CPU throttling under high GPU load?

Noticed some strange behavior in my T61p (see sig). Using Orthos beta cpu and ram stress test http://sp2004.fre3.com/beta/beta2.htm and just the gaming benchmark of Futuremark Vantage 1.0 http://www.futuremark.com/benchmarks/3dmarkvantage/download/, I discovered that when running the benchmark and CPU stress test together, the cpu freq immediately throttled back to 1.2 ghz (from 2.6, in both cores). Measured cpu freq with cpuz and some other gadgets just to be sure. This speed throttling was not related to temp - it happens as soon as I start Orthos, and as soon as I exit Orthos with Vantage still running, the freq returns to 2.6. When I run Orthos alone, the CPU freq stays pegged at 2.6 ghz, with 100% usage. HWMonitor shows nominal temps at the start of the run, when the cpu freq throttles back. It is on AC power, set to max performance in power manager. The only items that are not maxed out in this profile are the system temperature, and fan speed. I cannot find where to set those to the max in power manager (ver 2.41). Is it throttling back in anticipation of heat gain? What implication does this behavior have on gaming, when heavy loads are used on CPU and GPU? Is there some other reason to throttle the cpu speed?

 

Has anyone else noticed this on their T61p?

 

this is a bad way of testing since in order to test the GPU, CPU power is used as well, so you're not getting anywhere

anyways, it could be a power issue/profile to save power, so check into that

 

Good luck bsodder. My t61 ran like a champion so I can't tell you what I did.

Renee

 

Thanks for the response miro and Renee. I understand that the gpu processing is going to be slowed down by running the cpu test - but the idea was to stress them both. The cpu is the limiting factor, since it is pegged at 100% utilization, but then why would the cpu throttle back to 1.2 ghz. When the cpu throttles back, the core voltage drops to 1.01 volts, from 1.35 v. I have a T61p with the dual-dimm bsod fault in the motherboard, although mine has never exhibited the crashing that others were getting. According to Lenovo, the T61p crashes were a tolerance issue in the MB supply voltage. On some units, if a second memory chip was installed, the voltage dropped too low when the cpu and gpu were under full load, and the bsod occurred. When I noticed these results, I wondered if a subsequent BIOS release had been programmed to lower the cpu voltage when the gpu is under load, or when a voltage drop is detected, in the affected T61p motherboards - thus avoiding more motherboard replacements... Otherwise, it just makes no sense to throttle back the cpu when, as you said, the demand on the cpu is even greater, because the gpu is under load. I don't get it. I have confirmed my results with additional stress testing programs, however.

 

I thought that maybe this problem was caused by Power Manager, since it will not allow the user to set system temp and fan speed to max?? So I uninstalled TPM and tried the stress test again - same results. Windows PM was set to max performance, and I also set the minimum cpu speed to 100%. The BIOS settings are also on max performance. The cpu speed still throttled back to 1.2 ghz when I started the graphics benchmark, with cpu stress test running. It seems that Lenovo has crippled the T61p to avoid the motherboard under-volting problem. I can no longer run the cpu and gpu full on at the same time. I would love to know if others have this problem