*Official* NBR Desktop Overclocker's Lounge [laptop owners welcome, too]

It's a fun ride. I completely enjoyed using AMD's platform. You have more tools at your disposal, so you'll be able to push things a little farther on the CPU side. Memory tweaking is important (as it is on 11th), but while it is fun pushing AIDA scores, those gains rarely correlate to real world application gains after ~3600/3733 1:1 and tightening up primary and secondary timings. If you're chasing optimal memory timings for benchmarks, that's fun. I'll be curious to see you push a high 1:2 (the future path it seems with DDR5 down the road). Sometimes the latency tightening and bandwidth gains are appreciable, but turning around and testing in .exe's was a mild yawn at best.

I'm most curious to see ambient water vs chiller gains.

 

@Mr. Fox

When you get a chance (and play around with it for a bit), can you post a few shots of the per core controls of the Asus board and secondary timings screens in BIOS?

 

Yes, I will. In fact, I have a capture card that will allow me to create a video of the BIOS navigation.

The motherboard is rather disappointing in terms of build quality. It feel so cheap and flimsy compared to the EVGA X299 and Z490 Dark mobos and the old X299 Rampage board. While I am working I am allowing Hydra (successor to ClockTuner by 1usmus) to stress tune the highest possible clocks and voltage. It is on the second hour of auto-tuning and stress testing right now.

 

I had a very stressful time getting this system to function and couldn't imagine what the deal was. Whenever the GeForce driver would load, the screen would go black and had strange flashing behavior. Then it dawned on me... the vertical GPU riser cable is PCIe Gen3. The BIOS was set to Auto (Gen4). After I manually set PCIE slot #1 to Gen3 all of the problems I was having disappeared. I ordered an absurdly overpriced Gen4 riser cable to replace the ridiculously expensive Gen3 cable. I'll have to do back-to-back Fire Strike runs and see if there is a difference in graphics scores with Gen4.

 

I would love to record videos in my bios, and I’m general too. What are you using to record with? How does it work exactly?

I really want to record my monitor regardless if my PC is in windows or not.

 

How is this platform anyways? You mentioned reliability issues in the past. Has all of that been worked out, and are you going to run the X570 Dark?

 

All's well that ends well. First thing EVGA wants you to do is give it a shake. That didn't fix it but did change the noise so I ended up connecting my 3090 AIO directly to my motherboard to do a slow ramp up (10% intervals) while shaking it and giving it love taps and all the noise went bye bye. I remember doing that to an old Corsair H100 years ago to fix the bubbles/grind/noise. It's actually slightly quieter than before. I run tubes up, front mount so bubbles do stand a higher chance of traveling down but eventually migrating back up too.

 

It is one of these... https://www.amazon.com/gp/aw/d/B01MYWBG1I and it works nicely for screen capture at 1080p 60Hz. It records MP4 direct to USB. If you are using a 1080p 60Hz monitor you can leave it connected 24/7 if you wanted to. It uses HDMI passthrough, so you could use it for video capture with a Roku, DVD, Blu-ray player or game console. You start and stop recording by pressing a button on top. It's very slick and works perfectly.

It cannot capture video at a higher resolution or refresh rate and any passthrough video is limited to that. So, I hook it up only when I want to use it.

I have experienced similar things with DDC and D5 pumps off and on again. It can be annoying, especially in an otherwise quiet environment. The noise itself doesn't bother me much, but wondering about whether the pump is going to have a short lifespan can be a legitimate concern. That said, I have a Swiftec D5 pump that has been noisy since it was new and it has been running hard 24-7/365 for about 3 years now and it is still going strong.

 

Oh thats a neat adapter, it doesnt seem to go with the super 1337 Champ1onship Esports king I used to find when I was looking around a while back.

I just ended up using a 2 PC method since I had a few 9020m's laying about, works perfect for when in OS but obviously doesnt work for your scenario where you display the sBIOS and such. I dont even use mine too much, mostly for when me and a friend are bickering about what is and isnt there in a game / game menu so I can throw it on real quick and settle it lol.

All that technology and software, some people use it to make a living, I use it to settle arguments that ultimately change nothing anyways :/ I should probably get a life at some point here....maybe tomorrow (lmao)

 

I've been running an AMD proc since the 1950X (well I used them before that but alas) and my knowledge of these platforms have grown since then. TR4 had its fair share of troubles but I have not had any big problems with TRX40 (mobo specific troubles, yes) or X570.

If the X570 Dark launches soon, I'll pick it up, but I'm thinking hard about throwing my 5950X into my Gigabyte Xtreme build.

 

it is definitely not as simple and responsive as overclocking an Intel CPU. There are more poorly documented BIOS settings than I have seen in a long time. It seems like maybe even ASUS does not know what purpose some of the settings serve. While most things happen very fast, other things randomly seem to have a tiny bit of lag or latency.The 5950X is super-powerful, but not quite as crisp and snappy-feeling as 10900K in some ways. I think that might be inherent to Ryzen's chiplet architecture, but it looks like Intel is poised for similar product design.

 

Finally we have it Gold for @Prema Team

https://hwbot.org/submission/479318...20_(mobile)_594.51_dx9_marks?recalculate=true

 

Power limits are what is holding this back. If I bump the power limits to 650W and 350A, and set the C0% to <50% close to zero performance improves. It's no longer trying to be an "efficent" piece of crap, it's just going balls to the wall (which is what I want.) I was seeing like 350W and now it's closer to 600W with the same voltage and core clocks. Now, if I can just get it to 1100W like I did with the 7980XE we might be getting some really scary Cinebench scores.

https://hwbot.org/submission/4793201_

https://hwbot.org/submission/4793199_

 

I think that’s pretty good so far.

 

6900XT Liquid Devils are in stock on EKWB for

$1,899 if anyone is interested. Or in need!



https://www.ekwb.com/shop/powercolo...ercolor-liquid-devil-6900&utm_term=2021-08-03

 

I've noticed AMD cards popping up and staying in stock now on Best Buy and Microcenter. Prices are ridiculous as I guess nobody wants to pay almost 1k for a 6700xt. These are the official MSRP prices to boot.

I'm going to get me a 6900xt one day at 999.99 or close to it. I don't mind waiting.

At least the last two sets of Nvidia drivers finally fixed the flickering texture issue in WoW that Nvidia broke way back in November....then again I haven't played WoW all week. Something about 9.1 and Blizzard atm has really soured me on it hopefully short term.

 

https://hwbot.org/submission/4793457_papusan_3dmark2001_se_hd_graphics_520_(mobile)_26471_marks/

 

I got my eye on that Asrock 6900 XT Formula OC. It was 2500, now it's 2269. I'll get it once it goes below 2000 and I'll sell the 6800 XT that I have.

 

That's a nice break from the norm in aesthetics and lighting. I like it. It reminds me of my "if price were no object" love for the Toxic 6900xt Extreme Edition. It's probably my favorite 6900xt card.

 

Rumors. But more of the small power efficient phone cores, LOOOL

Why not just skip the few Big cores and fill up with the small one. Maybe we would see around 80-100 of them, HaHa

Arrow Lake and Lunar Lake are supposedly both using Lion Cove high-performance “big” cores and Skymont high-efficiency “small” cores. According to the leak, Arrow Lake would, once again, double the count of small cores compared to Raptor Lake, which is already rumored to offer doubled core count over Alder Lake. This means that Arrow Lake would feature 8 big cores and 32 small cores and 40 cores in total.
https://videocardz.com/newz/intel-a...es-appear-in-a-leak-as-meteor-lake-successors

 

I would be 100% fine if Intel manages to find a way to say, group 4 little cores into a single fast core depending on what the CPU needs. I doubt this will ever happen as even explaining it is weird, but it would be super cool and useful.

 

I want to try the 6900XT Liquid Devil Ultimate. But it’s just far to expensive to run something like this in a secondary PC.

Apparently this is the best bin available? But, why does it cost $999 more than the already cool 6900XT Liquid Devil? That’s a whole reference 6900XT on top. It sounds like the whole “Ultimate” naming scheme is an AMD moniker or something.

The Ultimate liquid devil cost $2,899. And I just don’t see paying that much. I mean, the 3090 Kingpin Hydro Copper is only $2300 and has 24GB of Vram and DLSS which I feel is far superior.

6900XT Liquid Devil $1,899
6900XT Liquid Devil Ultimate $2,899

They really need to get this pricing together. These MSRP’s seem too high.

 

Not only the better AMD cards have an disgusting price tag…. https://videocardz.com/newz/galax-g...-available-for-preorder-for-at-least-2672-usd

For pricing, the RTX 3090 HOF starts at AUD 3,899, the HOF Premium lands at AUD 3,999, and HOF Limited Edition SKU lands at a whopping $4,499. Converting directly to USD would be $3,007 for the HOF, $3,470 for the HOF Limited Edition, and $3,999 for the HOF Premium.

https://hwbot.org/submission/4671311_ogs_3dmark___fire_strike_geforce_rtx_3090_50114_marks

https://hwbot.org/submission/4689981_ogs_3dmark___time_spy_extreme_geforce_rtx_3090_14609_marks

 

Ultimate / Toxic (and a few others) use special binned AMD chips and charge a premium for them.

I do see many of these 2000+ 6900xt cards and ~1000 6700xt cards now readily in stock apparently buyers hit a wall of what they were willing to pay as insanity hopefully has its limits.

 

I wish there was a good way to delid and run this bare die. I am using KPX paste, but will probably switch to liquid metal when I have time to fuss with it. I got the PCIE Gen4 cable and it did nothing to improve benchmark scores.

@Rage Set - I was able to tighten up the timings without reducing memory clock. I was able to match the infinity fabric as well (2000 MHz) with the memory clock. There is definitely some responsiveness lag compared with Intel. Things are not quite as snappy with Ryzen. Boot time is longer as well. It is weird and quirky in many ways, but awesome in other ways. It's not a clear winner over Intel so far. But, I am not totally disappointed. It seems slower than 10900K and 7980XE in some ways, faster in others.

It was a terrible pain in the butt, but I did manage to get Windows 7 working.

 

I use to want a Galax HOF model OC Lab 2080Ti really bad, I always felt like it was the best model. But after getting my 3090KP I got in to a ‘Who’s is bigger’ with a guy who owns a 3090 Galax HOF. It’s the base model with an air cooler. But still a $3,000 dollar GPU nonetheless.

He said he could match my port royal on air cooling. And I’m on watercooling. And he almost did beat me!! I thought it was impressive, that he was within about 60 points of my watercooled KP3090.

But then I looked again, and turns out his card is actually running cooler than mine lol.

Galax HOF 3090 @53C 2160 internal clock. 15,485 Port royal

KP HC 3090 @58C 2175 internal clock.
15,561 Port royal

If someone were to buy one, I think the best model at $4,500 is probably gold!

 

You must have a good sample (well almost all 5950X's should be) if you got IF to 1/1 at 2000. I got my 5900X to 1900 but kept it at 1800 so I could get tighter timings at 3600. Another tip, some benches do better on Ryzen with four sticks of RAM...but you will have to loosen some of the timings. I saw how much power you pulled on your last bench, that was crazy.

You summed it up perfectly. It isn't a clear winner over Intel but likewise, Intel is not a clear winner over AMD any longer. Intel's 11 gen has similar latency tendencies.

 

If he really is using JUST the air cooler and it is running that cool, he should be able to beat your KP's Port Royal score. What's stopping him? He is probably using a portable AC like a lot of us use to squeeze that little bit extra.

 

One guy got Z590 DARK with 4000MHz Gear 1 on 11900K, and one guy got Z590 APEX XIII on same 4000MHz G1 11900K. I think it's a real tough luck there, IMC lottery. For Intel Dual Rank B-Die should be superb and 2x DRAM sticks as well for best performance (Don't have the screenshot from APEX XIII, it was on OCN, can't find the post)

More trash from Intel on this B.S, I call these as Joker Lake, Clown Lake. So many damn lakes and threw nm out of window to do the marketing PR bidding with Intel 7, 4, and 20Angstrom lol. Intel delivering on these promises is too much. Once ADL is out we will see how that Hardware scheduler is made, I think they are done with their core architecture series. SKL is the last uArch that was able to scale past 8C with Ring Bus and high speed DRAM and high clocks, end of an era for sure. Also DRAM is definitely going to take a backseat for 2-3 years of DDR5 debut, in the final stages of DDR4 we finally got 4000MHz C15/C16 binned kits.

AMD Zen 4 no ARM copy cores drama. I hope AMD gets all their driver mess right with the Zen 4 debut on their new 5nm platform AM5 socket and crushes Intel very hard that they would be embarrassed and go back to their stupid board and remove these core hyper nonsense about efficiency and ATX12VO joke. I do not want a monopoly but Intel is really horrible with these smoke and mirrors crap. Enough of their B.S you should have read their Interview on Anandtech by that VP of Xeon, it was complete woke cringe garbage. No wonder they are shoving this phone cores to save the planet from the evil men.

 

Ryzen 9 is not as nearly as geek resilient as Core i9. Maybe they think the complexity and excessive overuse of settings is impressive. The complexity is actually kind of annoying. As an example of what I mean by that is with Intel CPUs it will throttle and perform like crap if power or current limits are too low. With Intel, you simple disable or max out the power and current limits to make them higher than you will ever need them to be and you're good to go. In contrast, it seems that exhausting power limits with Ryzen 9 causes the system to power off with some goofy "protection" mechanism instead of throttling, and when that happens I have to turn off the PSU to reset the mobo or it refuses to power on again. So, after a lot of poking around in the dark corners of the Crosshair BIOS I did find the location to remove power limits. I had to accept a nasty and ominous warning from AMD about damage from overclocking not being covered under warranty (I have never seen any discouragement from Intel on that subject) before I could access those obscure BIOS settings. I set the power limits to 1000W and the current limits to 350A. Now the only limiting factors are voltage tuning and thermals. It doesn't shut off when I exceed power limits now.

I almost need a glossary of proprietary AMD terminology to figure out what the hell AMD is talking about with some settings. Many of their settings have names that are less than obvious, nebulous, or even obscure to the point that Google searching reveals nothing useful. The most common Intel BIOS settings are generally obvious on face value by what they are called and you have a pretty good idea of what they do in the context of what menu you find them.

I also don't like the C0% adjustments on each core. That seems kind of stupid and superfluous IMHO. I'd rather just have one BIOS option to disable all c-states, p-states and power management "features" to let it run balls to the wall with no concern about anything other than voltage tuning and controlling thermals. I think it could be made a whole lot easier through simplicity. I don't care about reducing power and having "gaming" profiles to balance performance with power consumption and gimmicks like that.

Now that I am used to EVGA firmware, I really love how simple they make it. They know that their customers buy Dark mobos with overclocking as the primary (perhaps sole) objective, and anything that interferes with that is disabled and hidden so it doesn't become a distraction when the focus is on overclocking. I actually like the EVGA firmware better than ASUS firmware now, but it took having long-term experience with both brands before I could appreciate simplistic elegance of EVGA's overclocker-focused firmware. (Thank you, Vince.) Having a crap ton of BIOS settings that are not capable of being fully understood seems impressive until you come to realize they are more inconvenient than useful. It will be interesting to see if EVGA cuts to the chase and makes everything simple with the X570 Dark like they have with the X299 and Z490 Dark. I suppose that could be dictated to some extent by the licensing agreement with AMD. Hopefully, AMD will let EVGA do what they think is best for overclocking enthusiasts and not require them to include the full gamut of weird and obscure firmware settings that add no value in terms of overclocking.

 

Put it this way, I don’t think the bin stop him

 

Those are sexy timings, with 16GB DIMMs no less. Makes me want the X570 Dark even more.

I completely agree with you. AMD requires Mobo manufacturers to keep a "copy" of AMD's original BIOS functionality/naming. Yes it is redundant. But I do understand why. A lot of mobo companies like ASUS like to rename certain BIOS features. They give them "clever" names but if I am experienced with a different brand, that functionality could be named something totally different.

Now that Intel is going down this road with their future platforms, a lot of Intel fans are going to have to relearn how to overclock. I am still bullish on both Intel and AMD though, at least for HEDT. I have a sneaky feeling that Intel is about to pull a Threadripper-like move on AMD.

 

Can someone dumb down Gear 1 vs Gear 2 really quickly, it seems extremely confusing on which is actually better for certain scenarios. I've tried watching GN's video on it, but I still can't grasp how it works.

 

I would like to know why it exists. Like so many new fads in tech, I question the basis for it existing at all. Lots of stupid stuff starting to surface.

 

So I managed to find all (3) 3090 Kingpin Hydro Copper bios versions.

(1) Stock 450 watt KP HC
(2) OC 480 watt KP HC
(3) LN2 520 watt KP HC

Anyways, PX1 software really has issues with the RGB on 3090 Kingpin Hybrids that are using the Hydro copper Kingpin full water-block kit. Apparently they render the RGB reversed.

So I just installed one of the bios, and once I opened PX1, it started updating the internal MCU immediately on my 3090 KP Hybrid. Once it finished I re-opened PX1 and now evga PX1 and everything else finally recognize my GPU as a 3090KP Hydro-copper. And not the Hybrid model.
And that auto fan stuff is GONE!!! Seems like wasted resources with a auto fan profile trying to adjust its self based on temps and a graph, when I don’t even have fans..

Just figured I’d share if anyone is curious.

 

My understanding of it is that the gear system allows for higher stable RAM speeds as the cost of increased latency. I'm not sure why anyone would want to increase latency since that seems pretty stupid to me.

Gear 1 has the lowest latency but gives you the least headroom for overclocking your memory. Of course your highest achievable speeds and lowest achievable latencies will depend on the quality of your RAM and your motherboard's memory controller.

Gear 2 allows you to run your RAM at higher speeds while remaining stable. However, this comes at the cost of increased latency. This option is actually useful for anything that requires maximum memory bandwidth.

Personally, I'd only run my systems using the Gear 1 option since low latency matters more in the vast majority of cases. This is the best setting. You want both high speeds and low latency. Good quality RAM can run at high speeds in Gear 1.

 

@Mr. Fox



The evga Hydrocopper manual says the left inlet is the correct way of flow, but the blocks arrows say the right inlet is the only way to run it.

This block looks like it is clearly setup for one direction of flow only (Right inlet only). And in this case, that would mean.. mine is hooked up backwards lol.

 

I have used it both ways and it doesn't make any difference on the 2080 Ti water block or the 3090 water blocks. The temps are identical for me either way. I alternate flow direction to keep the fins clean.

It definitely matters on the CPU water blocks though. The OptimusPC and XSPC Raystorm blocks do not work well if the water is moving the wrong direction.

Nice find, bro. I only use the 1000W XOC vBIOS for everything. I will have to ask Vince if he has one for the HC minus the adjustable bar support. (I cannot use adjustable bar support. That vBIOS will not boot with Legacy BIOS and I am not going to give up running Windows 7 for something as unimportant as adjustable bar support.)

 

I have some questions prior to trying to overclock the [desktop] i9-10850K this laptop, Sager NP9672M-G0 (X170KM-G), that I recently purchased directly from Sager with the below specifications.

Sager NP9672M-G0 (X170KM-G) Specifications:
- [DESKTOP] 10th Generation Intel Core i9-10850K Processor (10 Cores, 20MB Smart Cache, 3.60GHz Base / 5.20GHz Max)
- 17.3" Full HD 300Hz, Wide View Angle 72% NTSC Matte Display with G-SYNC Technology (1920 x 1080)
- NVIDIA GeForce RTX 3080 GPU with 16GB GDDR6 Video Memory
- Thermal Grizzly Kryonaut Thermal Compound - CPU + GPU
- Windows® 10 Pro 64-Bit Edition Preinstalled
- 32GB Dual Channel DDR4 SDRAM at 3200MHz - 2 X 16GB
- 500GB WD Blue SN550 M.2 PCIe NVMe SSD
- Intel Wi-Fi 6 AX201 M.2 AX Wireless LAN + Bluetooth Module

Specifically, I am trying to figure out if this laptop is not performing to the specifications that it should be with respect to the CPU performance. It is my expectation that since the i9-10850K is rated: from 3.6 GHz/3.6 GHz (single core/all core stock base speed) to 5.2 GHz/4.7 GHz (single core/all core stock max turbo speed); than with the fans at 100% speed along with the Thermal Grizzly Kryonaut Themal Compund (TGKTC) applied then it should be able to at least be able: 1) hit and 2) remain stable (without thermal throttling or power throttling); at the advertised 5.2 GHz/4.7 GHz (single core/all core stock max turbo speed).

BUT, I can't even HIT that stock max turbo speed 5.2 GHz/4.7 GHz (single core/all core) with 100% fan speed and the TGKTC applied let alone be stable at it!

The closest I get is hitting 4.9 GHz/4.68 GHz (single/all core) which I can only HIT for 1-3 minutes before thermal throttling kicks in at around 90-99 C, so I am not even stable at that speed.
For reference, my laptop comes with the "Control Center" software where that software has power mode preset options of which there are four where my laptop is running on the highest power preset, "Performance Mode" preset, which sets the power limits (PLs) to: PL 1 set to 130 W and PL 2 set to 251 W.

I read the post from Mr. Fox (#11505) which provided some helpful insight in to how Intel handles speed/throttling vs AMD, but I am not sure what I can do here to correct this i9-10850K to hit and be be stable at single/all-core stock max turbo speed.

So, since there are some experts here in overclocking; I figured I'd ask them for their assessment if I have something wrong with my laptop?

Intel Core i9-10850K Specifications By Core:


Intel Core i9-10850K Specifications from Intel Directly:

 

Intel needed do this because of today’s notebooks doesn’t offer good enough MB. Everything has to be cheap and quality becomes second. Tomorrow’s chips is notebook chips ported into the desktop world. OEM and notebooks is what comes first before DIY/Retail.

 

And, the OEM turdboxes and turdbooks often have little or no customization available. Like the old Dell/Alienware Level 1, 2 and 3 "overclock" garbage. Only respectable with a BIOS mod, but even then nothing to write home about.

 

This is the primary issue. It comes down to the silicon lottery and implementing extra cooling methods to remove (or greatly lessen) thermal throttling along with being realistic with clocks achieved in a laptop vs desktop cooling.

@Donald@zTecpc and @Mr. Fox created a custom cooling solution including binning chips suitable for their units to provide superior thermals paired with @Prema BIOS for complete control and fine tuning

The 10850k is a poorly binned 10900k so that's strike one. A desktop is much more forgiving with potentially hot/leaky chips or not so great samples.
You would most likely have to get under the hood and see how your pairing is (along with the thermal compound being used). Since it looks like you ordered a stock Sager unit, it could probably use some help.

 

@electrosoft, thank you so much for the reply. I appreciate getting the help in this thread as my laptop technically has a desktop cpu (i9-10850K) which I am looking in to overclocking, but I would want to research first why this laptop with this i9-10850K isn't even hitting stock max turbo speed, and it seems a lot of the experts seem to post in this thread.

After I initially posted I did take some time, to search within this thread for the key term "10850K" and read up on alot of the users' perspectives on it; it seems there is a consensus that the chip quality (bin quality I guess?), is not as stellar as the 10900K. Specifically it seems on a statistical average, the 10850K has slightly higher temps than the 10900K for a slightly lower speed.

If you don't mind me asking is there a link to where I can read about the custom cooling solutions by @Donald@zTecpc and @Mr. Fox?

I actually paid Sager's additional fee for their Thermal Grizzly Kryonaut Themal Compund (TGKTC) application, so I am rather surprised to have these issues with it; even though I paid the extra for TGKTC application - do you still suggest I get under the hood to see what thermal paste was used and how it was applied?

 

I don't remember if I posted pics of the Ryzen 9 setup. Not remarkably different than before. Only difference is mobo and CPU. I took the photo with the side panel removed because the glass tint is too dark and reflective for photos with the side panel installed. I wish the tempered glass was not tinted.

 

Stupidity everywhere. Some can’t make it right. Happy I’m not that stupid engineer. Why make it easier for people? … http://forum.notebookreview.com/thr...-20-more-bother.767713/page-718#post-11111353

 

I would suggest you look at how the thermal paste was applied. I would never pay extra for better thermal paste from the shop because this is something we can do ourselves for much less than the shop charges for it. Additionally, Kryonaut is not a good paste for this laptop since it burns up if regularly exposed to temperatures greater than 80°C. They should've applied Phobya Nanogrease Extreme Instead.

Next time, just get the stock paste, clean it off, and repaste with your thermal paste of preference. Mine is Phobya Nanogrease Extreme since it appears to work well with both desktops and laptops, and it yields slightly lower temps than Kryonaut in my experience. Furthermore, there are no thermal degradation issues. The only downsides are that this is an expensive paste, and it's hard or impossible to find in certain regions.

The contact pressure on the stock heastink isn't very great, but you can alleviate this by getting longer screws for the CPU side of the heatsink. I believe @JCordero31 performed this mod, so he should be able to tell you more about it. I ordered the zTecpc version, so I get their thermal mods from Mr. Fox, the details of which are currently not released to the public.

Another factor as Electrosoft mentioned is that your CPU may be a dud. However, this should only affect overclocking headroom in most cases. My guess is that there is some stupid power limit still kicking in despite you having raised them. The X170SM-G had a problem where the stock BIOS would power throttle the CPU if it wanted to draw more than 150 watts no matter how high you set the power limits. How much power is your CPU drawing before the clock speed won't boost higher?

 

As far as I know all stock Clevo firmware (including models in the generations before the X170) have silly power throttling nonsense and the only way I know of to correct it is with a @Prema BIOS mod.

 

This looks amazing! I really need to go vertical on my 3090KP HC. Unfortunately, my bottom retofitted 360mm radiator won't allow me to do this.

I also swapped my inlet line to the right port so water flows in according to the arrows on the KP HC water block. I immediately noticed no change at all lol. Just like you said! But then I happend to open up GPU-Z to check out temps etc. And I saw my GPU hot spot is about like 6C cooler under load, and roughly 3-4C cooler at idle. Not really sure how this could affect the hotspot temps while the actual load GPU temps are the same? Did you notice this too?

I'm still testing it. This block really isn't all that bad. It seems to do great on the memory temps, and the power temps and things like that.

There seems to be a few people in the wild who have gone crazy on these blocks, and managed 10C deltas at extremely high power limits.

The one thing that is hurting the KP HC water block the most is the thermal pad density, thickness, and thermal ability in general too.

Not only are these thermal pads 2.25MM thick around the GPU die, but they are also 50 shore scale hardness. And at 2.25mm thick and being this firm that's not gonna offer a huge amount of squish to offer that really good block to GPU die contact. These pads are capable of only 5/Kwm so nothing too crazy. The screws arounds the GPU die will bottom out inside of the standoffs.This prevents board flex, I think EVGA did this intentionally.

 

That's the hallmark of silicon quality. Worse binned chips usually run hotter and perform worse but still within spec.

The customized cooling is all ZtecPC collab, but there are still steps you can take. Use something like Cinebench R23 to set a baseline from 4.5->5.0 all core and log your results. Then elevate your unit for better air flow, repeat. Remove the bottom, repeat so you know where you stand in a climate controlled room. It depends on what software you're using too. If stress tests overheat your unit but your day to day software runs fine then is it worth it?

As for Kryonaut, it is.....ok......but for laptops that don't have the pairing pressure of desktop mounts, they are prone to pump outs and subpar contact issues. I usually use nanogrease extreme as it is very thick and can compensate for any gaps till I certify my pairing. If/when you know your pairing is optimal, you can then explore other thermal compounds or liquid metal.

You can kinda use XTU, but as @Mr. Fox said, PremaBIOS is what really opens the BIOS up to let you fine tune your settings, but at least you can get under the hood to verify the thermal application and most likely reapply with another compound like nanogrease.

 

Was evaluating a Zalman Z3 iceberg "Premium" (ok) case that was in all white. I normally am a black / grey / gun metal kind of case lover, but I went ahead and built out of it with components I have lying around and I gotta say.....I don't hate it.

(I like how off the tempered glass bezel you can see the reflection of the KPE 3090 OLED.....like he knows his time might be up )