After finally managing to get a stable 4G overclock on all 4 cores on my 4910mq for 24/7 use I checked hwbot and found I had the world record there for a score of 1046 for my 4910mq @ 4G. Digging deeper I see this score is about the same as 4770k overclocked to 4.4G.
What scores and clocks are other people managing to get for their mobile Haswell cpus? I'm surprised I can outscore desktop cpus with their higher ocs.
-
Meaker@Sager Company Representative
The 4770k scores you are comparing to are not performing correctly. Still a nice score
-
Submitted 2 more scores, 4.1 @ 1.15 vcore 1063 points, 4.2 @ 1.18 vcore 1083 points. 4.3 was experiencing current throttling @ 140 amps so I cancelled out for the time being. Max power draw I saw was 99W.
Definitely not going to be using these 24/7, how do other ppl run their cpus at 4.2G? Do they not throttle/overheat under load? -
Meaker@Sager Company Representative
What system are you using?
-
-
Meaker@Sager Company Representative
Typically 3.9-4.0ghz for 24/7 but it depends on the quality of the chip and cooling system tolerances.
-
What vcores are you seeing at 4G? I went with 4910mq instead of 4940xm as I'd heard temps would prevent enjoying the 4940xm unlimited multis which seems to be the case here.
However I wonder if I just got a bad quality 4910mq chip for oc? -
Meaker@Sager Company Representative
No you seem bang on for what I would expect.
-
Thanks, these Clevos don't seem to have much thermal headroom for overclocking the cpu.
Oh well, more modding needed -
Also, the CPU cooling is indeed downgraded from the EM series, but it was necessary for 780M/880Ms. It's not necessary NOW however, so I keep telling people about it. XD. -
Problem is the parts seem to vary in effectiveness even between the same model of Clevo. Eg. Ethrem had his NP9377 replaced for high temps, it returned with super low temps with just ic diamond pasting.
I haven't started to look at gpu temps under load yet so don't want to disturb the existing gpu heatsinks until I get a baseline of temps under stock n oc. 10 degrees increase may not affect the gpu if it never goes above 80 fully loaded while oc. -
Meaker@Sager Company Representative
Temperatures will vary from machine to machine due to chip quality and exact heatsink dimentions, the CPU cooling is perhaps the weaker part of the unit, you can get 3.9-4ghz to run all day long though and that's going to get you a LONG way, even with the next generation cards (if they work).
-
Thx Meaker. Is there any way to measure cpu fan speeds? I'm convinced the cpu fan on auto either has an additional increase after 95 degrees or it's the highest speed being delayed since after throttling for 45 secs it revs up 1 notch and temps drop below throttling.
I'm trying to work out which it is as it's quieter than max yet louder than anything below 95 degrees. If you know what the temps and stock fan speeds are, evwn better. -
-
If that's the case then the fan behavior on my machine is the highest speed doesn't kick in until about 45 secs once throttling starts. Maybe delayed? Or once temps trip 95 degrees.
Need some way to verify this, only difference from max fans is that only the cpu fan speeds up and not the gpu fans.
Once xtu stressing stops it seems to turn off the cpu fan for 5 secs before engaging the fans at a slower speed. -
u can check the fan rpms by using HWInfo
it gets that info off the machine´s embedded controller (EC)
-
-
-
-
hm, too bad, back in the day with my HM series clevo i was able to read the rpms in HWInfo...
-
Suggestion: ask people in room to get used to it. -
Meaker@Sager Company Representative
You could make the fan profile a bit more passive at idle but at load a certain degree of cooling is required.
-
Only louder machine is my brother's AW M18x R2 when he starts gaming.
-
Meaker@Sager Company Representative
Well that's a similarly high end machine so no surprise. You could disable sli and run with a single card to take a fan out of the equation I suppose.
Haswell overclocking and XTU scoring
Discussion in 'Sager and Clevo' started by kantana, Jan 12, 2015.