Clevo Overclocker's Lounge

giving your 4.4GHz CPU OC settings a try @Phoenix.

EDIT: I let it ran wPrime95 1024M and it came back at 255.xx so about 18 seconds faster than at 4.2GHz. AIDA64 stability test ran for 30mins and took about 20-21 minutes to peak at 92 Celsius with fans at max.

 

I still can't understand your Wprime 1024M score...!! Can you please show a Screenshots of your bench? Run 1024M one more time and show us.
Download and run Cinebench 11.5 http://downloads.guru3d.com/downloadget.php?id=2475&file=1&evp=841d9cacb0e0de71ffd9de0248868ad9

 

gimme a few mins

EDIT: Turns out the wPrime was only set to use 4 threads (effectively half the CPU) so I stick it to 8 threads and got this score I got for 4.4GHz...



Below is the best Cinebench R11.5 score.

 

Yeah, This is more correct Rep added.

 

Probably just using a single or wrong number of threads.

 

Yeah, He used wrong number of threads. But it couldn't be only 1 thread, because 1 thread can't match his first score.

 

First time playing with Time Spy:

http://www.3dmark.com/spy/145236

 

Hello everyone. I bought a couple of 512 Gb Samsung SM 951 on eB... and created a Raid 0

array for my new EVOC P870 powered by Prema Bios.


Here are a couple of Benchs i find quite decent, for a stock installation...


For the total price ( 440 $ here, in France ) it cost me less than a single 2,5" Samsung 850 Pro

1TB or a couple of 512 Gb ( 2,5" or M2 AHCi either )... A theorical 2x to 4x speed gain, even if i

don't really see it in my daily use ( except for large data files transfer ), it is worth the price, i

think !


The next OEM generation -SM 961- seems more powerful yet ( 1.5 x faster ) but it can't be

raided !

What is your opinion ?

By the way, any idea to improve these scores ?

Thanks

 

FYI: There's some good information about Time Spy, A-Sync compute etc on PCPER.

Video:
Article: http://www.pcper.com/reviews/Graphics-Cards/Whats-Asynchronous-Compute-3DMark-Time-Spy-Controversy
Article2: http://www.pcper.com/reviews/Graphi...Looking-DX12-Asynchronous-Compute-Performance

Particularly this image from Article2: http://www.pcper.com/files/imagecache/article_max_width/review/2016-07-14/daigram.png

Futuremark's statement: http://www.futuremark.com/pressreleases/a-closer-look-at-asynchronous-compute-in-3dmark-time-spy

Hopefully it puts the "this is a bad benchmark" stuff going on here.

TL;DR

1. The load generated by Time Spy absolutely eclipses what Firestrike did. Literally 10x more vertices being drawn and 2-3x the triangles. More importantly, compute shader calls are off the planet in comparison.
2. Your GPU OC that may have function in Firestrike is probably not as stable as you think for the above reason. Do not be surprised if you have to go for a more conservative clock.
3. 3DMark has and always will be a gaming benchmark focused on GPU performance. CPU scores are heavily skewed. CPU speed does very little to the "Total Score" in TimeSpy as a result and for very good reason. This reduces reliance on the CPU for the sake of comparison, making Total Scores more relevant between systems, especially between reviewers.
4. Maxwell and prior chips do NOT have ASync Compute support at all. Only AMD and NV Pascal architectures see a benefit from it.

 

The one place to discuss this has been closed off to any Posts other than Benchmark results, but maybe @Ethrem will approve this discussion in a sub-thread of it:

http://forum.notebookreview.com/threads/3dmark-time-spy-benchmark.793889/page-2#post-10300120

 

Problem isnt that Time Spy is a bad benchmark. But how its Pascal favored and has ended up being released just as the latest cards came out, as if its a combined marketing tool. And most parts of Async compute dont seem hardware based.

 

Hoping since this is specifically the Clevo OC thread it'll help more. Most people here are running unlocked vBIOS/BIOS systems where knowing about these limitations is a real issue and perhaps not as relevant in a general thread. Particularly since it's focused on GPU performance and most other systems run the usual gimped Maxwell BIOS with +135mhz limit in play.

 

Well, in all honesty, unless you work with a lot of small files all day every day, RAID-0 isn't worth it. Heck i work with files like that and i didnt find RAID worth the head ache. But thats just me.

 

It's Pascal favoured because Pascal has different hardware which can actually take advantage of ASync Compute. Same with AMD. Maxwell cannot execute multiple compute queues without serious driver-level emulation and stuffing around which no doubt creates more problems than benefits.

This is partly the case. If you read said articles you'll see that.

A-Sync Compute requires hardware and software changes. Much like multi-core CPUs and 64bit instruction sets did absolutely nothing back in the day without the relevant software to utilize them. Hell, with 4+ cpu cores these days we're STILL struggling to find things to do with all of them (from a gaming perspective). This same logic and relevant problems apply almost directly to A-Sync Compute and it's implementation as well.
It's not some magic bullet that you simply flick a switch with and see huge gains.

It's literally what the name implies. Multiple compute threads on the GPU. If you don't fill them with tasks, then it's literally useless. Current games and benches have NO way of doing this because they're literally not programmed to use more than one compute queue. Once people actually find more things to do with multiple compute queues, then you'll see pre-Pascal get left FAR behind. 3DMark Time Spy, which is a contrived test already makes very little use of multiple queues as evidenced by them. At most it uses 2 queues and the 2nd queue barely has anything in it anyway. It's not for a lack of trying either, there's just not much else to put in there currently.

 

Even so, im seriously skeptical about the whole Time Spy hallelujah Pascal train. Maybe boo boo on my part, but Pascal just has been hyped way too freaking much and i just refuse to buy into it.

 

The "Pascal hype" has been mostly about performance/watt, which is quite simply impressive. This is a fact and does not require your "buying into it". We have benchmarks now with power consumption recorded so hype has nothing to do with it now. That being said we don't have 10XXM chips yet so we'll have to wait and see on that front. If it continues the trend set forth by the desktop cousins, then performance increases should be pretty similar which is exciting and could indeed leave Maxwell users in the dust. This is a notebook/laptop forum after all. Related: unfortunately, AMD are still a generation behind in the power consumption game.

As far as Time Spy though, I think you may just be looking at the wrong parts of the internet. It benefits AMD WAY more than Pascal.

The reason Time Spy was released so close to Pascal is probably because it was the first NV hardware generation to have async compute abilities and more comprehensive DX12 instruction support from both sides. Basically, nobody knew what they were doing before and Futuremark even consulted with AMD/Nvidia/Intel on how to build Time Spy so that it more or less accurately represented an advanced "game scenario". Part of Futuremark's goals have always been to push new rendering techs earlier to "show off" what can be done and thus have always favoured whatever new rendering techniques came about.

This happened before with Firestrike and Tessellation. People complained about NV "gimping" Kepler and Fermi chips because Maxwell did so much better in Firestrike. It simply came down to the fact that Maxwell was way faster at Tessellation, not that Kepler/Fermi were made to go slower. Granted, they did have the odd dud driver in that time.

Furthermore, both Red and Green team sorted out how their Crossfire/SLI/MDA support modes for DX12 were actually going to function. This presented a much more even playing field for both teams without having big architectural changes or driver issues skewing results. Personally I think LDA Explicit without a bridge is the way to go. It could even lead to more specialised GPUs since it's up to the software/engine to divide up tasks.

 

Probably. Maybe im just reading into it too much, and have been annoyed at NVIDIA. And no doubt the Pascal cards will leave Maxwell in the dust. Lets see how it plays out.

 

Anyone is free to make a discussion thread about the benchmark itself, it just makes it really difficult to see the results when you have to go through so much OT to get there which is why I banned discussion in that thread.

 

@Phoenix what's your BIOS settings for 4.5GHz?

 

Posting up my top FireStrike 970m scores!

970m w/ 3610QM Drivers v368.39
Core +354
VRAM +560
mV +100

Overall: 8029
GPU: 10037!!

Link: http://www.3dmark.com/fs/9486694

 

No offense but : http://www.3dmark.com/fs/4349123

It was in my old Alienware but my 970M in that was a beast before Windows 10 killed it!

 

Thanks for your comment.

I do agree, especially after all i had to go through: multiple mistakes and bad settings, freezing and brick-like laptop

but i have learnt and improved from this...

Nevertheless, i love the opportunity to build one large ssd from 2 small ones, plus Pci-E speed

for a lower cost than AHCI !

 

Nice!!

Got me beat by 242points! I will be OCing my 970m even higher for sure!

You got me a goal of 10279 to beat!

What was your stable gaming clocks? core/VRAM/pwr ?

Also what were you core/VRAM/pwr increases to get the 10279 score?

Thanks!


____________________________________________________________________

Update: Just got a little closer:

http://www.3dmark.com/fs/9510693

GPU:10104

Only 175 points to go

 

Anyone know how to get past +100mV limit on MSI afterburner? I think I can use some more juice.

I tried this:

But it didn't work.

 

Can you all post comment in this review http://www.tomshardware.com/reviews...aming-laptop,4680-2.html?579773ba2491b=reload
@Mr. Fox @bloodhawk @Phoenix @Prema @TBoneSan @ssj92 @dspboys +++ all other who prefer real performance from laptops!!

 

Nvidia inspector also shows +100mv limit?

 

Simple. There are two superior options to MSI Afterburner.

NVIDIA Inspector Download

EVGA Precision X Download <<< this is an old version I use that has never bricked an LCD on vulnerable LVDS connections... sharing it from my Google Drive.

If the main reason you like MSI Afterburner is the on-screen display, use HWiNFO64 and RTSS and it will do far more, far more superior than MSI Afterburner can for on-screen display.

Posted my two cents in the forum there.

 

Hey guys was wondering if the 87c peak temp i got from running wprime(8 threads) on max fans +delidded cpu w CLU is normal? i am running at 4.2ghz.

 

For those that care, something I have noticed is that when stuff gets crammed into NVRAM over time this can lead to corruption and cause your machine to become unstable. That is one of the reasons I hate storing system settings in NVRAM. It is prone to issues and I much prefer the old legacy (non-UEFI) BIOS design even though it does not look as good. I'm more about function than having a nicer UI, but that's just me.

If you find your machine suddenly getting unstable for no apparent reason, or your CPU overclock settings suddenly do not work as great as they used to, pull the AC cord, shut down the laptop and pull the CMOS battery out for a few second to flush all of the saved up crap and that may help. Just be sure to write down or photograph your custom settings first.

I have also noticed that tweaking the RAM with Thaiphoon Burner can cause wrong values to populate NVRAM, same as swapping out memory modules without clearing NVRAM first. After clearing NVRAM, run a complete boot cycle as least once with the default memory profile before attempting to set an XMP profile. If you try to make too many changes without a clean boot in between that can stuff up the NVRAM with garbage, too. At least it seems that way based on my observations.

I believe this is one of the reasons changing memory modules can cause instability and failure to POST like what @dspboys experienced trying to upgrade to DDR4-3000 RAM. Below are two screen shots showing what I am talking about with regard to the crap piling up in NVRAM. The first one shows a ton of garbage on memory SPD with most of the values being invalid and unbootable. The second screen shot taken 5 minutes later show what is stored for SPD after clearing my NVRAM.

 

LOL those timings.
Thanks for looking into this. Indeed helpful for anyone looking to upgrade their memory.

 

Definitely high. But depends on what the ambient temps are like. Also are you un-dervolting ?

 

-140 on core and -70 on ring as what @Phoenix suggested. Could i go lower on both?

Sent from my SM-G935F using Tapatalk

 

@Mr. Fox,

What is writing to NVRAM? Isn't it just BIOS settings? If one gets to a stable spot, and makes few (if any) changes to the BIOS, could the NVRAM overflow corruption be avoided?

TIA.

 

Definitely high then. What is the ambient temperature like?

You can give it a shot though. When i was at 4.2Ghz , i was running it at -155 and -160 (Core/Cache) , it will be different for you.
Did you disable C-States by any chance?

I suggest this to everyone to leave C-States on. Turning them off causes the processor to always run at the max speed, and an unnecessary increase in temperatures. Let the processor drop clocks , doesn't lead to any performance loss in normal use.

 

Memory timings are part of the BIOS settings. Unless you are hunting for your ideal BIOS OC settings there should seldom be a reason to change anything. While you are hunting for the sweet spot, or just experimenting because you enjoy being a geek like so many of us do, changing a lot of things around, that might lead to some data corruption, especially if the machine fails to POST correctly. It's not affecting the BIOS (UEFI firmware) and the corruption is not permanent. It is only the non-volatile information being saved to scratch space on the EEPROM, which isn't a huge problem if you flush it and start off with a clean slate. I think some of the data may also be stored in a non-accessible area of your HDD/SSD boot drive as well. This is not something to be fearful of, just wanted to create an awareness that flushing NVRAM can sometimes be helpful in fixing weird behavior. When you flush the NVRAM, almost everything reverts to BIOS/UEFI defaults and you lose custom settings. That's why taking a photo or writing them down is a good idea. In the best case scenario you will have to reset the CPU and memory OC values and ACPI settings. In the worst case scenario, custom drive settings will prevent you from booting Windows or can corrupt your Windows installation (changing AHCI to RAID or vice versa).

 

Uh oh, I think Brother @CaerCadarn clicked the wrong button by mistake...
http://www.tomshardware.com/forum/id-3130891/avadirect-avant-p870dm.html#18336888

 

C states is enabled. You mean -155 on core voltage offset and -160 on ring voltage offset right?



Sent from my SM-G935F using Tapatalk

 

I disable c-states in the BIOS (my preference since 2920XM days) and use ThrottleStop to enable CPU downclocking (check C1E box) for normal use. You can also just switch to the balanced or power saver profile and the CPU will still downclock with c-states disabled in the BIOS.

 

I am seeing better temps now at the settings you recommended @bloodhawk


73c peak. Idle temps around 38c in an a/c room now and got my aircon set to 18c.

Sent from my SM-G935F using Tapatalk

 

Yeah, those were the offsets i was talking about.
Are you using CLU on the IHS as well ? If so, then i have a feeling its not making proper contact.

Were you able to get your Keyboard issues sorted?

You sure it was a mistake ?

 

Nice!

Make sure you run a bunch of stability tests, those offsets might not work for everyone.

Apologies for the DP.

 

No clu on my ihs. Only the inside is w clu and the outside w icd. I only have half a tube of clu left or less cause i decided to sell 2/3 for 32usd each .Yah issue sorted out infact it was cool of them to deliver my system to me the following day! Just had someone apply carbon fiber on the top area bec the top part is a scratch magnet now it looks pretty cool imo no more fingerprint and scratches! What tests should i run to check if my system is stable? Wprime on 8 threads enough?

Sent from my SM-G935F using Tapatalk

 

That looks nice. Where did you get the skin? I would not mind having one to cover the white lid on the P750ZM, but I've been worried the skins might curl up around the edges and start wanting to peel off. Been thinking about taking the white lid off and painting it.

 

That looks dope AF.

Glad things got sorted. 3EZ4RTZ.

 

A local store in philippines does custom skins for everything "decalboyz" i think was the name of the shop. Had tons of options including a croc skin but that was too flashy for me. Wanted a clear skin but according to the guy it tends to bubble up i guess this was the best option. Just brought my laptop to the shop and after 20min voilà! Perfectly cut skin!

Sent from my SM-G935F using Tapatalk

 

Huh?
Yeah, well they actually don't...
Just helped a 'VERY TALENTED' customer of theirs, who replaced his entire old rig (pics) with a single OriginPC EON17 SLX (P870DM), to 'Mod Up' the BIOS for his mission to solve a major 'mathematical puzzle':

Spoiler

http://meboo.convexoptimization.com/Meboo.html

All the best Jon!

 

Are they growing laptops like potatoes now? Just kidding..

But i have a feeling, its just a slightly unlocked version of the stock BIOS, without any proper under the hood tweaks. Im sure Clevo charges them more than normal for that, or they might be doing it in house.

 

Exposing the basic Intel XTU OC Options also in BIOS is something all newer Clevo BIOS do and not vendor specific:

http://forum.notebookreview.com/threads/clevo-overclockers-lounge.788975/page-424#post-10298620

 

Ah, i think i saw that in @jpsm 's BIOS he got from Sager.
Even then, i would love to see what is so special about that EON17 BIOS.

 

That's the point, literally nothing. They probably just compared two BIOS revisions or CPU samples against each other...the older stock bases simply ran better...