Clevo Overclocker's Lounge

Damn. Any news on if that is a LGA chip in there?

 

I think it has the same tdp as the desktop version https://www.amd.com/en/products/cpu/amd-ryzen-7-1700

 

Yeah for actual workstation work, that chip will easily beat a 6700k/7700k at stock. Not that we can sustain high clocks (4.9Ghz+) for long rendering loads, using the 6700/7700 on our Clevos, but if it IS LGA we might just be able to throw in a 1800X in there with a bit of cooling mods.

 

AM4, which is PGA. I'm guessing the firmware has a white list block you'd have to break first, and that the gpu is soldered... But, after getting around the restriction, 1700 and 1800X are the same socket.

Sent from my SM-G900P using Tapatalk

 

Ah yes, i forgot the AMD PGA part.
But yeah the white list part wont be that big a deal. Till the time it can be done, i wouldn't really case about the BGA GPU. since this would be a seriously amazing portable render rig / extra cores for remotely offloading test/preview renders to.

 

Remember, the 1700 overclock nearly as well as the 1800X, gets slightly lower bench numbers, but I'm not sure you would need the 1800X for it except numbers chasing.

Sent from my SM-G900P using Tapatalk

 

But for gaming w/c is what matters to me I still go with 6700k/7700. Even the ryzen 1800x is still below 7700k in terms of gaming. But nevertheless if it would give intel a good competition then we all benefit from it!

 

Interesting. I haven't followed the AMD developments as much. But if the 1700 can do what the 1800X can just very marginally lower, then even better!!

True, thats why i specifically mentioned rendering workloads.

 

http://www.tomshardware.com/reviews/overclocking-amd-ryzen,5011-5.html
Just fastest article I could find, but all pretty much agree with these findings!

Sent from my SM-G900P using Tapatalk

 

Mr Fox
so are you only using SKylake IHS and not modded welded shim? Or Both?

I want to improve my temps on my 7700k. I was thinking on doing the welded mod buf if your are getting good results with the Skylake IHS then I might just go that route... seems easier too than the shim..

 

I do not have a shim. I sent the prototype heat sink with the soldered shim back to HIDevolution when I installed the Bitspower Skylake IHS. I did not try the two together. I should have, just did not think of it at the time.

 

These are decent temps with cache ratio and voltage set lower. Have you already tried this? If you have XTU installed, you may get frequent crashes with trying to launch ThrottleStop. I have to uninstall XTU or disable the Peformance Tuning Ultility service to avoid it, or open XTU first and change multipliers up or down, then apply, before opening ThrottleStop or I get a BSOD or lockup. Both the Tornado F5 and the P870DM3 have this behavior.

 

What are the settings for your 7700k at these temps, if you don't mind? I imagine you are using a static core voltage?

 

That screen shot shows the ThrottleStop settings for 4.8GHz. I set the BIOS to 47x4, 160000 for PL1/PL2/PL3/PL4 and static voltage at 1150 (1.150V). Just a minute and I will post screenies of the other profiles for you.

 

Haha. My apologies, the screen was blown up so large that I didn't even see the throttlestop settings when I was admiring your thermals. Have you found setting PL3&4 to have any impact on performance or thermals? As of now, I believe mine are disabled.

 

I did not think PL3/PL4 had any effect, but if I leave them to zero (Auto/BIOS defaults) it does throw errors in HWiNFO64 at higher overclocks (5.0GHz and up) so apparently they do if the BIOS default is not enough. I also have C-States disabled in the BIOS.

 

I have not tried throttle stop. I dont have intel XTU installed either. I got all the cores at 4.5 in bios and static voltage of 1060. My temps are still in the mid 80s while playing battlefield 1 (pretty cpu demanding game) usually in the low 80s for most games.

 

ummm... I dont remember messing with the cache ratio... its probably set to the default value. I will look in to it tonight. What cache / voltage do u recommend? I am at 4.5 on all cores at the moment.

 

How is this for P775, 6700k 4.0GHz and GTX 1080?

EDIT::: 80CPU, 76GPU Max temps during the run, the magic of @Prema infused on the CPU ♡

I'm still curious whether the score is good or not!

 

friend that's literally stock settings

 

Aka work as intended

 

Yes, I am curious if the scores are good for the settings!

I would love to have some gains but... It doesn't last. anything over 4.0 GHz will throttle the CPU and I am on stock vBIOS...

 

The Mod has the EC throttle removed, but on P7 you still have to use TS to stop the OS from disabling Turbo.

 

We're forgetting something though. Ryzen CPUs NEED fast RAM. If they're tossing 2133MHz cl15 or 2400MHz cl17 at that thing, it's not going to make a lick of sense.
And the overclocking is also board dependent on the 1600/1700 versus the X chips, from what I see.

It's very much possible that this laptop is going to keep the CPU at stock speeds and no higher, so let's not get hopes up anytime soon. This ** IS** ASUS we're talking about here...

 

I put my answer regarding Ryzen and Asus here No Prema and no Svet. Expect a disaster!!

 

Even if you had unlocked firmware, if the machine only uses low-speed RAM it's borderline worthless, plus cooling issues. ASUS isn't like Clevo who has a lot of potential in the cooling system; their systems basically benefit from a repaste and that's it. If that isn't enough you're screwed.

Remember, the crossbar linking the CPU onto itself ticks at RAM actual speed. Higher RAM speed LITERALLY means the CPU is faster, period. It's not like intel where it just benefits in some cases, for Ryzen it essentially bumps IPC. You want 3000MHz+ RAM for it. And we all know how much ASUS cares about their laptops...

 

I think I'm doing something wrong, don't know how to properly set up TS and the first one that comes from Google is Obsidian TS. I just open it and it makes Turbo go well?

Also, 220 seconds for TS 1024M 8 threads is a good or a bad score?

 

Does the power window matter or is it recommended to leave that at 0 (default)?

 

After the AGESA 1.0.0.6 update, support up to 4000 MHz on divisor was added, including half-steps and sub-timings. Many have reported 3200 and 3400 on 4 dimms (including some at 3400 with dual rank), previously very hard to accomplish. That will be the tests you want to compare to. The IMC favors Samsung B-die chips. So, they will incorporate it as they have the early beta on the CrossHair VI with that update for the microcode.

You may be right on overclock. That is TBD.

Sent from my SM-G900P using Tapatalk

 

Your statement about RAM only applies to desktops. Where are they getting LAPTOPS to run decent RAM at 3000MHz for Ryzen? This is the main issue. I'm talking about that ASUS laptop. In all likelihood it will sell with 2400MHz cl17 RAM at the most. And even if they toss 2666MHz at it, let's look at what we've got:
G.Skill 2666/2800, 18-18-18-43.
Kingston 2666, 15-17-17-35
Hynix 2666, 19-19-19-43 ( I'm not even joking about that piss-poor garbage excuse for RAM).

That Ryzen chip is going to be gimped, even if you can keep clockspeeds up.

As for the 4 DIMMs issue on desktops, it's more because they were forcing the RAM to run at 1T. The Flare X chips had the best chance by far of doing this, being AMP-designed. You had to run through memory training on the boards. I had people tell me the B350 Tomahawk MSI board wouldn't run at 3200MHz and they had to settle for 2932MHz before, and I made them kill the CPU overclock and try tossing in the timings manually and run a couple boot loops (like how I was taught to get the timings trained here) and it worked for them, and then they just overclocked it again after. So now I suppose it's just easier, or people are tossing the timings to 2T and it's working better. *shrugs*.

 

It is easier with this update. WAY easier. But you have a good point, mobile doesn't have great ram support to begin with. Why I brought up B-die is that if it is what G.Skill uses on its 3000-3200, so long as proper support is given, it should see support. This is also speculating due to the upcoming mITX AM4 Ryzen boards, which will have similar support, potentially, to this laptop board...

Sent from my SM-G900P using Tapatalk

 

Take a look how to use TS by @Mr. Fox

 

Yeah it's just it's gonna be a laptop board, using SODIMMs. You know how ASUS treats their laptops. Plus those SODIMMs aren't the greatest either; they wouldn't even be used in the existing systems that can use them without serious work from Prema

 

As posted previously. 42x with static cache voltage set in ThrottleStop between 1.050v and 1.060 (as low as stable will let you, so start with 1.050v) and if that is stable, leave it there. If you're going to stay at 45x4 you can likely use static core voltage between 1.060 and 1.070.

 

Thanks Mr Fox. Going to try that see if it improves my temps.

 

*sigh* all this back and forth concerning coupled or uncoupled cache and core voltage.... with haswell it was independent from each other, on skylake i couldnt set those two independently, but now kaby lake is separate again? hmmm....back to the drawing board and tweak some more with TS.... @Mr. Fox keeping me on my toes here

 

It was separate on Skylake and worked fine for me setting cache voltage in BIOS at whatever value worked best. Having them coupled or fused to the same value is better for high overclocks with high cache ratio, but worse for temps at all clock speeds. Sadly, it is not separately controllable with Kaby Lake unless you use ThrottleStop because the BIOS settings are missing for cache voltage and the BIOS (or XTU) matches them to the same value by default. If you are going to use Intel's default approach here, better to set cache max OC ratio to 83 and core max OC ratio to 83, then the BIOS will dynamically control the cache ratio. Cache will always be 3 bins (300MHz) lower than what you have your core multiplier set to. But, temps are going to suck because the cache voltage will always be too high. It will be stable, just too hot.

In fact, this stupid approach on Intel's part probably has something to do with the whiners complaining that 7700K shouldn't be overclocked because it runs too hot. Of course it does if they don't know what they're doing. The combination of engineering error and user error is not a good one, LOL.

 

This sounds interesting, but I don't know what to say. If they developed better fans and better fins, wouldn't it be interesting to make huge headtpipes for P775 out of those?

http://www.pcgamer.com/asus-stuffs-...source=facebook&utm_campaign=buffer_pcgamerfb

 

You assume clevo doesn't do the bare minimum so that the laptop doesn't explode.

 

Oh, I'm actually intrigued by the idea that someone could discover an even better cooling mechanism that could make Clevo laptops better. I know they do their math and I count on Clevo doing their best, just intrigued by the possibility of improving an already great product.

 

I'm actually retweaking my CPU too from adaptive -75mV to right now 0 Offset and Static Voltage 1.145v which results in about 1.22v when stress testing with x264 encoding. good to know that it's 3 bins below the max core ratio, i set e.g. in my bios the max ring ratio to x42 but it still stays at x40 ratio whatever i try, i can only force it when using min x42 and max x42 or using Throttlestop, but i noticed when tweaking that going from x40 cache ratio to x42 ratio my CPU get's unstable incredibly fast and i BSoD out with 0x1b errors, when voltage was too low i still get the old trusty 0x124. But shouldn't acutally my CPU using when set x42 ratio in premas bios? cause cpu ratio is x45 and yeah 3bins lower would result in x42.

So far what i noticed between adaptive and static, it still flunctuates ... a lot acutally, when using Throttlestops integrated 1024M test i'm at 1.18-1.19v regardless of adaptive and static. wPrime and x264 using 1.22v as expected, but like you said some months ago static seems like a more "stable" fluncutation, too bad we don't have LLC settings on the Notebooks . Adaptive "default" seems like the lowest it can go anyways e.g. i can set 1000mV in bios and voltage stays at 1.35v stock and have to use -80mV to get actually 1:1 values for what i set it, so 1.22v with stressing, 1000mV still 1.22v but if i set 1250mV in bios with adaptive i get 1.25v so it somehow works but only with higher values, sadly not below.

 

yeah im unnerved by these fluctuations as well. for all their shortcomings, haswell cpus had one BIG advantage: static voltage meant STATIC voltage!

however, for now i at least have the option back to set cache voltage separately from core, could give me more thermal headroom and better stability at higher overclocks

also, people, dont forget: overclocking cache IS NOT WORTH IT! +1ghz cache gives FAR LESS of a performance boost than ONE multi. keep that in mind!

Sent from my HUAWEI NXT-AL10 using Tapatalk

 

Yeah on the haswell it was really good after you figured out that static voltage results in the voltage you set + C-States is was practically nothing in terms of efficiency loss either, using static + C-States now too ofc.

what i forgot to mention tho is even when using static now and the flunctuations are still there i noticed 2 °C less with the "statistically" same voltage, x264 with adaptive netted about 91W and 80-81 °C across all cores and now 87-88w with max 78-79 °C, same room temps ofc cause when i dialed back to adaptive -75mV Offset 81 °C was back.

And yeah i know that Cache OC is practically useless outside of benching but i was just curious cause it didn't work out of the box, as far as prema bios is OOB, on bios level.

 

I feel like a dunce, I was overvolting my cache when I did my overclocks.

I couldn't undervolt my core without undervolting my cache, but it seems I can undervolt my cache while overvolting the core. GG throttlestop and me!

 

no worries, i was doing the same thing.however, with my previous 6700k it was absolutely not possible to set them separately, even with TS. now with the new beta mod and the 7700k the settings do stick though! thats why i didnt think to try it again until @Mr. Fox came up with it.

Sent from my HUAWEI NXT-AL10 using Tapatalk

 

Yeah I didn't realize it worked in reverse because it doesn't work forward. If I set static voltage to 1.05v (what I use for 4.4, I may be able to set lower even) and I set static cache to say 1.1v, my core will sit at 1.1v static.

but if I raise my core... I don't know. I have no way to check cache voltage; I guess it must be evident in temperatures? I never thought to do it that way.

 

Static meant static was what it told you, you had no idea if it was actually static of course

The load line calibration options are unlocked in a prema BIOS if you feel brave about working with ohm values

 

true that, ive seen the LLC options in @Prema bios, but i havent touched those yet

 

Just be careful with them, small changes can have large voltage swings.