Hunky Dory
Petty Officer 2nd Class
- Joined
- Jul 20, 2003
- Messages
- 102
So I had two old laptops of very similar capability sitting side by side. A primary difference was that o'Dell had PC2100 memory and toshibi had PC2700, both 512MB. Both had Pentium 4 HT processors; o'Dell at 2.8 and toshibi at 3.0. Both had XP Pro SP3. Both were updated. Both were stable. I wondered what would happen if I raced the two. I don't know why...
Don't get too caught up in the configurations because there is a more basic puzzlement to follow. Suffice toshibi is made of slightly faster stuff.
So I went to calculator set it to scientific view and set both machines to find the factorial (n!) for 50,000. I ran procmon, CPU-Z and perfmon. I quiesced everything til the system idle process got 99%. CPU-z said that o'Dell was coasting at some 1500 mhz while toshibi never clocked less than 2990. Everything looked ready.
I poised the mouse pointers over the n! button and simultaneously left-clicked the trackpad buttons. Both fans went to loud. Perfmon showed CPU utilization @ a constant 50% with tiny bumps for running the monitors -- both machines. Process Explorer or taskmgr showed both threads in use at ~50%. Seems normal for a single thread application. CPU-Z showed both clocks running at or very near rated speed (o'Dell had jumped up to 2.8mhz). Perfmon showed HD activity at or near zero. Same for paging and network traffic.
The awkward surprise was that o'Dell consistently outran toshibi. In fact, toshibi seemed to slow during subsequent tests. Aha, reboot required! So I did. Same story. o'Dell needed about 20 seconds, Toshibi no less than 27 (I ignored the "long time" dialog, it closes itself when the calculation completes). They arrived at identical results. There were no errors, crashes or BSODs.
The point is that everything looked the same and everything looked normal. But it wasn't.
I have resolved the problem, all is right with the world; Toshibi now beats o'Dell every time (by a half second or so). So here is the puzzlement. I changed nothing but I did something. What was it? And through what mechanism was toshibi previously slowed?
In other words, even though I know what fixed the problem I don't know why. Do you?
Don't get too caught up in the configurations because there is a more basic puzzlement to follow. Suffice toshibi is made of slightly faster stuff.
So I went to calculator set it to scientific view and set both machines to find the factorial (n!) for 50,000. I ran procmon, CPU-Z and perfmon. I quiesced everything til the system idle process got 99%. CPU-z said that o'Dell was coasting at some 1500 mhz while toshibi never clocked less than 2990. Everything looked ready.
I poised the mouse pointers over the n! button and simultaneously left-clicked the trackpad buttons. Both fans went to loud. Perfmon showed CPU utilization @ a constant 50% with tiny bumps for running the monitors -- both machines. Process Explorer or taskmgr showed both threads in use at ~50%. Seems normal for a single thread application. CPU-Z showed both clocks running at or very near rated speed (o'Dell had jumped up to 2.8mhz). Perfmon showed HD activity at or near zero. Same for paging and network traffic.
The awkward surprise was that o'Dell consistently outran toshibi. In fact, toshibi seemed to slow during subsequent tests. Aha, reboot required! So I did. Same story. o'Dell needed about 20 seconds, Toshibi no less than 27 (I ignored the "long time" dialog, it closes itself when the calculation completes). They arrived at identical results. There were no errors, crashes or BSODs.
The point is that everything looked the same and everything looked normal. But it wasn't.
I have resolved the problem, all is right with the world; Toshibi now beats o'Dell every time (by a half second or so). So here is the puzzlement. I changed nothing but I did something. What was it? And through what mechanism was toshibi previously slowed?
In other words, even though I know what fixed the problem I don't know why. Do you?
