AMD R9 290/x Crossfire (Faster and smoother than SLI?)

[wand3r3r]

Since the "smoother but slower" (viral) campaign of a year ago seems to have "fizzled out", strangely enough, it's interesting take a look at the current situation of Crossfire with the new XDMA in the R9 290/x cards.

It looks like the campaign has paid off quite well for the improvements in the new AMD cards.

Since heat is brought up, with R9 cards you can space the cards apart as far as you want without a bridge. Yeah the reference fans suck, but even with a weak fan crossfire appears to be vastly improved and soundly beats the NV alternatives, while being noticeably cheaper.

Personally I'm sick of reading all the marketing style viral campaigns, so let's reevaluate the situation. These things shouldn't just get swept under the rug when the side that started the viral marketing is losing or no longer ahead in the metric. :whiste:

So much talk about smoothness, frametime, and the actual experience of multi-GPU video cards while gaming has been discussed lately. In the past, AMD has been highly criticized of having a sub-par CrossFire experience, and rightly so. There were major issues with smoothness where games would stutter or feel choppy, even though the framerates looked good. We've been telling our readers for years that CrossFire just didn't feel as good as SLI while gaming.

Those times have changed, at least on the new Radeon R9 290/X series. The new CrossFire technology has improved upon the CrossFire experience in a vastly positive way. Playing games on the Radeon R9 290X CrossFire configuration was a smooth experience. In fact, it was smoother than SLI in some games. It was also smoother on the 4K display at 3840x2160 gaming, and it was noticeably smoother in Eyefinity at 5760x1200.

http://www.hardocp.com/article/2013/11/01/amd_radeon_r9_290x_crossfire_video_card_review

Two of these cards are priced roughly similar to say a GeForce GTX Titan, yet you'll gain a good chunk in performance. Obviously though, Crossfire is not for everybody. Even myself, I'd rather have the fastest single GPU based graphics card over a multi-GPU solution. Then again, over time things like micro-stuttering is slowly becoming a thing of the past. On that note the latest frame-pacing drivers definitely seem to work nicely. And if you get a 290/290X then the new XDMA Crossfire interface eliminates this in Ultra HD as well. That's a win-win.

http://www.guru3d.com/articles_pages/radeon_r9_290_crossfire_review_benchmarks,23.html


Edit:
An AMD representative on a Q&A on Toms said this yesterday.

2) Frame pacing for multi-GPU systems, at any resolution (e.g. Eyefinity/4K), is fully resolved in hardware on the 290 and 290X.

3) Frame pacing for the R9 270, R9 280 and HD 7000 Series systems will require a software solution. Our engineers are working on that right now, and we intend to release a driver this quarter to resolve the issue.

http://www.tomshardware.co.uk/forum/id-1863987/official-amd-radeon-representatives.html#11877136

 

[Nomanor]

mods, lock this [Redacted].



Profanity is not allowed in the technical forums. If you have an issue take it to Moderator discussions instead of calling us out.

-Rvenger

 

[OCGuy]

Subjective terms like "smoother" aren't scientific. It has to be able to be measured and recreated to have any credibility.

Also you are linking to/discussing 290 reviews, of which there is a thread.

 

[Spidre]

Personally I'm sick of reading all the marketing style viral campaigns

uh huh...

 

[x3sphere]

PCPer measurements still show less frame variance on 780/Titan, so Nvidia is still objectively better

http://www.pcper.com/reviews/Graphi...4GB-Review-Trip-Hawaii-399/Crysis-3-CrossFire

In the [H] benches Crossfire was putting out way more frames than SLI on average. Given the same frame rate, they'd probably find SLI superior still.

And putting the results aside, it's worth noting that Ryan from PCPer says he found the SLI setup smoother than 290 CF in some games, particularly in Crysis 3.

 

[parvadomus]

mods, lock this [Redacted].

Are you talking about your account?

 

[wand3r3r]

PCPer measurements still show less frame variance on 780/Titan, so Nvidia is still objectively better

http://www.pcper.com/reviews/Graphi...4GB-Review-Trip-Hawaii-399/Crysis-3-CrossFire

They could be, it's good to evaluate based on facts. I just found it interesting that AMD has took the criticism seriously and actually put effort into improving crossfire.

Sometimes it may take viral marketing type situations to get improvements. I just find it ironic when it fizzles out. In this case it appears to have been very successful (proven by reviews).

 

[wand3r3r]

Subjective terms like "smoother" aren't scientific. It has to be able to be measured and recreated to have any credibility.

Also you are linking to/discussing 290 reviews, of which there is a thread.

I recall a thread you made a little while ago which you didn't want to lose in the 290 thread (as well as posting it across threads etc.). What's the difference now? :whiste:

In the end this isn't about the 290, it's about R9 290 and 290x crossfire. It's an entirely different subject.

On topic, yes it should be backed with data. Bring on the data.

 

[GaiaHunter]

PCPer measurements still show less frame variance on 780/Titan, so Nvidia is still objectively better

http://www.pcper.com/reviews/Graphi...4GB-Review-Trip-Hawaii-399/Crysis-3-CrossFire

In the [H] benches Crossfire was putting out way more frames than SLI on average. Given the same frame rate, they'd probably find SLI superior still.

And putting the results aside, it's worth noting that Ryan from PCPer says he found the SLI setup smoother than 290 CF in some games, particularly in Crysis 3.

According to PCPER even though the SLI has lower frame variance, the CF (290X) setup takes less time to render frames making it a better experience.

http://www.pcper.com/reviews/Graphi...deon-R9-290X-CrossFire-and-4K-Preview-Testing

At 4K we see another improved result for AMD! There is no difference between the FRAPS and Observed FPS graph which tells us that there are no reported runts or dropped frames from the multi-head CrossFire configuration. A look at the Frame Times graph shows a bit more frame time consistency with the GTX 780s in SLI than the R9 290Xs but considering the frame rates are so much higher, AMD clearly gave us the better experience here.

Well that changes things. At 3840x2160 the R9 290X is clearly faster than the single GTX 780 and with multi-GPU that difference scales even further. Frame times are much lower with the Radeon cards in CrossFire, and even though the variance is higher than we would like, the result was a solid overall experience even with Very High image quality settings.

The 290 CF setup does not seem to have the FPS advantage the 290X CF has but it isn't a one sided victory for the 780 SLI.

With our ASUS PQ321Q monitor we again find that the R9 290 CrossFire setup is producing a better overall frame rate than the GTX 780s in SLI though this time with similar levels of frame time variance.

 

[joshhedge]

What are the advantages of AMD no longer using a Crossfire bridge vs Crossfire over PCI-e?

 

[96Firebird]

Good to see they got the issues fixed with 4K/mulit-monitor, where 2+ cards is really almost required.

Edit - Wait, do they have a fix for the other cards yet? Or is it just the 290(X) that have frame-pacing because of XDMA?

 

[wand3r3r]

PCPer measurements still show less frame variance on 780/Titan, so Nvidia is still objectively better

http://www.pcper.com/reviews/Graphi...4GB-Review-Trip-Hawaii-399/Crysis-3-CrossFire

In the [H] benches Crossfire was putting out way more frames than SLI on average. Given the same frame rate, they'd probably find SLI superior still.

And putting the results aside, it's worth noting that Ryan from PCPer says he found the SLI setup smoother than 290 CF in some games, particularly in Crysis 3.

Can you provide the full picture? Rather than "some games" with one example, list them as well as the ones which are not better. "They'd probably find" is entirely subjective. I'm not going to take the time to debate this without hard data.

 

[x3sphere]

According to PCPER even though the SLI has lower frame variance, the CF setup takes less time to render frames making it a better experience.

Not across the board though.

Grid 2:

Interesting - even though the AMD R9 290 cards in CrossFire produce a better average frame rate, the truth is that the NVIDIA GTX 780s in SLI produce a better overall experience thanks to a smoother frame rate and a very close average frame rate.

Crysis 3:

The GTX 780 in SLI clearly makes a smoother gaming experience here than the R9 290s in CrossFire and you can see it with the much wider band of orange in our Frame Times graphic. In addition, the GTX 780s in SLI are able to produce a higher average frame rate across our testing as well at 2560x1440.

 

[wand3r3r]

What are the advantages of AMD no longer using a Crossfire bridge vs Crossfire over PCI-e?

A side benefit is that you can place them in any PCI-E slot so you can have air gaps between them. (Assuming you have the space)

Since AMD advertises it as "no performance loss vs. bridges" and compatible with framepacing, I'm not sure if there are other benefits?

 

[GaiaHunter]

Not across the board though.

Grid 2:



Crysis 3:

I didn't notice you were pointing to the 290 CF instead of the 290X CF.

I already edited my post to reflect that.

 

[x3sphere]

Can you provide the full picture? Rather than "some games" with one example, list them as well as the ones which are not better. "They'd probably find" is entirely subjective. I'm not going to take the time to debate this without hard data.

Well, it would help if [H] was more specific as to what they meant by "some games" as well. I just went through the [H] review, and they only really comment on the smoothness in Far Cry 3. Unfortunately PCPer didn't test that game so we don't have any frame time data.

At any rate, it's a fact that having a higher frame rate makes it easier to gloss over any inconsistencies SLI/CF brings. The quote GaiaHunter posted seems to back up what I said:

At 4K we see another improved result for AMD! There is no difference between the FRAPS and Observed FPS graph which tells us that there are no reported runts or dropped frames from the multi-head CrossFire configuration. A look at the Frame Times graph shows a bit more frame time consistency with the GTX 780s in SLI than the R9 290Xs but considering the frame rates are so much higher, AMD clearly gave us the better experience here.

So, 290/X CF still have worse frame variance but the increase in frame rate is making up for it. At 4K AMD has a huge advantage in frame rate though, which isn't the case at other resolutions like 1600p.

 

[GaiaHunter]

My main problem with these new methods is that where is the point the numbers don't matter in the physical world.

I mean we all understand FPS.

If I have a 60 Hz refresh rate monitor, having a minimum of 60 FPS or having 300 FPS is irrelevant.

At what point does frame variance stop to be of relevance?

 

[Saylick]

What are the advantages of AMD no longer using a Crossfire bridge vs Crossfire over PCI-e?

The benefit lies mostly in that there is far more bandwidth available for frame buffer transfers using the PCIe bus than in the old CF bridge, which means CFX can now scale properly with ultra-high resolutions (read: more frames than 1600p @ 60 Hz). In other words, so long as the PCIe standard is updated frequently enough to always provide enough bandwidth, the new PCIe CFX implementation is more-or-less futureproof in comparison to the old CF bridge implementation, and is thus the reason why AMD chose to go down this route.

Of course, the other alternative is to develop a NEW CF bridge standard, but going down this route doesn't solve the problem of bandwidth limitations in the long run. Eventually you would run into the same issue as before. If this is the case, you might as let the continually developing PCIe standard do all the work for you.

 

[Leadbox]

So, 290/X CF still have worse frame variance but the increase in frame rate is making up for it. At 4K AMD has a huge advantage in frame rate though, which isn't the case at other resolutions like 1600p.

The frametime variance for both solutions is comfortably in the imperceptible range,I feel Pcper is speaking more to the data than than what they see play out on the screen.[H] on the other hand is a purely subjective opinion, and lets not forgot that it was subjective opinion that led to the scientific method being developed

 

[SolMiester]

Since the "smoother but slower" (viral) campaign of a year ago seems to have "fizzled out", strangely enough, it's interesting take a look at the current situation of Crossfire with the new XDMA in the R9 290/x cards.

It looks like the campaign has paid off quite well for the improvements in the new AMD cards.

Since heat is brought up, with R9 cards you can space the cards apart as far as you want without a bridge. Yeah the reference fans suck, but even with a weak fan crossfire appears to be vastly improved and soundly beats the NV alternatives, while being noticeably cheaper.

Personally I'm sick of reading all the marketing style viral campaigns, so let's reevaluate the situation. These things shouldn't just get swept under the rug when the side that started the viral marketing is losing or no longer ahead in the metric. :whiste:




http://www.hardocp.com/article/2013/11/01/amd_radeon_r9_290x_crossfire_video_card_review



http://www.guru3d.com/articles_pages/radeon_r9_290_crossfire_review_benchmarks,23.html

Sorry your sick of the viral campaigns?....however they do appear to have kicked AMDs' butt into doing something about it....Not sure what the sweeping under the rug is supposed to imply. NV is useless now XF has caught up?

 

[wand3r3r]

Sorry your sick of the viral campaigns?....however they do appear to have kicked AMDs' butt into doing something about it....Not sure what the sweeping under the rug is supposed to imply. NV is useless now XF has caught up?

I acknowledged that already, this particular one turned out to have merit.

I'm sick of the these huge viral marketing campaigns/issues (in some cases valid, mostly not imo) which when the issue is fixed, reversed, or found untrue suddenly everything is quietly ignored.

At least give credit (good or bad) where due. Changing the goalposts every new card release is tiresome to watch.

 

[boxleitnerb]

What's most important when considering smoothness with multi-GPU is fps.
With AFR, 50 fps will obviously be smoother than 35 or 40. Framepacing/metering generally works better at higher fps than at lower fps.

That said, it is difficult to compare smoothness in the [H] review because of the - sometimes significant - performance differences between the AMD and Nvidia solutions, despite their playable settings approach.

I still believe if they had tested the CF setup in quiet mode or the NV setup with higher temperature allowance, the results would have been different - fps wise and in terms of smoothness as well.

 

[ICDP]

I have to be honest, the only issues I had with these viral campaigns was the single AMD GPU being lumped in with CF. A few games such as Skyrim and Borderlands 2 had single GPU stutter issues that could be fixed with profile/driver updates.

Crossfire on the other hand was a complete mess out of the box. I used CF 7950s for a while and the stutter was abysmal and unplayable without some extensive tweaking. It could be fixed with some FPS caps and vsync etc but that did not excuse AMD for allowing such serious problems in the first place. Why should we give AMD credit for fixing what should have been fixed years before?

If I purchase a car with a faulty gearbox and it takes the dealer years to admit there is a problem. Should I praise him for finally fixing it? Don't answer that, it's a rhetorical question.

 

[Erenhardt]

I have to be honest, the only issues I had with these viral campaigns was the single AMD GPU being lumped in with CF. A few games such as Skyrim and Borderlands 2 had single GPU stutter issues that could be fixed with profile/driver updates.

Crossfire on the other hand was a complete mess out of the box. I used CF 7950s for a while and the stutter was abysmal and unplayable without some extensive tweaking. It could be fixed with some FPS caps and vsync etc but that did not excuse AMD for allowing such serious problems in the first place. Why should we give AMD credit for fixing what should have been fixed years before?

If I purchase a car with a faulty gearbox and it takes the dealer years to admit there is a problem. Should I praise him for finally fixing it? Don't answer that, it's a rhetorical question.

I think it is more like expecting '70 Camaro to run on nano-fuel

 

[Teizo]

OP says he's sick of viral campaigns, then starts his own. Nice.

That said, glad AMD got their CF bug worked out.