The-Funster
asked on
AMD A8-5600K Crossfire performance problem
Hi everyone.
I'm in dire need for assistance of someone who knows about crossfire using an AMD A8-5600K/Sapphire HD6670 1GB PCI-E card.
The System's specs are:
Sapphire HD 6670 1GB GDDR5 PCI-E Graphics Card
AMD A8 5600K Black Edition 3.6GHz
Gigabyte GA-F2A85XM-D3H A85X Socket FM2 Mainboard
Seagate 1TB Barracuda 3.5" SATA-III Hard Drive - 7200RPM 64MB Cache
Kingston 8GB (2x4GB) DDR3 1600MHz HyperX Genesis Memory
CIT 750W Power supply
Microsoft Windows 8 64-bit
I based my choice of hardware on online reviews and benchmarks results.
Crossfire is enabled according to the Vision Engine Control Centre and the latest AMD drivers are installed. Otherwise all other hardware and system settings are unmodified.
I thoroughly benchmarked the system with several synthetic tools and several games prior to enabling crossfire. I then enabled crossfire and retested. The results are :
Before After
3DMark Vantage 7386 8306
3DMark 2006 11995 14175
3DMark 2011 1794 2203
PCMark 7 2682 2687
Unigine Heaven fps 35.5 fps 29.4
Score 893 Score 741
Min fps 19.8 Min fps 16.1
Max fps 68.8 Max fps 64
As you can see, apart from the Unigine scores, every benchmark has improved after enabling crossfire. (Although not as much as I was led to expect.)
With fps in games however, things are far from convincing.
Before After
Dirt 375 69
Fraps Portal 284 300
Fraps Skyrim 54 52
Fraps Alan Wake 49 60
Fraps Far Cry 3 16 18
Fraps Sniper 70 48
Fraps Aliens CM 59 54
As you can see, performance either drops or improves only minimally.
So my question is this. Can anyone advise me on what I need to configure in order to get the performance increases looking like the ones I have seen in all the reviews of AMD APUs using crossfire with the HD6670.
Thanks
The Funster
I'm in dire need for assistance of someone who knows about crossfire using an AMD A8-5600K/Sapphire HD6670 1GB PCI-E card.
The System's specs are:
Sapphire HD 6670 1GB GDDR5 PCI-E Graphics Card
AMD A8 5600K Black Edition 3.6GHz
Gigabyte GA-F2A85XM-D3H A85X Socket FM2 Mainboard
Seagate 1TB Barracuda 3.5" SATA-III Hard Drive - 7200RPM 64MB Cache
Kingston 8GB (2x4GB) DDR3 1600MHz HyperX Genesis Memory
CIT 750W Power supply
Microsoft Windows 8 64-bit
I based my choice of hardware on online reviews and benchmarks results.
Crossfire is enabled according to the Vision Engine Control Centre and the latest AMD drivers are installed. Otherwise all other hardware and system settings are unmodified.
I thoroughly benchmarked the system with several synthetic tools and several games prior to enabling crossfire. I then enabled crossfire and retested. The results are :
Before After
3DMark Vantage 7386 8306
3DMark 2006 11995 14175
3DMark 2011 1794 2203
PCMark 7 2682 2687
Unigine Heaven fps 35.5 fps 29.4
Score 893 Score 741
Min fps 19.8 Min fps 16.1
Max fps 68.8 Max fps 64
As you can see, apart from the Unigine scores, every benchmark has improved after enabling crossfire. (Although not as much as I was led to expect.)
With fps in games however, things are far from convincing.
Before After
Dirt 375 69
Fraps Portal 284 300
Fraps Skyrim 54 52
Fraps Alan Wake 49 60
Fraps Far Cry 3 16 18
Fraps Sniper 70 48
Fraps Aliens CM 59 54
As you can see, performance either drops or improves only minimally.
So my question is this. Can anyone advise me on what I need to configure in order to get the performance increases looking like the ones I have seen in all the reviews of AMD APUs using crossfire with the HD6670.
Thanks
The Funster
Also, a fundamental flaw in benchmarking is not understanding what any specific benchmark actually does, and whether or not drivers and hardware is optimized to perform outrageously well on any particular benchmark.
That is why there is no substitute for running whatever applications you ordinarily run and using them as a basis to measure performance.
In your case, you proved that your particular games are not compiled and optimized to take advantage of this new hardware/engine combination.
Before wasting any more time fretting about benchmarks, you should contact the software vendor for your games and ask about any updates, drivers, patches that take advantage of this new card. You will likely discover (i have no idea, I'm just responding from perspective of somebody who has done a heck of a lot of code optimization over the years) that the vendor doesn't take advantage of this new functionality in their code and that the higher benchmark numbers will only manifest themselves in code that is optimized to take advantage of them.
In other words ... the card is only faster when running code that has been tweaked to take advantage of the newer functionality, and much slower on legacy code because now the hardware has to emulate the old way of doing things.
So check for an update that takes advantage of the new features.
That is why there is no substitute for running whatever applications you ordinarily run and using them as a basis to measure performance.
In your case, you proved that your particular games are not compiled and optimized to take advantage of this new hardware/engine combination.
Before wasting any more time fretting about benchmarks, you should contact the software vendor for your games and ask about any updates, drivers, patches that take advantage of this new card. You will likely discover (i have no idea, I'm just responding from perspective of somebody who has done a heck of a lot of code optimization over the years) that the vendor doesn't take advantage of this new functionality in their code and that the higher benchmark numbers will only manifest themselves in code that is optimized to take advantage of them.
In other words ... the card is only faster when running code that has been tweaked to take advantage of the newer functionality, and much slower on legacy code because now the hardware has to emulate the old way of doing things.
So check for an update that takes advantage of the new features.
ASKER
Hi.
Thanks for your replies.
I've ran benchmarks and checked fps in games while monitoring for resource usage as you recommended but other than the running game/benchmarking software nothing shows as using any significant resources.
With regard to crossfire compatibility, all the tested games are patched and up to date. Also, comments from other crossfire users suggest that few games have negative performance issues using crossfire. According to some users Far Cry 3, Dirt 3, Alan Wake and Aliens Colonial Marines improve greatly using crossfire. In fact it would seem that many older games improve fps with crossfire enabled even though they have no crossfire specific support.
I appreciate your comments however.
Thanks.
The Funster
Thanks for your replies.
I've ran benchmarks and checked fps in games while monitoring for resource usage as you recommended but other than the running game/benchmarking software nothing shows as using any significant resources.
With regard to crossfire compatibility, all the tested games are patched and up to date. Also, comments from other crossfire users suggest that few games have negative performance issues using crossfire. According to some users Far Cry 3, Dirt 3, Alan Wake and Aliens Colonial Marines improve greatly using crossfire. In fact it would seem that many older games improve fps with crossfire enabled even though they have no crossfire specific support.
I appreciate your comments however.
Thanks.
The Funster
So then if we cut to the chase, the problems are limited to some games you run on your system. So curious, can you do a fresh install of dirt or one of the others that shows the greatest hit on a USB stick and then run benches from the stick with a fresh copy?
Maybe your HDD is pounded looking for data or due to other things that you would only see with a lot of historical data and files from a game that has a lot of history?
Also when you looked at the perf monitor, look for the QUEUE DEPTH. If that number is consistently > 1, then even though it might show 10% I/O utilization, root cause for that is because the computer is waiting for the disk drive to give it data. So unless you look at the queue depth then in this case, that 10% utilization would mislead you into thinking your problem was the disk drive.
By doing a fresh install on a USB stick, and then looking at the numbers you could learn if disk I/O was a big factor. let's face it, reviewers are going to be using fresh installations on SSDs and other disks that won't tend to be bottlenecks.
YOur numbers indicate the bottleneck is NOT the graphics card. So you need to dig further. (Network traffic, possible virus,etc..) but check disk first
Maybe your HDD is pounded looking for data or due to other things that you would only see with a lot of historical data and files from a game that has a lot of history?
Also when you looked at the perf monitor, look for the QUEUE DEPTH. If that number is consistently > 1, then even though it might show 10% I/O utilization, root cause for that is because the computer is waiting for the disk drive to give it data. So unless you look at the queue depth then in this case, that 10% utilization would mislead you into thinking your problem was the disk drive.
By doing a fresh install on a USB stick, and then looking at the numbers you could learn if disk I/O was a big factor. let's face it, reviewers are going to be using fresh installations on SSDs and other disks that won't tend to be bottlenecks.
YOur numbers indicate the bottleneck is NOT the graphics card. So you need to dig further. (Network traffic, possible virus,etc..) but check disk first
Just as a first thought:
As you're using hybrid Crossfire you really should plug your monitor into the card with the higher spec, as your Trinity APU is a 7560D are you connected to that rather than the 6670?
Are any of the components already clocked?
As you're using hybrid Crossfire you really should plug your monitor into the card with the higher spec, as your Trinity APU is a 7560D are you connected to that rather than the 6670?
Are any of the components already clocked?
ASKER
Hi. Thank you both for your replies.
The benchmarked system is a fresh install of Windows 8 with only drivers, Windows updates and games updates installed. Fraps FPS counter logged frame speeds during normal gameplay. Disk activity is 0 for almost all the in game test periods with the exception of moving into a different game area in Aliens Colonial Marines where I moved between zones. The system was taken off the network once all updates etc. were competed.
Today I tested the same games on a clean install of Windows 7 on a different HDD and the results are pretty much the same.
I'm beginning to wonder if the mainboard I bought does not have reliable support for Dual Graphics. (Although Gigabyte make a lot of noise about the benefits of this feature.)
Component speeds have not been modified. I've checked and rechecked the manual for any possible missed configuration settings but all the boxes are checked.
I've contacted AMD for advice but I don't know how long they normally take to reply. I think it's going to be a case of hurry up and wait.
Thanks.
The Funster.
The benchmarked system is a fresh install of Windows 8 with only drivers, Windows updates and games updates installed. Fraps FPS counter logged frame speeds during normal gameplay. Disk activity is 0 for almost all the in game test periods with the exception of moving into a different game area in Aliens Colonial Marines where I moved between zones. The system was taken off the network once all updates etc. were competed.
Today I tested the same games on a clean install of Windows 7 on a different HDD and the results are pretty much the same.
I'm beginning to wonder if the mainboard I bought does not have reliable support for Dual Graphics. (Although Gigabyte make a lot of noise about the benefits of this feature.)
Component speeds have not been modified. I've checked and rechecked the manual for any possible missed configuration settings but all the boxes are checked.
I've contacted AMD for advice but I don't know how long they normally take to reply. I think it's going to be a case of hurry up and wait.
Thanks.
The Funster.
SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
You really are asking a lot from combining such different GPUs via hybrid Crossfire though
Look at the benchmarking for mixing the Trinity and 6670
Look at the benchmarking for mixing the Trinity and 6670
ASKER
I accepted my own comment as a solution because none of the other comments actually got my issue resolved. However, dlethe offered some very useful advice and so I have awarded him points for at least sticking with me.
Run performance monitor in another window while doing your benchmarks and then you can see if the system is starved for RAM, I/O, or CPU.