AMD APU's

[BnBorg]

I am running four APUs, two E-350s, an A4-3400 and an A6-3670.

For the two E-350s, Core Temp 1.0 RC3 x64 reports about twice the bus speed but the same multiplier as CPU-Z, as has been previously reported. Core temp's on these APUs seem reasonable AFAIK and agree with what the CPUID HWMonitor 1.19.0 x64 shows for the integrated Radeon 6310. HWMonitor does not show core temps on these APUs, but they are about 10C higher than the Asrock motherboard CPU monitor. I have seen them vary from 46C to 72C. These machines are running Windows Server 2008 R2 x64, and have been very stable.

The A4-3400 is shown by Core Temp with the bus speed 35% too high (135 MHz instead of 100). The core temps on the A4 are wrong, showing up in the single digits usually (8 or 9C). Note, it is 83F (29C) in this room. The good news is that I do not think it is a bug in Core Temp 1.0 RC3, since it is in close agreement with both HWMonitor and the Biostar Temperature Monitor. This APU is on a Biostar A55MH motherboard (Hudson D2), running Windows 7 x64. This computer has problems with spontaneous shutdowns.

Core Temp shows the A6-3670 at 13C, also in close agreement with HWMonitor and Biostar. The bus frequencies are shown at 90 MHz where CPU-Z reports 103 MHz. This is on a Biostar TA75M motherboard (Hudson D3), currently running Windows 8 CP x64. This computer has had display corruption after waking up from S3 sleep.

I hope these reports help.

[kasparhauser]

This is just to corroborate BnBorg's observation re AMD E350. I have the same problem with the frequency displayed. The temperature, however, seems to be correct. Here's hoping for the next release.

And btw.: other that this minor glitch CoreTemp really is a very neat little program!

[The Coolest]

Sorry for the very late response, which is probably irrelevant by now but I will still try to reply.

There is a bug where the bus frequency on *all* APUs is doubled than what it is in real life. It will be fixed in the next release, coming early November.
I've never seen an issue like BnBorg described of the bus speed being 'randomly' inaccurate, can't even guess what could be wrong. This is the first time I hear about such an issue, with any platform.

Regarding the low temperatures. The sensors in these chips usually under-report by between 10-20C under load. When the CPU is idle, what I noticed is that temps go to a very low value. I assume this occurs due to AMD's power saving feature, which disables some of the thermal sensors on the chip, making the readings even more inaccurate than they are under load.

[The Coolest]

RC4 is now available, it fixes recognition problems of BD CPUs and Trinity APUs, fixes the bus speed bug with APUs.

[MWisBest]

RC4 is now available, it fixes recognition problems of BD CPUs and Trinity APUs, fixes the bus speed bug with APUs.

I'm still having issues with my A8-4500M (Trinity mobile APU). The bus speed is sitting there stuck at 135. CPU-Z reports it varying from about 100 to 115 as the frequency changes and whatnot, however CoreTemp doesn't even pick up on the bus speed increase and just stays there at 135 still, rather than going up by 15 to 150 or something like that. The multiplier is read correctly though, and the temperatures seem to be accurate.

My A8-3500M (Llano mobile APU) is working flawlessly.

[@lber]

Hi.
I post here, because it's the same problem I have, while it is not a real one.
I have a netbook (MSI Wind U270) with an E-450 AMD APU.
In 1.0 RC3 FSB speed reported 200MHz FSB. It's always the case in RC4.
while in RC3 the multiplier was at max 16x which would report a 3.3GHz CPU speed (it's an 1.6GHz dual core processor), RC4 version reports x32 multiplicator.
I now have a reported cpu speed of 6.4 GHz.
I think FSB speed is theoretically 100MHz.

Thanks.

@l

[The Coolest]

@lber

Please post a screenshot and a register dump from Core Temp (look in the 'Tools' menu)

[Sporkinator]

AMD APU's? I noticed a bug:

CoreTempBug.PNG

Notice the incorrect frequencies in Core Temp.

[@lber]

Hi, here you'll find screen capture & dump register.

Thanks.

On idle


On load (7zip bench test)

[The Coolest]

Thanks. Next release should have this problem sorted out.