Clevo Overclocker's Lounge

Shhhhhh, i was just testing you....mum's the word.

 

The mobile one looks pretty underclocked....
Great for benching AoTS and a better card than 980.

 

We can all hope so, but do we have any proof? I want to see if it can handle extreme overclock/overvolt, how much it can handle and what the Futuremark scores look like. If it does extremely well (meaning as good or better than NVIDIA's best) the only question will be, does it last less than a year for many like their last best mobile MXM did? It needs to overclock like a holy terror, get insane benchmark scores and live for as long as you feel like using it. Buying replacement GPUs out of warranty because you got sold a piece of trash sucks. I've had it happen to me with about 8 AMD GPUs from 6900M series forward (close to 100% failure rate), and Maxwell is not scoring any points with me on quality control either.

 

Same quality problem on desktop version vs mobile Maxwell also? Or is there a difference?

 

Nvidia do tend to skimp on their reference desktop cards power circuitry wise so that can be different.

 

Polaris should be better at future proofing in DX12 then GTX 980 is...

 

Here is one over clocked to around what you have posted.
http://www.3dmark.com/3dm11/11315966

 

At the moment on the desktop it looks like the new products are simply not competing.

 

@johnksss that link ended in a RNF. Did however see your SLI entry with 61k graphics score, very nice.

 

The problem with this is the expectation factor.
And what i don't get is this 1.25V limit everyone is worried about. Heck, i can't even get mine past 1.075V@2050/5500 without throttling,let alone anywhere near 1.25V. And not one link with anyone on air or water showing this fact. So if you know where to find any links to that, then please share. Because i'm thinking...1.25 is the least of someones worry at this point.

Thanks.
Try it now.

 

So, am I interpreting what you are sharing to mean it's OK stock, but sucks at overclocking... along with NVIDIA's signature throttling problems using their traditional cancerware stock vBIOS?

 

im guessing the 1.25V limit is only a concern for extreme LN2 overclockers ^^ everyone else on air or water should not, under regular circumstances, be able to reach that limit...

 

On stock vbios with no mods i am able to beat all my old titan x/gtx980/gtx980ti scores. All while not going over 1.075V.(Since this is good for about 2.1 to 2.2 on a good working day) So if i were able to get to say....hit 1.1V and no throttling. That would yield what most people are looking for. a 2.2 to 2.3 ghz core card.

Im still testing, but have seen a few things in the vbios, but am unable to save changes as of yet. To see if im even remotely close to being right.

Yeah, well 99 percent of them think they should. lol.
Yet not one has shown a bench where their voltage is even above 1.075V. All this talk about 1.25 is moot if you cant even get to 1.15V or 1.1V. That should yield some over head room. since the power target is locked down so low....Speculation of course.


Edit:
So being able to run a 1080 or 1080's in a laptop should show a nice boost in all around everything.. Be it mobile or not.

 

I hate artificial limits. More voltage typically means more overclocking headroom as long as you can keep it cool enough to use without overheating or thermal shutdown. I'd want to have the ability to push it to the edge of its functional limits even if I couldn't keep it cool enough using stock air cooling when doing so. I would choose thermal limitations over firmware or hardware limits any day of the week.

 

These last couple of generations don't seem to scale as well as older generations did with voltage, especially maxwell.

I remember my fermi card would keep going higher and higher as long as I added voltage, it became a temperature issue at that point not voltage lol

 

It all has to have a ceiling somewhere, it just can't go on for ever and ever...unfortunately.

And yes, i remember the fermi days. I had Vince make me a card back then.

Also, this card has only just come out so with anything....we wont know the full extent of what it can or cant do for a while....

 

TBH, it looks like just a process limit on air/water. Apparently when cooled properly, you can hit it higher.

Edit: Have EVGA FE coming in on express shipping.

 

When im able to edit mine, i'll let you know if that is the truth. But 1.075V@65C is plenty of wiggle room for improvement if you ask me.

 

Sweet! tell me if you break 3ghz on LN2 haha. if you are planning it.

 

nah, im going to stick with air..

 

What's your opinion of tec/block on water loop

Sent from my SM-G900P using Tapatalk

 

I wanted to remove this **** BGABooK from first place in Fire Strike Extreme (single graphics). A BGA machine has nothing to do in front of a socket machine!!! Now is the garbage removed from first place .


http://www.3dmark.com/3dm/12418032

 

After seeing your screenshots of 1080m, I thought you had one.

 

Agree with you and did mention that 120% TDP limit was being hit at 1.050V here.

But people want to experiment and that's not always a bad thing. An excellent guide for those wanting to run higher voltage / power ( @tgipier?).

As you say @johnksss, still some unansewered questions. My Galax card limit is 1.093V with 100% extra voltage running while default is 1.062V so that 100% gives a whopping extra 30mV lol.

I run my benches with fixed settings, something that's easily done with a little software but still subject to adaptive changes with temp and power limit. The adaptive temp is strange in that it seems the clocks will drop down a touch as the temperature increase regardless of what frequency the bench starts with (well I never actually tried with the much lower clocks).



Here's the default voltage curve of my GTX-1080 and while minimum is about 0.45V the card seems capped to 0.625V minimum in operation. It gets pretty flat at the top end and not sure what that dip is about at the lower end. Of course all these points this can be programmatically changed which is a nice feature.


Then to add another question there's this post from @Mr.Fox showing 1.225V for a 1080m. Not sure what to make of that.

Let the fight for first place begin or has it ended already.

 

Some with a 980m should run bench with single graphics and remove this BGA junk from the first page in Firestrike Extreme!!
Edit. Something is wrong with this score!! Is Futuremark on berry-picking? This is perverse

 

@Mr. Fox post shows the GPUZ pic, but as for the transistor count and die size, it shows for GM204.

Edit: Nevermind. I clearly didnt get the joke here..

 

It looks like pascal had a lot of core variability. You need a good sample and good power delivery/cooling to take it far.

250w tdp limit of an msi desktop card covers you there but one ws only reaching 2050Mhz with that limit.

 

He linked to the post to indicate that we should not believe everything as it is presented to us.
We have a sample, but it's not with Fox...he's just playing with his 980, which has about identical results clock for clock...that way you got the hang of it and no NDA was compromised...

 

Im starting to get the hang of it. im up to 1.09V...whoo. hoo...lol

 

that cant be right, especially beating khenglish and his monster 980M with DT970 vram...

Sent from my Huawei Mate 8 NXT-AL10

 

i couldn't catch the score either....
But i don't think msi came with 8 gb 980m either. Only 4gb
http://www.3dmark.com/compare/fs/8229521/fs/6967989/fs/8709361

 

It depends on the model for vram amount.

 

They had 4GB and 8GB cards...the 8GB are identical to Clevo's and can run the same vBIOS.
1753Mhz on vRAM...could be legit if hard-modded...but may have very well been its last bench, too.

 

talk about a suicide run

Sent from my Huawei Mate 8 NXT-AL10

 

Oh I see. Thanks for explanation.

 

Even voltage can not really help as it's the bus itself that is the bottleneck rather than the chips.

 

For that FS bench it seemed to have a huge impact as the core is just sitting there at some 1350, compared to John's 1550 and Khenglish's 1430.

 

aaaah, well that definitely needed further explanation, didnt get it

Sent from my Huawei Mate 8 NXT-AL10

 

Yes if you lower the core you do tend to get beeter memory clocks, less interference.

 

hummmm.....

Very well could have been. It seems to be his one and only bench ran on that day.

That is interesting..considering all the cards one puts on the same exact bus.

1350 translate to some like 1440 or so on core.

 

I'm talking about the memory bus rather than the gddr5 chips themselves.

 

They are all on a 256 memory bus.
You were 1650 or so.
That guy is 1753.
Khenglish is 1873 with 7ghz chips.

At some point there has to be ceiling.

 

I really don't know what 3DM is doing, sometimes (like on my W230SS) it shows Turbo clocks, then just 3D base clocks and other times only adds the Turbo on top of 540Mhz, all while using the same vBIOS...then they even change already uploaded results later on...

Or look at this one: http://www.3dmark.com/fs/8197706

Where the heck does it get it's clocks from? I didn't bench @ 1807Mhz /1633Mhz
My vRAM can barely handle 1350Mhz and the core runs about 1500Mhz...

 

I quit trying to understand that one. lol

 

The culprit must be Win Crap X!!! HaHaHaaaaaaa . The same crazy OS the BGA machine used in the golden Firestrike Extreme bench before the card went to heaven!!! LOL

 

@Prema

Ummmmm, aren't you suppose to be working on something?..

 

I wish there was a way to automatically exclude outliers when searching through 3dmark results, namely @Mr. Fox and @Papusan. I only want to compare against other mortals

So after having no luck with newer drivers I tried the good old 362 drivers that seem to still be many folks favorite, and I beat some of my own scores and overclocks. I created a new windows account so I could disable all sorts of things I don't need when benching but later realized it also screwed up my main account (not sure I understand why those settings are shared across different user accounts). Temps were maxing in the 60's for CPU and GPU. Nothing out of ordinary at all for what I've seen the last few weeks. I've been trying to find the limit on the GPU on stock voltages so no overvolt last night. CPU the entire time was 4.4 @-100mv which seems to be very stable as I've been benching and gaming with it for a while now.

When I hit the limit the screen changes to one solid color. 9 times out of 10 I have to hold the power button, but occasionally the system recovers and says the video driver crashed, but I have found it's best to reboot anyway because there are usually lingering issues at that point. It seems to do this whether it's the core or memory clock that's too high which makes it hard for me to figure out what to lower or try next.

Late last night after doing this for a few hours, I found out I could not overclock the GPU as much anymore. Even the "daily" OC of 175/300 that I had been running and playing games with for a long while would crash. I can no longer even get close to 210/410 like I did last night. I just can't figure out what changed.

Any words of wisdom for me? One thing I want to try is go back to stock on the CPU and see if it changes anything. I hadn't tried it yet because I honestly had not suspected it until just now.

This is harder than usual because the temps are so dang low I can't even figure out what is going on. Temps are good, plenty of power available, I know I can try to overvolt but I really just wanted the best I could do on stock voltage but that seems to be a moving target.

 

seems like there are no more new BIOS/VBIOS for our machines

 

Probably not.....Considering the machine is working pretty darn good now.
At the end of the day it's still a laptop and it will not surpass a desktop (People who know what they are doing and not the average joe) with the same main components. This needs to be accepted unfortunately.

 

There may be one more if clevo allows kaby lake instead of cutting off support, but that may fall to prema if they decide not to for higher sales of new versions. ..

Sent from my SM-G900P using Tapatalk