Clevo Overclocker's Lounge

The more the cores the merrier. Specially for CPU specific benches. And if you can push the clocks, you are in for a treat.

 

Check out my 4790K score on there. It is compared to the top, which is LN2. It doesn't show an LN2 for your hexacore, meaning most likely no one took the time to LN2 or DICE it for that specific test (I smell opportunity)! Meanwhile, the top CPU frequency is LN2 for your CPU model. Just trying to inform you. Meanwhile, if you'd like, check out me on HWBot (same handle and avatar)!

 

@ajc9988 , I see what you mean. Sorry for making confusion/conflict.

Also though there is a ranking list for water air and Ln, for specific CPU's, that is what I was talking about.

 

I think there is some misunderstanding as to how over all benchmarks are calculated at the bot.
It goes from the very best of each category to the lowest. And every single category you are up against anyone that can use ln2/dice/ss/lne/water/chilled water and so on.
Those other links are only for comparison or try to gage you against other water or air systems, but that is not 100 percent right because people will get lazy and just post whatever their last benchmark settings had. to speed up submitting multiple scores. This is why a laptops running desktop cpus will never place in the top 10 unless it's a no points category where no one cares.






@bloodhawk I might be inclined to have a look at your machine to see why you are having firestrike issues

 

@johnksss , That is amazing scores man. I am in complete envy and awe right now.

 

I'm back to benching with the 365.19 drivers, which seem to work very well with a 1.4 GHz clock. Now to tweak the memory and CPU clocks... @Prema has some tricks, but he ain't revealing them.

 

@Takaezo Thanks.

 

What were you using? 362.00?

 

I think so, yes.

 

Sure! How do you want to go about it?

 

Let me know which works best for OC. Try to keep the two (core and ram) as close together as possible. Go up until you get a driver crash, then up the voltage one notch. Continue until you get the black screen (too much voltage). Then increase the ram while holding either the core at a little less frequency or last safe voltage. After you find the higher ram, see if the core can now handle a little more frequency or voltage. At this point, you are maxed out. One or the other higher (core or ram) doesn't produce as good of results unless it can go much higher than the other, then it is fine. Just some advice you probably already know...

Sent from my SM-G900P using Tapatalk

 

The thing is, the stock voltage on my GPU is already 1.1 V, and I literally only have 0.1 V headroom. My GPU is running at 1.2 V already.

 

How's your cooling? I understand the concern for voltage damage or blowing a capacitor. Does it black screen? If you are ever concerned about going higher on voltage, don't! Stay within your comfort level or what you know to be safe for your hardware and cooling capacity. None the less, if you are already at the highest voltage at which you are comfortable, stay there and test upwards until highest stable.

Sent from my SM-G900P using Tapatalk

 

Teamviewer 10 since i have paid version and no nag.

 

I got some stuff to do today and tomorrow. Will Saturday work for you? Ill PM you about the details.

 

I should be around. Just hit me up.

 

Cooling is excellent - my GPU doesn't exceed 75 °C, and no, my notebook has never black-screened. I do get driver crashes when I benchmark at such high clock frequencies, though. At any rate, 1.2 V is the hardware limit of the voltage on my GPU - I cannot go any higher even if I want to.

 

Then you are voltage limited. Unless you modify the firmware (or pay someone to help you), you are limited. You would need to have higher voltage or back off your OC until there are no more crashes of the driver and you will have your stable oc. If I remember correctly, the voltage settings don't effect ram voltage on Nvidia Maxwell cards. As I'm still relatively new to video overclocking, I'd want a more experienced user here to confirm. But it is my understanding that this allows for you to run up the ram without an issue, until crashing, artifacts, etc. So back then both off until you are confident in no driver crashes. Then increase the ram while testing and looking for artifacts or errors. Once you have that down, you should have your stable overclock on your card! Hope that helps!

Sent from my SM-G900P using Tapatalk

 

New improvement by BGA-KILLER in Cinebench R15 - 1065cb @Phoenix Moved up 4 places on Hwbot - Cinebench R15 Ranking.
http://hwbot.org/submission/3218735_papusan_cinebench___r15_core_i7_6700k_1065_cb?recalculate=true


Another Cinebench R15 - 1062cb from today!

 

Sometimes though rare it can be done. For example I once had a GFP with a mobile dual core Haswell CPU on air over all LN2 desktop CPU's until Skylake was released and put an end to it.

You know it surprises me for years now that people talk about maximum voltage but very seldom current. Typically with silicon too much voltage will kill a chip instantly while too much current can be a little more subtle from snowballing to gradual degradation.

 

Got any links to that?

 

If you call pulling a 3 year old SSD from a W230ST & plugging it straight into a W230SS and continue to use the same old OS installation, which already has been migrated from 7 to 8, 8.1 & 10 (without even manually updating any driver except for GPU) and having 1GB space left, a secret then, well, it's out now.

 

Did you ever get your p170hm3 up and running?

Sent from my SM-G900P using Tapatalk

 

I am traveling and won't be back in its physical proximity for another 2 months...

 

Btw, loved that OC a month or so ago on the locked Haswell! Does that work for Haswell-EP micro codes as well?

Meanwhile, as always, wishing you safe travels wayfairer!

Sent from my SM-G900P using Tapatalk

 

I followed the video guide on page 1 for 4.2ghz overclocking in BIOS (I own a p750dm-g with Primamod Bios) and once I boot into windows and play any sound from laptop speakers, I get a crackling noise come out at the same time.. Even after multiple resets, it still has that sound; the only way to fix it is to revert back to default bios settings.

This tells me there must be a particular setting from the video that when enabled/disabled, it causes the speakers to make the crackling sound.

Until this is resolved, I'm using intel XTU to keep cpu at 4.2ghz with -140mV... however I would prefer for the BIOS to control the overclock and not XTU..

 

ok so now it's you and me, what I noticed is that messing with the cache overclock caused that crackling.

I need to recheck the settings but for now, just set the ring OC Ratio min/max to 0 (which means default multiplier) and also make the cache voltage offset the same as the CPU voltage offset.

Tell me if that fixes the issue, if not, then we will have to do some more tweaking in the settings to find out what's causing this.

 

Thanks Phoenix! That seems to have fixed it, I just changed the value from 42 to 0 - the rest of the settings are exactly the same as the photos.
I'll keep testing, but it seems to have gone, no more crackling!

 

and here is the detailed post about this, it sucks that we can't touch the cache ratio but oh well, its performance benefits are negligible anyway and just causes extra unnecessary heat:

Crackling noise from speakers

 

Thanks Sir.

 

Need to cool that puppy down! ;-)

Sent from my SM-G900P using Tapatalk

 

Not really, since it's not a 24/7 thing.

 

It was in jest...

Sent from my SM-G900P using Tapatalk

 

I figured. All is good.

 

High numbers is more important

 

If you mean the OC then
http://hwbot.org/news/12173_0_0_bea...with_air_cooled_mobile_haswell_in_2x_cpu_xtu/
http://hwbot.org/news/12230_interview_with_0_0_overclocking_locked_mobile_cpus
https://gaming.msi.com/article/i7-4600m-overclocking

Although the DDR3L RAM can clock up to 2800MT/s, being single rank probably wasn't doing me any favors.

If you mean the voltage and current then I don't have any links to hand but googling electromigration and junction breakdown should help.

 

Question: is anyone else here finding their Clevo's lacking on certain benchmarks, such as GPUPI for CPU runs? I ran one on the P170HM last night with the AC full blast, outperformed in other categories, but 6 seconds slower at 100M. My 2960XM was comfortable at 4.4 (could've gone higher, just was out to steal some points, low hanging fruit sort of thing). Even with taking 5 categories, including the cinebenches, geekbench3 single and multi, and hwbot prime, I couldn't beat a chip clocked at 3.7GHz with slower ram. My throttle down was still over 4.0GHz. This means the issue is in the OS (which I have optimized fairly well), the driver (openCL drivers, some are better than others for different intel models), firmware (I've had this with multiple firmware and have different BIOS I use, depending on longer or shorter runs as one I have works better for wPrime1024, SuperPi32M, and GPUPI for CPU 1B) while the other shaves off a little from the shorter tests (wPrime32, PiFast, SuperPi1M, GPUPI for CPU 100M). Any advice I could get on optimizing these further than I have would be greatly appreciated, because a 3.7GHz on a dell m6600 should not be beating a P170HM at 4.4GHz by 6 full seconds. @johnksss @Mr. Fox

BTW, johnksss - no way am I shooting for your cups on that platform yet, just wanted to up my cups for the novice level because I haven't submitted much since Dec. or Jan.

 

Alrighty boiyos, fixed my throttling issues. BIG THANKS TO @unclewebb !! And im sure most of you guys are gonna love TS 8.20 once its ready!


Look at the Physics scores. that's literally a 400-500 Point jump from my previous runs @ 4.8Ghz.
Cant wait to get the new memory and then im coming for you @Papusan .
http://www.3dmark.com/fs/8539359
These runs were to mainly test out the processor OC, didnt even bother to take off the bottom cover and only had the fans at max speed without the cooling pad.

 

Atleast on the P870Dm the throttle issue was there because of some bug in the EC, and @Prema 's BIOS fixed that.

 

Personally I always keep my benchmarks with the system still able to unplug and be mobile otherwise I feel it kind of defeats the point

Still, an interesting test to run.

 

Haha true that. Im actually going to mod my U3 and then put the system without the bottom on it, so that i can run higher clocks for rendering. The only time i have been mobile with this system was last week.

But i get your point, if i ever had to take it out of the house, it would be a pain to put the bottom on and screw things in.

 

I have the Prema BIOS and still have the CPU throttling issue when the GPU is active like in Fire Strike or games.

I even tried Throttle Stop but it didn't help, how did you fix it bro

 

This is the normal speed burst down clock on the Sandy bridge mobile series. It will burst to 44, then go down to 41. I've seen some go from 42-44 down to 37. It's not the heat, but the architecture. Need to also play with some other settings. But this being the case, that should out perform the same processor at 37. Plus, I'm trying to figure out why I under perform in that category generally. I noticed the new opencl driver from Intel released last month... Hmmm... Might open throttle stop back up to figure it out...

Sent from my SM-G900P using Tapatalk

 

What clock speeds are you running the benches at? Might be a bunch of things, took me a week to nail this down lol.
Also try giving the latest drivers a shot, my GPU and CPU temps have dropped by 3-4C. These runs above were with the latest ones. One thing that you will notice though is that the GPU Power consumption will no longer be visible in HWInfo.

That is weird. For our Skylane CPU's it would clock down to 4Ghz if the voltages were ridiculously high. Maybe the something similar?

 

it happens on any overclock, be it 4.2 on all cores, 4,4, 4.5, or 4.6

it dips to 4 GHz for a fraction of a second, so my average clock speed during the benchmark is like 4597 MHz. if I was @ 4.6 GHz but it ticks me off why is it dipping to 4 GHz. I tried with C states on or off, doesn't make a difference and neither does Throttle Stop

 

C-States might not be the complete issue.
I think i had this as well, so i basically went ahead and setup base stable settings (for me 4.6Ghz @ -75 Offset and ring 4.3Ghz @ -150 Offset) in the BIOS. But i set the multiplier under the Overclocking Feature menu to 55 and then i locked it to 46 under the CPU Information page. After which i control everything using TS in Windows.
This helps me in finding the exact voltage which the processor is stable at without throttling, usually its because of marginally excess voltage. As mentioned by the veterans here - Skylake is sensitive to voltage.
Only thing is that 4.5-4.6Ghz isn't that high of a frequency for the system to feel the Voltage is too high, unless you are going apeshit crazy. For example my 4.8Ghz holds @ 1.420V for @Papusan i think it holds around the same @ 1.425V (+/-) 5-10 mV and these voltages the system seems to be fine.
But if i take my voltage anything over 1.445V for 4.8Ghz it starts freaking out. However for 5Ghz @ 1.475V it works like a charm. So im not entirely sure at whats going on.

You will have to fiddle around a bit to figure out the exact voltage or offset where it gold the frequency steady.

Also keep in mind the current clocks always go 3-4Mhz up and down so sometimes the average does get affected by it. Last but not the least, make sure the Max / Min processors states in the power settings are set to 100%.

 

Not weird for SB mobile. This is when they were playing with turbo and it is time limited burst as well as tdp limited burst. And yes, I'm using a 2011 year platform, so things have changed significantly in how they do the turbo. With that said, a 4.1 should outperform a 3.7 in almost all use cases. I listed the main differences: MB, firmware, drivers, ram (even though mine was faster by 266mhz), OS optimizations, etc. I spanked his cinebench 15 by 160-200 points! Decimated his scores on geekbench. Meanwhile, I get my ass handed to me on gpupi?!? The m6600 was the business Dell. This would have compared, to a degree, to the alienware mb on certain components and firmware. His score compared to a 4770K stock 3.5ghz score in gpupi (the sample image on the rules page). For comparison, my 4790K 5.1ghz score on wprime32 compare with johnksss's 5.1GHz scores with his alienware and 2960K. That means it very well is possible, but is maddening...

Sent from my SM-G900P using Tapatalk

 

Firestrike. Dang..., Missing only 8 points for a nice clean 12000P in Firestrike with stock Gpu Maybe break 12000P will be the normal with new/better drivers from Ngreedia?
http://www.3dmark.com/fs/8540223


@bloodhawk 1.475v isn't much. You can go up to 1.5V on Skylake You said GPU Power consumption will no longer be visible in HWInfo... I can see this in Hwinfo with latest driver.

 

Weird for me its like this :


If i go back to 362.00 it comes back right away.

 

Weird... 365.19