TR Forums

tldr: This is just an unusual experience report. It wasn't anything I couldn't fix on my own; I'm just sharing a little story.

So, leading up to the holidays, I built myself a new computer based on an X6 1100T and the M5A97 EVO motherboard. I also got one of the swanky all-in-one water cooling kits to go with it in a Define R3 case, and wanted to do some modest overclocking. So with Prime95 in hand, I fire up AMD overdrive and find... all six cores running at ~3850 MHz? Stock? At idle!? Double-check task manager, yes it really is idle. Then I started up the Asus motherboard management software and see... 1.55V core? By this point in the machine's lifetime, it has gone through a Windows 7 install, played a few hours of Civ V, compiled some code, and has run some Debian Sid, all with no trouble. So the machine is at least stable enough to go through most daily tasks at those settings. But a torture test run of Prime95 led to thermal throttling of the CPU in just a few minutes. Under Linux, Prime95 destabilized the machine and rapidly led to a kernel panic.

I turns out that this motherboard had a HT clock of 233 MHz, with all of the voltage control sliders maxed out**, and with turbo core and cool'n'quiet disabled. I set the clock back to 200 MHz, re-enabled turbo core and C'n'Q, and dialed the voltage control back to normal. With those settings I modestly overclocked the machine by multiplier adjustment to 3.8/4.1 GHz for a prime95-stable system. I didn't even try to find the failure point, I just found a happy place and stopped there.

In the end, I'm happy with the machine I've got. But it was very strange to go through this. My first guess is that this is someone's returned motherboard after they did some of their own overclocking. Except that it was not marked as an open box SKU, and all of the packaging looked factory new when I opened it up. A friend mentioned that maybe the factory end-of-line test setup could have gotten cut short, such that the board was not reset to standard settings before packaging. I suppose that's possible, but I don't have any evidence to either support or refute the idea.

I'm not asking for help in any way, just curious if anyone else has a similar story to share.

PS: The only supported resolution for the GUI setup interface is not compatible with my monitor (a 2004 vintage 20" apple cinema display), so I had to lug the machine over to the TV to actually perform any of the aforementioned BIOS configuration changes. Others have complained about this problem on the asus forums, only usually in reverse: for them, it's the TV that doesn't support the resolution. Grumble, grumble.

** Voltage control on the EVO motherboard is not as simple as just picking a voltage, and it is not 100% clear what each of the sliders actually do other than, "Higher setting increases the voltage. Sometimes."

I generally clear the CMOS on new boards right out of the box. I've seen boards come with weird (non-default) settings too, so it is the only way to be sure you're starting from a clean slate.

I have the same board and I must say the BIOS acts strangely at times and you really have to take it easy on the VRM stuff to keep them running cool. Regarding the BIOS the first thing I do when I fire up a system is go into the BIOS and check everything out or at least do a reset.

I have very good water cooling and airflow and still this thing likes to throttle like crazy and does not like high voltage. I was hoping to use 1.5v or so but it was just not possible without throttling (which sucks because I'm pretty sure I could've hit 4400MHz+ since it was running fine except for the throttling). I have run 1.55v or so on similar chips on older **** boards for years without problems and this one is supposed to have good VRM but I guess maybe the non-digi or the Gigabyte may have been a better choice for extreme overclocking. Still after lots of work, I finally got my 1090t stable to ~4275 on the first 3 cores with a 19.0 multi and ~4160 with an 18.5x multi on the last 3, and the voltage starts around 1.39v at idle and goes up to 1.43v load. I was able to leave on C1E and cool and quiet and just about all of the energy conserving features the board has everything has worked out just fine.

Question for you, do you ever find that the board sometimes takes a little long to get through the boot sequence?

Question for you, do you ever find that the board sometimes takes a little long to get through the boot sequence?

It definitely got longer as I added HDD's to it. It is much longer if a USB block storage device is attached at boot time. With 2 SSD's and 2 HDD's attached, I see a consistent 5 seconds or so to get to GRUB. Some of that time could be GRUB itself. Honestly, since the screen is blank for most of the POST process, I can't reliably divide that time between POSTing and GRUB's initialization.

madmanmarz wrote:

I have very good water cooling and airflow and still this thing likes to throttle like crazy and does not like high voltage.

One thing I noted is that AMD Overdrive and the Asus motherboard utilities are clearly showing different temperatures for the CPU core. Thermal throttling kicks in at 65C as reported by the asus software, which corresponds to about 45C on Overdrive. However, the offset is not constant. Either they are using different sensors entirely, or they are using different scale factors in the measurement equation. At least under Linux, there are two separate "CPU" temperature sensors visible, one reported directly through the CPU using the k10 driver, and another reported through the it87 driver.

A log entry is recorded by the BIOS every time the cpu throttles back to the x4 multiplier, and there is a configurable alarm threshold for temperature alarms, too. I suspect that changing the alarm threshold also affects throttling, but I was too timid to try it myself. Considering how easy it is to change the multiplier or clock frequency, I rather prefer to have an independent safety device to back me up.

just brew it! wrote:

I generally clear the CMOS on new boards right out of the box. I've seen boards come with weird (non-default) settings too, so it is the only way to be sure you're starting from a clean slate.

Ouch. I didn't realize that QC had slipped that far. My prior computer was a single-core socket 939 that got upgraded to a 4200+ almost 6 years ago, so its been maybe 8 years since my last total system build. I don't recall this being standard advice back then, but I will keep it in mind for the future. Thanks.

jbrandmeyer wrote:

madmanmarz wrote:

I have very good water cooling and airflow and still this thing likes to throttle like crazy and does not like high voltage.

One thing I noted is that AMD Overdrive and the Asus motherboard utilities are clearly showing different temperatures for the CPU core. Thermal throttling kicks in at 65C as reported by the asus software, which corresponds to about 45C on Overdrive. However, the offset is not constant. Either they are using different sensors entirely, or they are using different scale factors in the measurement equation. At least under Linux, there are two separate "CPU" temperature sensors visible, one reported directly through the CPU using the k10 driver, and another reported through the it87 driver.

A log entry is recorded by the BIOS every time the cpu throttles back to the x4 multiplier, and there is a configurable alarm threshold for temperature alarms, too. I suspect that changing the alarm threshold also affects throttling, but I was too timid to try it myself. Considering how easy it is to change the multiplier or clock frequency, I rather prefer to have an independent safety device to back me up.

Negative. I even tried disabling all temperature monitoring and it doesn't make a lick of difference.

madmanmarz wrote:

jbrandmeyer wrote:

madmanmarz wrote:

I have very good water cooling and airflow and still this thing likes to throttle like crazy and does not like high voltage.

One thing I noted is that AMD Overdrive and the Asus motherboard utilities are clearly showing different temperatures for the CPU core. Thermal throttling kicks in at 65C as reported by the asus software, which corresponds to about 45C on Overdrive. However, the offset is not constant. Either they are using different sensors entirely, or they are using different scale factors in the measurement equation. At least under Linux, there are two separate "CPU" temperature sensors visible, one reported directly through the CPU using the k10 driver, and another reported through the it87 driver.

A log entry is recorded by the BIOS every time the cpu throttles back to the x4 multiplier, and there is a configurable alarm threshold for temperature alarms, too. I suspect that changing the alarm threshold also affects throttling, but I was too timid to try it myself. Considering how easy it is to change the multiplier or clock frequency, I rather prefer to have an independent safety device to back me up.

Negative. I even tried disabling all temperature monitoring and it doesn't make a lick of difference.

PS It seems like it's throttling based upon the heat of the VRMs and not the CPU but honestly I have no idea. Regardless even 1.45v would result in throttling even with good water cooling and a portable a/c unit blowing directly into the front of the case.