After finally and successfully unlocking my X3 720 BE, I noticed AOD reporting the PCIe BUS speed as 108 MHz. I know AOD has issues with reporting the temperatures of unlocked CPUs, as do other monitoring programs. This PCIe BUS speed is baffling me.
I've got a GigaByte GA-MA790X-UD4P motherboard running the F5 BIOS. I know there have been several BIOS revisions since the F5's release last June, but I've not updated because none of the other versions seem as though they'd have a direct impact on my current setup. Though, I doubt the BIOS is the source of the problem, unless it's directly caused by the identification of the X3 720 as an "X4 20," as you can see here:
The really messed up part is when I attempt to change the PCIe BUS speed within AOD back to the standard 100. It immediately jumps to 143 with a big red box around it. The closest I can get it to 100 within AOD is 101, and even then it actually sets to 102 and runs at 101.6 MHz.
If I manually set the PCIe BUS speed to 100 within the BIOS, it runs at 99.9 MHz, which seems acceptable. But, in order to modify the PCIe speed, I have to also manually set the FSB speed. When it's set to 200, it actually clocks in at 201.4 MHz, adjusting my CPU frequency from what you see above to 3.222 GHz. I realize GigaByte's baseline setting is fluffed a bit, but it's fluffed even more when set manually.
I'm just totally baffled by this 108 MHz portion, and I'm kinda stuck and uncertain what to do. Perhaps there are other programs I could try to verify these speeds?
I've done some testing, and I now trust that AOD's report of 108 MHz is accurate. How it's possible the default PCIe speed for the CPU changes from 100 to 108 when the 4th core is unlocked is beyond my knowledge. Though it's certainly understandable considering the CPU is not correctly identified. It also falls in line with how monitoring software reads the Core temps as 0 or -256.
The reason I trust the AOD number is because I decided to bench the system using 3D-Mark 06. I ran the test the previous day with the triple-core settings at 3.5 GHz and my HD 4890 at 925/1050. I ran it last night with 4-cores at 3.21 GHz and my score dropped. I figured it was a clear sign that clock speed is more important for the graphics portions of the tests. I ran it again today after manually setting the PCIe speed to 101 in the BIOS. Here are the numbers:
Well, I don't know what it was - setting the HT multi manually, initializing "PEG" video vs "PCI Slot", or what - but I've managed to get the PCIe clock to 99.9 MHz while set to "AUTO" within the BIOS. I guess that will have to do.
I ran 3D Mark 06 w/ 4 cores @ 3.2 GHz, 4890 @ 925/1050
3DMark Score 16285 3DMarks
SM 2.0 Score 6445
SM 3.0 Score 7729
CPU Score 4467
Much more like what I expected, and aside from the CPU score, it's very similar to the 3-core @ 3.5 GHz scores. Seems the graphics tests benefited more from the clock speed as opposed to having more cores.
Would have been great to get some input from some other people, but I guess it wasn't quite so simple, huh?