Need help with testing Scalability of Intel chips.

Hey I got a bunch of questions because I have been “thinking” again. I would like to get some info on some topics I am curious about. I have been typing them into a word file as they come up since I keep forgetting them before I get a chance to post ^^…blasted memory -_- So here we go and I got links to the other topics below.


2.) Has anyone or can anyone try a test scaling a current Intel CPU? (IB) As in a quad core and a ULV? If I tried to test the scaling of my 3720qm how do you think scaling would work the best? Try locking to one, two, or three cores? Clock mod? Chip mod? Freqs? Multiplier mod? Also I thought in jay’s thread that Haswell has this new ability to control voltage per core or something along those lines. How does that affect scaling of a processor and how does that compare between IB and Haswell (HW? or HS?)

I am asking this because I am curious on how well a quad core can scale down to a lower TDP and how an ULV can scale down. Do they scale efficiently or no. How does the performance per watt compare and so on. It has been proven that higher the frequency the more inefficient it is. So how well does that work for quad core and dual core. Does a dual core scale better because it has a lower overhead due to a smaller die? If so by how much? I have been wanting to try this for a while now. I have a quad core but no ULV IB so would anyone want to try this with me and try to scale a ULV at 5w 10w 15w and 17w (obviously that isn’t scaled)? I would try 10w 15w 20w 25w and so on. I know the lowest my 3720qm will idle is 7-8w with it throttled at 37.5% clock and chip mod. At 25% it nearly freezes. I know uncle webb and others say intel chips will turn off cores and they use almost no power so why does a chip idle run at 8w at best even though much isn’t used?

There is a very good reason for this question and a practical application for this but I don’t want to get into details for certain reasons. If you think it is stupid than please don’t troll and go post elsewhere. There is a practical application for this idea.

As always I appreciate the advice!

Questions and URLS
1.) Curious on data about computer and electronic sales per year
http://forum.notebookreview.com/off...er-electronic-sales-per-year.html#post9014927
2.) Need help with testing Scalability of Intel chips.
http://forum.notebookreview.com/har...ting-scalability-intel-chips.html#post9014929
3.) Another RISC/CISC question need help
http://forum.notebookreview.com/har...risc-cisc-question-need-help.html#post9014928
4.) Need help with some virtual machine questions
http://forum.notebookreview.com/win...me-virtual-machine-questions.html#post9014930
5.) Thin film, li-polymer, and nanowire battery question.
http://forum.notebookreview.com/har...er-nanowire-battery-question.html#post9014931

 

When a core goes into a lower power state, it has to be "woken up" to be used again. The lower the power state it is put into, the more clock cycles must be used before it wakes up. Your cpu consuming 7-8 watts at idle (which to me actually sounds pretty high) is the balance between putting cores into lower power states and waking them up again for whatever the OS needs during idle usage.

 

well i am glad to see if anyone else with a 3720qm can post their numbers.


I have just turned off F@H and given it time to rest and find equilibrium and with only standard programs installed i am bouncing from 1-5% while typing this and i am getting 13-15ws right now. I only hit 8.5-9.6w with 37.5% clock and chipset mod. It also runs at 19-35% utilization.


At 25% everything is super slow and i am using 45-60% CPU with 10 FID and 25% clock and chip mod, which runs at 8-8.5w.


Side note i just tried chipset clock mod alone at 25% and i am using 18-22w so that is even worse. Runs better with less waste with nothing done. Why does chipset mod cause more power to be used?