal’ in cross fire mode ?
To date all review show the Fury X to be using a meager 20w more then a 980t ti, not 50w
PCPer : Something doesn’t add up.
unless you only use Metro as your metric?
because in this game the extra 980 ti card is barely adding anything, hence why you only get an 18% speedup.
In this game the two 980 ti in SLI are running at not even half load. The less the GPU do, the less power it uses.
In contrast, the Fury X provide a 70% boost, showing that both GPU are loaded at a much higher level.
And the FPS reflect the higher GPU load. The 980 ti in SLI mode is almost 50% slower.
All that to say… your power numbers are almost almost meaningless as presented.
Crysis3 might be a good game to use for the power usage, as both system seem to scale equally well.
I think you answered your own
I think you answered your own question. The Fury X is scaling slightly better in SLI and thus using proportionately more power. Also, with the HBM memory being more efficient, the GPU power used by the AMD is higher than you might think by just looking at total power of the card. So when you go to Crossfire and use more of the GPU that’s why you get the big difference.
(I meant “Crossfire” not SLI)
(I meant “Crossfire” not SLI)
I have a question for you
I have a question for you Ryan, why weren’t the 15.7 drivers used? Did you make this video prior them coming out and just got around to uploading it or what? Just curios because I figured with all the new updates in it I would like to have seen those results especially in a two way configuration. Thanks for all the hard work you put into doing these because I know they are time consuming.
Yeah, all of this testing was
Yeah, all of this testing was done before 15.7 was out, sorry.
But, based on my Fury CF testing, there aren't any changes and I confirmed that with AMD. There should be no significant differences between the 15.7 driver and the 15.15 beta driver I used.
1. TriFury/TriSLI = CPU/DX11
1. TriFury/TriSLI = CPU/DX11 bottleneck perhaps?
2. whats temperature results? how did 3 fury cards do?
Speaking of DX11:
@Ryan: Are
Speaking of DX11:
@Ryan: Are you planning to test this again after Windows 10 comes out and some DX12 and Vulkan games are released or existing games are patched to support?
Thank you
Andrew
Are you sure it is a
Are you sure it is a driver/game problem and not just a CPU bottleneck?
So Quadfire FTW, right?! 😛
So Quadfire FTW, right?! 😛
Amazing presentation, Thank
Amazing presentation, Thank you for the information!
I’d be interested to see the
I’d be interested to see the power consumption with FRTC/Nvidia FRL, possible?
Frame rate limiting will
Frame rate limiting will definitely lower power consumption…but is this something users really want? I don't know many that are willingly running at lower frame rates just to save power.
I’m thinnking that it might
I’m thinnking that it might be something that I’ll use when I get a FreeSync monitor in order to keep the FPS within the functional range, with the added benefit of using less power and keeping everything cooler. Does that not make sense for some reason I’m not aware of?
well a fixed 60 HZ update
well a fixed 60 HZ update makes for better animation when the game engine knows how much to move objects every update.
So for a freesync/gsync setup it would make sens to set the max HZ to something like 60HZ or any other value that can be guaranteed most of the time, to get the best animation.
If you always go for max fps the game engine is always going to be wrong about how much to move objects as the rendering time is not known in advance.
I’m currently looking at an
I’m currently looking at an Asus MG279Q which has a FreeSync range of 35-90Hz, so I figured I’d set FRTC to around 85-90Hz to ensure I’ve always got FreeSync working.
… Thinking about it
… Thinking about it further, I’d like to know if there’s any performance implications such as an impact on frame times / any strange behaviour etc. when hitting the limit while using FRTC.
Hi Ryan.
Thanks for the
Hi Ryan.
Thanks for the thorough review. One thing not mentioned (although I might have missed it) was that the 980ti keeps dropping frames once multiple cards are being used. This seems to be prevalent on the new Kepler cards (specifically from the 970 upwards), but no review sites mention this anymore – to the point where almost nobody is doing framerate analysis on multi-gpu Kepler setups. It would be great to see if this holds for all new Kepler cards and how it compares vs the 7 series cards and AMD.
I think you mean Maxwell. 😉
I think you mean Maxwell. 😉
My bad… You’re absolutely
My bad… You’re absolutely correct. Not enough sleep this side 😉
And i was jelous of triple
And i was jelous of triple and quad graphics users in youtube.Looks like two is good for me. haha.
Interesting numbers, thanks
Interesting numbers, thanks for the review.
Would it be possible to rerun
Would it be possible to rerun some of these with the new 15.7 driver for comparison? It claims to add a lot of updates to crossfire profiles.
Hello, why not use the 15.7
Hello, why not use the 15.7 drivers for AMD?
Thanks for the article Ryan,
Thanks for the article Ryan, nicely done.
Is there any way you can get
Is there any way you can get some benchmarks done inside a gaming case? I really want to see how well the GTX 980 TI SLI would do in a real world gaming setup (inside a case). Sure, throttling is to be expected with those cards that close together.
Our cards were actually
Our cards were actually separated out more than usually thanks to the large size of the Zotac ArcticStorm card so I don't think chassis installation would affect performance.