Clevo Overclocker's Lounge

Weere you able to check what your max CPU voltage was during the bench ? @4.8Ghz.
Mine usually stays around 1.430-1.439 with firestrike and Cinebench.

 

Yeah, like 1.365V load voltage and 1.390V~1.410V at idle.

 

Holy **** is that for 4.8Ghz or 4.7Ghz ?

If that's 4.8Ghz, then WOW, you have one amazing chip!

 

Here

4.8GHz. Yes, it's definitely a little bit better than the one that came in the Sky X9 review unit. That was a good one, too. I think very similar to yours, in fact. Brother @Papusan has the best one I have seen, but this one seems pretty close.
__________________________
OK, here's a Cinebench at 4.7GHz on El Cazador. No AC, just fans. I had it apart to do some tweaking and it seems to be running cooler now. Pulled the IHS and started over from scratch with the paste. I've not done a lot of CPU benching because the little guy always ran so hot. Not sure how this compares to others with 4790K in the same little bruiser. Max temp was 85°C at 4.7GHz. On AC cooling it used to get hotter than that with the original screwed up heat sink. With the new one I am able to use liquid metal paste. Had to bend it a little more to improve fit while it was apart.

 

Some are really upset if the idle voltage is around 1.390/1.435V with processor nice OC'd, but forget that the load voltage can be 100mV lower than idle (OC'd 4.7/4.9 GHz) under load. 1.350 or 1.365V in load voltage is decent with 4.8 GHZ. This is even lower than a lot of 6820BGA processors <heavy> OC'd up to 4.2 GHz
Ps. Nice score from that little baby monster in Cinebench11.5
10.20 pts is only about 0.15 from highest score in Hwbot with same Clockspeed That Baby rock!!

Edit. This bench is from Auros with 980

 

very nice CB scores fox! ive only gone up to 9.40 at 4.8 ghz, i guess the rest is due to windows streamlining

edit: thx to @Papusan for pointing out my error. corrected in a later post

also, how exactly did u "bend" the heatsink into shape on the cpu side? did u remove the plate and hammer it down before resoldering it to the rest? or something else entirely?

Sent from my Huawei Mate 8 NXT-AL10

 

You should have more than 9.40 regardless of streamlining Os with 4.8 GHz. 9.40 is more like 4.3 GHz with Hotwell.

 

worth OC'ing the CPU?

 

Not in gaming. But bench and work with need of processing power, Yes.

 

ok, will stay at stock for everything...not a big deal.

 

You have a unlocked processor, Prema bios and one of the most OC friendly laptop ever created. I would push the processor a bit. See how it goes. But not much over 4.4/4.5 GHz if you haven't delid cpu or use Liquid ultra. Of course if your heatsink is ok and not uneven/warped.

 

like 4.2 or 4.3 if it can do it? what's the best tutorial for this?

 

You have a guide for different Oc in the first post in the thread. Also the noob thread from @Phoenix http://forum.notebookreview.com/threads/clevo-p870dm-n00bies-guide.789463/
Remember to follow max temp in your first bench with Oc. And Hwinfo is the best tool for that. + don't use XTU. Will cripple your setup if you put in profiles.

 

yes youre right, my bad! checked my scores and it was actually 10.26 at 4.7 Ghz thx for reminding me of my own scores haha, mustve been confused with CB15, where my highscore was 940

HA! so still staying ahead of brother fox, even though by a small margin

 

Yeah, I thought your bench numbers was very very wrong I think 4.7 GHz is good for +- 10.30
And 4.8 GHz should go a bit higher

 

i have it clocked at 4.2 with @Phoenix settings. what's a good way to test stability?

 

Bench with the benchmark tests showed in first page. Wprime 1024M stress test is fine. Remember to follow temp first time.

 

tried loading wprime but it gave an error. testing with AID64 and max temp so far is 76 celsius. what's a good test length?

 

5-10 min is ok. Did you run Wprime as adm? Also Cinbench R11.5/R15 is a very nice bench. + 3DM11.

 

cool. i am running wprime now while downloading cinebench and typing here

 

just remember: the more stuff ure doing while running the benchmark, the lower the scores

 

Anyway. Follow temp after your first OC Then push all the way without any thrash in the background Every point count. And repeat bench!!

 

so far 4.2 seems to be good...and good enough so far.

 

And the bench results is correct for your clock speed? +Temp is ok.

 

with fans at max it's below 80 celsius in some tests i have tried. 3DMark was good as with cinebench - that was 273.xx seconds for 4.2GHz.

 

?? 273.xx sec in Wprime 1024. What version?
Cinebench 11.5 should be around 10.25

Edit. You can compare your results here http://hwbot.org/hardware/processor/core_i7_6700k/

 

Did you select 8 Threads? Seems that you did stress testing it with only one Thread.

 

But undervolt a little bit if you using stock speeds. That will drop temps quite a bit.

 

Or do the opposite . Increase multi. Stock voltage can be used up to 4.5/4.6 GHz. Test and see what happens.

@CaerCadarn Even with only 1 thread isn't 273.xx the correct score with Wprime 1024M version 1.55 or 2.10.
273.xx is a abnormal score.

 

Cinebench 11.41 @ 4.8Ghz

 

I have attached a couple of photos of the original P750ZM heat sink. It might be easier to show what I was dealing with than try to thoroughly explain it. Clevo should be ashamed of producing such crappy heat sinks. Hopefully, they will give this greater attention in the future. When they fit right they are really good. When they do not they are junk. They are very inconsistent in their QC.

As you can see in the first photo, the heat plate for the CPU is not even close to being flat. I was able to bend that plate with my bare hands. I flattened it out, improved the fit and added a 30mm x 30mm copper shim to improve pressure. It helped, but it was not enough. You can see the shim I added in the second photo.

The second photo is also the old heat sink, but the arrows show the direction I had to bend the new heat sink that was purchased. The radiators were too close together to fit into the pockets in the chassis and the GPU spring tension arms did not line up with the screw holes in the GPU support plate. Again, I was able to do this with my bare hands, spreading the two halves of the heat sink away from each other. I had to do it carefully, so as not to destroy the heat sink.

 

Now put it on top of the cpu....

 

And tighten the screws in the correct order

 

Was talking about the ruler.

 

We know the result already

 

If the heat sink were made correctly it would fit the top of the IHS. The original was so deformed that it did not make contact with the IHS except very minor contact on one outer edge of the IHS. Before adding the shim I could actually lay a narrow strip of paper on top of the CPU in the middle or the edge of the IHS closest to the GPU, bolt down the heat sink and slide the paper out, LOL. So, QC is clearly pathetic. The threaded stand-offs on the CPU support plate that the CPU heat sinks screws go into may also be a bit too tall for a snug fit. New heat sink is working pretty good now. It was in a bind with the CPU and GPU pulling things opposite directions and the screw holes not lining up.

 

Correct Fox. Then I can put in this http://forum.notebookreview.com/thr...enix-has-arisen.781814/page-727#post-10298036
I don't like the design and Qc have to be better

 

Ignore what I said before for now. I will retest PCIE x16/x8 with SLI once I get a second flex bridge.

 

Just a note perfectly flat is not ideal for chips with a heatspreader

 

will have to read @Phoenix's settings cos i don't know where the voltage is, just the multiplier.

 

Also read @Mr. Fox exelent review https://biosmods.wordpress.com/2015/10/21/sky-x9/5/

 

Right, because the IHS is not flat. But, no contact or poor contact is very bad no matter how you look at it. Improving that that is critical, even if the final fit is not 100% perfect. Some of the messed up heat sinks I have seen are not even "almost" good enough. And, by the mixed results reported by several it seems to be hit or miss. I've been very lucky with both P870DM-G machines, but it took a second try on the Panther and the second try on the P750ZM still wasn't very good without tweaks to get it closer. I also see reports of CPU thermal challenges with older machines like those owned by @Ethrem and @D2 Ultima so it's not something Clevo just suddenly developed an opportunity to improve upon.

@Phoenix

 

Added although I still beat you on the Write / Copy memory benchmark but you beat me overall.

 

No need to change the leaderboard. I thought you would enjoy seeing the results. That's a really hard one to make a call on with so many different metrics involved. Might be good to break it up into several different categories if you can figure out a good way to do that. Maybe something like Read, Write, Copy, Latency categories. Too bad AIDA64 doesn't assign a "score" for a calculated overall result. That would make it easy. MaxxMem2 does that, but it does not function correctly for DDR4.

 

it's cool, I will run a few benches later and defeat you brah

The main focus here is RAM speed and not cache speed but I do take that into consideration on the overall score.

The reason I added your score is you beat me an way too many metrics to ignore your result.

 

interesting, thanks for the explanation @Mr. Fox, definitely food for thought

 

That's not the premamod but the newer type of stock BIOS, they finally added some options, but it's just the same stuff you also see in XTU, hence you can't reach your old OCs.
But no worries, for more serious tweaking and no throttle etc just flash the Mod I made for you earlier and all will be good again.


EDIT: So Futuremark added TimeSpy plugin (paid ) while I was gone...meh...was more hoping to finally see VR Mark by now, but still nothing:

http://store.steampowered.com/app/464170

EDIT2: Is an OS run via USB 3 a bad idea for 3DMark benching? Running out of drives here...

EDIT3 (yeah I know ) : So Microsoft auto-updated my 10 year old Clevo M570 to W10 without even telling me that they have no GPU driver for the Quadro 2500M...took me half a day to create a Franken-Driver that worked.