Crossfire 7870's


Recommended Posts

I have

http://www.newegg.com/Product/Product.aspx?Item=N82E16814102981

 

I'm looking at a deal for one of these

http://www.newegg.com/Product/Product.aspx?Item=N82E16814202024

 

Would I lose performance with slightly different cards or would I be best to match my card as much as possible?

It's Tahiti vs Pitcairn architecture. I don't know what that means, perhaps you do?

Link to comment
Share on other sites

As far as I can tell, the 7870 XT cannot crossfire with any 7800 series cards. You'll need a 7850, 7870 (GHz), or R9-270(x) to double-up with what you have right now.

Link to comment
Share on other sites

You may want to look up micro-stuttering. ATI card tend to have it still in crossfire AFAIK:

 

http://www.tomshardware.com/reviews/radeon-geforce-stutter-crossfire,2995.html

Now I am not disagreeing with you as MS is in both nVidia and ATI/AMD but you really should use a more recent article than that "By Igor WallossekGreg Ryder

AUGUST 21, 2011"
  • Like 1
Link to comment
Share on other sites

Microstuttering in crossfire setups is nowhere near what it was back then. I've been using dual 7970's for over half a year now, and it's never been a problem for me. Especially not after the frame pacing drivers that AMD released not too long ago.

 

Also, yes, that card will work with yours. Any 78xx card will work with any other 78xx card. The only exception is the 7870 XT, which uses a Tahiti core. The R9-270 & R9-270x will both work as well since they have the same core.

  • Like 2
Link to comment
Share on other sites

 

Now I am not disagreeing with you as MS is in both nVidia and ATI/AMD but you really should use a more recent article than that "By Igor WallossekGreg Ryder

AUGUST 21, 2011"

 

 

Honestly, you need look up what you are saying instead of concluding that what I linked wasn't relevant because of the date. Perhaps, I was simply providing an article that demonstrated the issue for the OP in case he didn't know about it. Moreover, perhaps, I specifically mentioned ATI cards and told the OP to look up the issue for a reason:

 

http://www.tomshardware.com/news/AMD-Radeon-CrossFire-Drivers-Driver,23214.html

http://en.wikipedia.org/wiki/Micro_stuttering

 

I didn't want to get into it before because there really shouldn't be an argument about it, but SLI hasn't had an issue for years. By far the largest offender in micro-stuttering has been AMD. Earlier this year, nVidia even provided a tool called FCAT as a marketing gimmick just to show how ATI cards still had issues with micro-stuttering. I could provide benchmark articles, but I am fairly sure you are capable of searching yourself for benchmarks that show the results with CrossFire and SLI. Unless something has changed drastically in the last 6 months: AMD still has issues. And, I really doubt anything has drastically changed since architecture issues don't particularly fix themselves with driver updates. You can alleviate some of the issues, but software solutions for hardware problems generally don't end so well.

 

Not, that I would ever buy NV cards since they tend to be much less cost effective in terms of price/performance ratios.

Link to comment
Share on other sites

August's frame pacing drivers are the first in a stream of updates designed to implement a proper frame metering solution, which AMD/ATi has historically lacked (as pointed out in that first aritcle you linked). The initial release only supports Dx11 titles in the standard range of resolutions (everything below 1440p, if memory servers) on single-monitor setups. The next revision is expected to include improved performance, support for Dx9 and 10, higher resolutions, and eyefinity. I believe AMD claimed that it would be released in late fall, but it looks like that got delayed. Not really surprising. It took them several months to get the first release stable enough for a beta release. I heard that the early preview version sent out to a few reviewers was unbelievably buggy. The proper beta release was pretty stable though.

  • Like 1
Link to comment
Share on other sites

August's frame pacing drivers are the first in a stream of updates designed to implement a proper frame metering solution, which AMD/ATi has historically lacked (as pointed out in that first aritcle you linked). The initial release only supports Dx11 titles in the standard range of resolutions (everything below 1440p, if memory servers) on single-monitor setups. The next revision is expected to include improved performance, support for Dx9 and 10, higher resolutions, and eyefinity. I believe AMD claimed that it would be released in late fall, but it looks like that got delayed. Not really surprising. It took them several months to get the first release stable enough for a beta release. I heard that the early preview version sent out to a few reviewers was unbelievably buggy. The proper beta release was pretty stable though.

 

I still question whether they will be able to provide a silver bullet via software-only in the end. I think it is something that is going to require a hardware redesign to completely eliminate. Perhaps, that is why the stalls have come into play?

 

As for the practicality of the issue, I don't have a dual-card setup so I can't comment on how perceptible it is to the average person at the end of the day though.

Link to comment
Share on other sites

Here's a link to benchmarks and things with the frame-pacing drivers:

http://www.tomshardware.com/reviews/catalyst-13.8-frame-pacing-crossfire,3595-10.html

 

EDIT: actually these results are much better than I anticipated. Though, not perfect.

so my point exactly, its not as bad as your 2011 article puts it, were in 2013 theres a reason i said your article was outdated

Link to comment
Share on other sites

I still question whether they will be able to provide a silver bullet via software-only in the end. I think it is something that is going to require a hardware redesign to completely eliminate. Perhaps, that is why the stalls have come into play?

 

As for the practicality of the issue, I don't have a dual-card setup so I can't comment on how perceptible it is to the average person at the end of the day though.

It can be done entirely on software. Not perfectly, of course, but it's not like a hardware solution would fix it either. You'd probably need two GPUs exactly right next to each other on the same PCB with access to the same memory in order to totally eliminate any latency that could lead to that. Not even the dual-GPU card solutions from either team offer that. The only thing that can be done is to hack away at the severity of the issue until it's so hard to notice that nobody complains anymore. AMD's software-only approach seems to be working well so far. It's a bright future if they keep expanding on it.

Link to comment
Share on other sites

so my point exactly, its not as bad as your 2011 article puts it, were in 2013 theres a reason i said your article was outdated

 

I'm honestly not sure what to make of that given that you downplayed the issue to sound like nothing and also implied NV cards have the same problem. :huh:

Link to comment
Share on other sites

It can be done entirely on software. Not perfectly, of course, but it's not like a hardware solution would fix it either. You'd probably need two GPUs exactly right next to each other on the same PCB with access to the same memory in order to totally eliminate any latency that could lead to that. Not even the dual-GPU card solutions from either team offer that. The only thing that can be done is to hack away at the severity of the issue until it's so hard to notice that nobody complains anymore. AMD's software-only approach seems to be working well so far. It's a bright future if they keep expanding on it.

 

Of course, you can't eliminate it completely*, but hardware solutions do go a long way. Does NV even have frame pacing in their drivers? I'm not aware of it if they do (I can't say I've particularly looked very hard though). I think the difference in architecture alleviates the issue much on its own.

 

* That is a little too strong. Some clever hardware engineers may make the issue a thing of the past at some point. You never know. Operate on the same clock tick somehow?

 

EDIT: found this: http://developer.download.nvidia.com/whitepapers/2011/SLI_Best_Practices_2011_Feb.pdf

  • Like 1
Link to comment
Share on other sites

I'm honestly not sure what to make of that given that you downplayed the issue to sound like nothing and also implied NV cards have the same problem. :huh:

AMD has been catching a lot of flack for the issue in recent years since they never offered a proper solution to fix it. nVidia jumped on it much sooner. A few years ago, they released a software-based approach as well, and then followed it up with a hardware solution. I don't know the full details of what that means, but I'd be willing to bet it's about 50% effective and 50% marketing. Whatever it is, it works, but it's not perfect.

 

* That is a little too strong. Some clever hardware engineers may make the issue a thing of the past at some point. You never know. Operate on the same clock tick somehow?

Of course, you can't eliminate it completely*, but hardware solutions do go a long way. Does NV even have frame pacing in their drivers? I'm not aware of it if they do (I can't say I've particularly looked very hard though). I think the difference in architecture alleviates the issue much on its own.

Again, they have bragged about some kind of hardware solution, but I don't know the details. I'm sure there's a software fix in their drivers somewhere as well. Even today, nVidia cards do suffer from microstuttering issues as well. It just doesn't get much attention because AMD/ATi has overshadowed them there. It's an inherent problem with any multi-GPU setup, regardless of which vendor is chosen. As you said, maybe there will be one day in the future when it's not an issue anymore, but I don't see that day on any calendar for the next few years.

  • Like 1
Link to comment
Share on other sites

AMD has been catching a lot of flack for the issue in recent years since they never offered a proper solution to fix it. nVidia jumped on it much sooner. A few years ago, they released a software-based approach as well, and then followed it up with a hardware solution. I don't know the full details of what that means, but I'd be willing to bet it's about 50% effective and 50% marketing. Whatever it is, it works, but it's not perfect.

 

Again, they have bragged about some kind of hardware solution, but I don't know the details. I'm sure there's a software fix in their drivers somewhere as well. Even today, nVidia cards do suffer from microstuttering issues as well. It just doesn't get much attention because AMD/ATi has overshadowed them there. It's an inherent problem with any multi-GPU setup, regardless of which vendor is chosen. As you said, maybe there will be one day in the future when it's not an issue anymore, but I don't see that day on any calendar for the next few years.

 

Looks like the hardware solution was this: 

http://us.hardware.info/reviews/2654/7/nvidia-geforce-gtx-680-reviewed-txaa--adaptive-vsync

http://www.tomshardware.com/reviews/radeon-hd-7990-devil13-7970-x2,3329-10.html

 

Moreover, looks like it worked pretty well in software (on AMD cards, next page) as well according those results so I wonder why AMD Hasn't done it this way...

Link to comment
Share on other sites

Microstuttering in crossfire setups is nowhere near what it was back then. I've been using dual 7970's for over half a year now, and it's never been a problem for me. Especially not after the frame pacing drivers that AMD released not too long ago.

 

Also, yes, that card will work with yours. Any 78xx card will work with any other 78xx card. The only exception is the 7870 XT, which uses a Tahiti core. The R9-270 & R9-270x will both work as well since they have the same core.

Thank you. So different Sapphire 7870 cards running at 1000 and 1050mhz difference in core clock performance won't take a performance hit over say 2 identical 1000mhz?

 

I looked over the micro-stuttering article as well, I was not aware of it but not concerned with it regarding the 7800 series. Help appreciated guys.

Link to comment
Share on other sites

Thank you. So different Sapphire 7870 cards running at 1000 and 1050mhz difference in core clock performance won't take a performance hit over say 2 identical 1000mhz?

 

I looked over the micro-stuttering article as well, I was not aware of it but not concerned with it regarding the 7800 series. Help appreciated guys.

 

I don't know myself. One problem is benchmarking websites almost never do that kind of configuration

Link to comment
Share on other sites

Looks like the hardware solution was this: 

http://us.hardware.info/reviews/2654/7/nvidia-geforce-gtx-680-reviewed-txaa--adaptive-vsync

http://www.tomshardware.com/reviews/radeon-hd-7990-devil13-7970-x2,3329-10.html

 

Moreover, looks like it worked pretty well in software (on AMD cards, next page) as well according those results so I wonder why AMD Hasn't done it this way...

AMD got a huge slap in the face earlier this year from http://www.pcper.com/ , so right now AMD is just releasing a software fix until they can work in a hardware version an that does take sometime to fix. So maybe late next year in the next-gen of AMD graphic cards we'll see a hardware fix in place.

  • Like 1
Link to comment
Share on other sites

Thank you. So different Sapphire 7870 cards running at 1000 and 1050mhz difference in core clock performance won't take a performance hit over say 2 identical 1000mhz?

 

I looked over the micro-stuttering article as well, I was not aware of it but not concerned with it regarding the 7800 series. Help appreciated guys.

I can't find the article to save my life, but I'm pretty sure someone tried comparing crossfire results from cards of different clocks with the same cards downclocked to the slower one. The results showed that there was still a noticeable increase despite only one of the two cards being clocked higher. This goes against the general idea that crossfire depends on the "weakest link" in the chain and is limited to the speed of the slowest card. If all else fails, you could always just overclock the slower one ot the GHz edition speeds. From what I've seen, virtally any stock 7x70 can bump right up to the GHz speeds at stock voltage, no problem.

 

Microstuttering exists on any multi-GPU setup, even 78x0-based. But again, it's a bit of an overglorified issue. Just make sure you install the latest beta drivers, and keep an eye out for new ones.

 

On that note, be prepared: multi-GPU will give you many headaches. It can work, but you'll need to be ready to fight with it. I've had to reinstall my drivers probably ten times over the past few months just because I keep encountering strange issues.

Link to comment
Share on other sites

This topic is now closed to further replies.
  • Recently Browsing   0 members

    • No registered users viewing this page.