AMD llano cpumark artificially inflated scores

[David (PassMark)]

See this post for details about how a CPU bug might be impacting the results.
http://www.passmark.com/forum/showthread.php?t=3656

[Omega73]

We are looking into it, but haven't come to any conclusion at the moment. Current line of thought is that it might be a CPU bug. But it isn't conclusive at the moment.

Can you try going into the Edit / Preferences window in PerformanceTest. Then selecting a different number of processes. Then post your result for the integer test for, 1,2,3 & 4 processes

I did other tests.
I replaced the cpu A6-3650 with A6-3670 but the results are the same.

System Test:
http://img.overpic.net/images/f/9/m/xf9m9pw2416mg3nac72mw.jpg

4 processes
http://overpic.net/viewer.php?file=xjssf1gaishpnou9c2kl.jpg

3 processes:
http://overpic.net/viewer.php?file=xan1bnwiy7n807jstvjhw.jpg

2 processes:
http://overpic.net/viewer.php?file=xshxeiiekxztuytgpemkx.jpg

1 processes:
http://overpic.net/viewer.php?file=x46lz9ah01epca2k0nv.jpg

CPU-id:


I read your conclusions on BUG-CPU:
http://www.passmark.com/forum/showthread.php?t=3656

I have a lot of CPU but I have different motherboards foxconn to be tested.
Currently all of the BIOS for the motherboard Foxconn give the problem!!

[David (PassMark)]

Thanks for that. As the number is threads is reduced from 4 to 1 the result goes from bad, to really really bad.
Integer maths 547 with 4 threads. 137 with 1 thread.

The implication is that all threads are in fact running, but slowly, and contributing to the result. (The other alternative was that 3 of the 4 threads had locked up, and the low result was due to threads not finishing)

On Monday next week we'll see if we can get a public patch for the bug CPU prepared, as discussed in the other bug thread. Then if possible I'll get you to try it out to see if it works around the problem.

[Ian (PassMark)]

We have made an application that applies the work around for the AMD Llano errata 665. This patch is only temporary until the next reboot, but if you are seeing very low Integer and Prime number CPU results in PerformanceTest, or having CPU hangs, then you can run this application to see if the problem is caused by CPU errata 665.

From our experience, it would appear that few BIOS's currently implement this work-around.

The 32-bit and 64-bit patches are here:
http://www.passmark.com/ftp/LlanoCPUerrata665.zip


The AMD A-Series CPU (Llano) has the following errata:

"665 Integer Divide Instruction May Cause Unpredictable Behavior

Description
Under a highly specific and detailed set of internal timing conditions, the processor core may abort a speculative DIV or IDIV integer divide instruction (due to the speculative execution being redirected,
for example due to a mispredicted branch) but may hang or prematurely complete the first instruction of the non-speculative path.

Potential Effect on System
Unpredictable system behavior, usually resulting in a system hang.

Suggested Workaround
BIOS should set MSRC001_1029[31].
This workaround alters the DIV/IDIV instruction latency specified in the Software Optimization Guide for AMD Family 10h and 12h Processors, order# 40546. With this workaround applied, the DIV/IDIV latency for AMD Family 12h Processors are similar to the DIV/IDIV latency for AMD Family 10h Processors.

Fix Planned
No"

[Omega73]

Hello, I tested the Patch.
Unfortunately I have not improved anything...

BEFORE PATCH



PATCH


AFTER PATCH

[Ian (PassMark)]

In your case it would appear that the patch was already applied by your BIOS - so actually our patch didn't do anything in this case.

[Ian (PassMark)]

I still think the most likely cause of the low Integer and prime numbers test results relate to the 12h series CPU errata 665. To try and determine this, I have updated the software such that if the patch is already applied, the patch can now be removed (until the next reboot).

Once this patch is removed, the CPU Integer result should increase dramatically. It may unfortunately also have the effect of hanging your system during the test. In the testing I have done on a system here, I did not get a system hang. This is just a test to help determine the cause of the problem - not a solution.

Please try the updated patch utility from here:
http://www.passmark.com/ftp/LlanoCPUerrata665_v1.1.zip (Now out of date, see below for new version)

Please let us know what effect this has on your CPU performance results.

Thanks.

[Omega73]

I tested the patch v.1.1 on my motherboard Foxconn A75M with the lastest bios (B44F1P06) and finally found some improvement!

System Test:
http://img.overpic.net/images/u/b/w/xubwhxgcaxnj3o0gs7y5w.jpg

Before Patch:


Patch v.1.1


AFTER PATCH v1.1

[Ian (PassMark)]

Thanks for the feedback. While I would have expected the Integer and Prime number results to be even higher, it does appear that the CPU performance issue you are seeing is related to CPU errata 665, possibly also with other factors we are not aware of. I will ask AMD for comment and post the result here.

[Ian (PassMark)]

OK, we have worked it out. The cause of the low PerformanceTest Integer and Prime number results is due to the workaround implemented in BIOS for Llano CPU errata 665.

V1.1 of our patch tool only applies/removes the patch to 1 CPU core and it needs to be applied/removed from all CPU cores. We have updated our utility to do this. Please try our updated patch utility to see the impact on performance, v1.2, from here:
http://www.passmark.com/ftp/LlanoCPUerrata665_v1.2.zip

We tested this on an A6-3650 system in our lab and found that with the patch not applied the CPU results were:
Integer test: 3074
Prime number: 2054

When the work around patch for Llano errata 665 is applied the CPU results dropped to:
Integer test: 528
Prime number: 751

The CPU tests that do not utilize integer divide are unaffected.

This is consistent with the baselines submitted to the PassMark benchmark site for Llano CPUs. Where the high CPU results are for systems without the BIOS patch for errata 665, while the lower results are for systems with the BIOS patch.