Please help: 2 of the 4 290X are way too slow when performing dictionary attack
#21
epixoip:

Crossfire is ALWAYS disabled and "Performance/Overdrive" displays only GPU #1

When enabling crossfire it shows in the "Overdrive" -> "Select the GPU to be configured" dropdown (after an OS reboot) only GPU #1 and GPU #2
#22
hm, ok. maybe that is normal behavior for Windows then.
#23
Latest update AND IT IS NOT GOOD:

1) I swapped the 2 Asus which were on the MB with the 2 Sapphire which were connected to the USB risers. Result: The Saphire GPUs were working on the 1/3 of the Asus (so, no change compared to the previous situation).

2) I installed 1 Asus and 1 Sapphire on the MB and 1 Asus and 1 Sapphire on the USB risers. Result: consistent as the previous. The Sapphires were working at the 1/3 of the Asus.

So, it was definitely NOT the connection type, ie, it plays no role whether the GPU is connected on the MB or via the USB risers.

A small point I wish to make at this point: the troubleshooting and testing is notoriously difficult as it seems that the MB/OS/Plug&Play have their own minds regarding the #x number they assign to each GPU with every new test (unplug/re-plug any of the GPUs). To make matters even worse, -like previously said-, the AMD CCC sees ONLY 1 GPU (the rest of the 3 show as "disabled") and so does the ASUS GPU Tweak.
So, the only utility I have found and use is the Sapphire TRIXX which is able to see all 4. But there is no way to distinguish the cards as they all have the exact same name and so identification is brutally performed via FAN controlling (setting from 0 to 100% to distinguish which each card is).


So, what gives? At that point I saw only 2 possibilities:
a) all 4 cards perform the same but OCL somehow reports the performance of the 2 Sapphire at 1/3 of the ASUS ones
b) the Sapphire GPUs are either hardware or BIOS "inferior" to the Asus when it comes to Dictionary attacks

Since I couldn't do anything with (a) I decided to focus on (b).
Using TRIXX I saw that the only difference between the Asus and Sapphire was the BIOS version (I always set the Sapphire's clock and memory to match the Asus so there is no difference there (@1050/1350)).

Asus Bios: 15.39.0.6.3515
Sapphire Bios: 15.41.0.2.0

Searching the net I found a thread that deals with exactly this issue, ie, flash a Sapphire 290x to Asus 290x as the asus rom is considered better and fully unlocked (Thread: http://forums.overclockers.co.uk/showthr...t=18552408)

Having decided to proceed with flashing, I removed the 2nd Asus from the rig leaving in one Asus and the 2 Sapphire. I then proceeded and flashed one of the 2 Sapphire. So, once flashed I had 1 Asus, 1 Sapphire with the Asus BIOS and 1 Sapphire with the original BIOS (both the Asus and the flashed Sapphire reported in TRIXX "15.39.0.6.3515").
Result: the newly flashed Sapphire MATCHED FINALLY THE SPEED OF THE ASUS!!!!!! (whereas the "original" Sapphire still worked at the 1/3 of the Asus speed).

Then I rebooted and OCL retested. Results remained the same, so I proceeded to flash the 2nd Sapphire. Once done, BIOS version matched the other 2. So, all 3 cards now had the same BIOS (the Asus). Rebooted and proceeded to test in OCL...

RESULT: scratching my head in disbelief I couldn't -and still can't- comprehend what was in front of my eyes... the 2 Sapphire GPUs were back working on 1/3 of the Asus and their BIOS version had reverted back to the original "15.41.0.2.0".

At this point I am at a total loss! I don't have a clue how the cards returned back to the original BIOS version (they could be locked but how can someone explain why when the first flashed, still reported the newly flashed BIOS AFTER THE OS reboot?).
The problem seems to be with the Sapphire cards but I don't have a clue what to do next nor how to proceed.
I will be waiting for the delivery of the HDMI fake plugs so at least the AMD utilities will be able to see all 4 cards and take it from there.....
#24
1) Did you check you didn't accidentally toggle the switch to swap from BIOS1-> BIOS2 on the VGAs ? May explain why the BIOS managed to restore themselves.

2) Did you try flashing the latest BIOS on the motherboard?
#25
blazer:

1)if you are referring to the "normal/uber" switch, no, I haven't touched it during the troubleshooting. It has been left at "uber" since day #1

2) have to recheck the BIOS version. A couple of months ago it was at the latest
#26
I am not sure if you stated this but have you tried having only one card in the computer at a time and flash one card at a time then put all 4 cards in. Then i would leave the 2 saphire cards in the MB just to make sure as little conflict can happen with those cards since they seem to be glitchy.
#27
SomeGuy:

Yes, I have this in mind. Once I receive the HDMI dummy plugs it is one of the things I will try.
#28
(07-25-2014, 08:26 PM)john2014 Wrote: SomeGuy:

Yes, I have this in mind. Once I receive the HDMI dummy plugs it is one of the things I will try.

well best of luck to you. I will keep an eye on this and remember this issue when i build my next computer.