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

Do it! Join me on this journey lol.

https://www.overclock.net/attachments/bf1c8246-bb41-4763-9e6e-3a822da79ce7-jpeg.2530217/

I need that Samsung backed GSkill ASAP. I grabbed my Dominators as a stop gap until these are widely available or they get the speed up a bit more.

 

See it as a invite http://forum.notebookreview.com/thr...ers-welcome-too.810490/page-607#post-10924726

@Mr. Fox @Johnksss

 

We would love to have you. I will look for the password and send it to you.

Edit: PM sent

Anyone else that wants to join, don't hesitate to say so.

 

Best GPU I ever owned. Zero complaints (except the price of course).
Last GPU I enjoyed as much was the Kingpin 1080ti
I skipped the KP 2080ti because I thought it was too expensive (Irony!)

Since they announced the Z690 Dark right at launch, it shouldn't be the long wait like the Z590 Dark. I enjoyed playing with 11th, but testing 5+ chips and all but one and returning one DOA wasn't exactly light on my pocket. I'm going to sit back and wait for the dust to settle while performance and benchmarks make their way out and we see how it performs on W10 and W11 outside of the hype bubble. I'm expecting a hot and heavy chip with good performance with the 12900k.

Only problem potentially with the 12700k/kf is another 11700k situation where it is weaker bins with 4 e-cores purposely turned off that don't OC nearly as well.

Wife has a socket 1200 system, my desktop is socket 1200 and now my laptop is socket 1200. I'll be staying on the Socket 1200 train till at least January.

 

I like what I'm seeing with Alder Lake thus far, but the DRM compatibility problems are giving me pause. This is certainly not Intel's fault, and they've been working with providers to get stuff working again. The concern is still there, though, plus I want to see what AMD has in store for Ryzen 6000. I've been doing most of my gaming at 4K so I've not seen much reason to move on from my 9700K.

Since I saw it mentioned, my favorite graphics cards that I've owned are:

1) Visiontek GeForce 3 - Back when I was more into modding and overclocking, I did that to this card and got some solid results.
2) Creative 3dfx Voodoo Banshee - My first proper 3D accelerator, purchased with the first paycheck from the job I worked in college.
3) Sapphire Radeon 9500 - Soft-modded and OC'ed to 9700 Pro levels. The last genuinely great ATI/AMD GPU until my 2011 MacBook Pro.

As much as I like my current 3080 Ti, I know I won't have as fond a memory of it as I do these older cards.

 

12900K has landed. At least the box is pretty. Have to wait for the motherboard and ram unfortunately so no testing yet.

https://imgur.com/a/ClrFBjW

 

I am glad the 5950X should be back in my system later today so I can put the pathetic 2600X back on the shelf. I am eager to see how it works with the delid. I hope it brings about a meaningful improvement.

 

I hope it brings meaningful improvement too and it may lead me down that road as well if they are still delidding them. Did they put LM on it? I am surprised you've allowed them to reseal the IHS.

 

Yes, they put Liquid Pro on it per my request. I thought about leaving it unsealed, but I did not want to risk anything happening being shipped with the IHS detached, like smacking one of the dies or SMDs next to the die and messing something up. They use SuperGlue gel to mount the IHS, which is fine as long as you soak the CPU assembly in acetone for 10 of 15 minutes before trying to remove it. The IHS will literally fall off once the acetone melts the SuperGlue. If you try to delid it by force with SuperGlue attaching the IHS it can actually tear the outer silicon surface layer off of the PCB and ruin the CPU.

I am going to see if I can figure out a way to run it bare die. I may look into modifying the OptimusPC block by having the cold plate machined down on the area that overhangs the IHS dimension to see if that allows direct cold plate contact with the dies without making contact with the chintzy plastic housing part of the socket ILM. Whether that will work or not depends on how much taller the ILM is and if the cold plate can be machined down enough to provide clearance.

I am actually surprized that nobody has already developed a water block with a cold plate narrow enough to cover only the CPU and not overhang the air space above the ILM.

This might even move the Optimus waterblock enough laterally to provide adequate ILM clearance with the IHS removed. http://der8auer.com/ryzen-3000-oc-bracket/

 

Intel could easily make a 10+8 if they reduced the profits. The Alder lake die is around 35% smaller than for Rocket. But I expect they tryed to do as little as possible to come on par with the best of Ryzen 5000 series Cpu's. Use 8 P cores all over (forget the small E-cores) for i7 and i9 is Rocket lake up again. Profits beats max/better performance.

 

I'm not going to say much on Alder Lake as I don't know too much about it, but as much as I dislike the idea of the E-cores, it's actually kind of good from my perspective.

For example, myself and many others are on Haswell CPU's, and the E-cores have been stated as basically being Skylake cores, so if I were to go out and buy a 12900k right now, it would be as if I'm using "16" cores as I'm coming from a CPU that performs worse than a Skylake chip. I'm not really good at explaining what I just said, but it makes sense in my mind, so I apologize if my rationale sounds like the dumbest thing ever.

 

I understand what you say, but Alder lake will never ever be a real 16 cores Cpu. Even the bot had to make changes due the use of two different type cores.


As a fotnote... The E-cores have IPC as 10 gen Intel chips. No HT and come with castrated clock speed to be able to make it more power efficient.

The 12th gen Hybrid chips line-up should rather look like this to make an difference...

i9 - 10c+8c (28 threads)
i7 - 8c+8c (24 threads)
i5 - 6c+4c (16 threads)

With 2 more P-cores for 12900K we would see +12600 cb in CBR20 and a healthy lead over Ryzen 5950X.

This is what we got.

 

But what would they launch as a 13900K Q4 2022?

 

More E-cores And higher Clock speed bump as the usual step with Skylake gen chips

Edit. See also the rumors. More E-cores, LOOL https://www.digitaltrends.com/computing/intel-raptor-lake-challenges-amd-apple/

The top-end Core i9 configuration will feature eight Raptor Cove cores and 16 Gracemont cores that equal 24 cores and 32 threads. This would represent a dramatic doubling of “efficiency” cores, considering the current rumors surrounding an eight-core/eight-core configuration in Alder Lake. If the leak turns out to be correct, that would mean Intel wouldn’t succeed Gracemont cores with something, but instead would just increase the amount baby cores. Performance-wise, the boost clock features a 200MHz increase.

Yep, this is awful prediction/rumors if true. I rather expect 2 more P-cores and maybe more E-cores with a speed-bump and IPC on top. This depending on AMD.

 

Here's hoping for a beastly bin! By this time next week p00f! the 11th gen forums over on overclock will suddenly became barren....

Definitely looking forward to your results. You are poking and prodding the 5950x for all its worth.

Finally unfurled the X170SM-G and taking a look around...

@Mr. Fox what a great way to implement a tension system that is the equivalent of eating your own chow for the CPU and GPU.

 

I think Intel plan is like this, they wanted to create a new Chipset and embark on that DDR5 standard and execute the same on the Xeon and LGA1700 + BGA trash as well.

To expand more ....

First looking at the Baby core BS. We already knew, 10nmESF cannot clock all these real 14C28T at 5.1GHz so they added this nonsensical stuff and bloated the benches (Remember there are Class 0, 1, 2, 3 for the Intel ITD, that's the core part of this new CPU design without the ITD the CPU cannot exist in x86 world) as a bonus they get the Apple Lapjoke OEM marketing fluff with more cores and super efficiency thanks to Intel's 10nmESF.

Second Foundry and I/O Since AMD has TSMC problems and cannot cope with the Intel's Foundry power (no matter what Intel's Foundries smash TSMC output by exponential factor) so they push out the ADL now and get all that new Intel shiny marketing PR with E/P cores + DDR5 + ST performance for games + PCIe5.0 all OEMs on board since it's Intel so win-win. New kid is on the block and grab all sales until Zen 4 arrives, so milk the E cores now and they refresh it in Q3-Q4 2022. Exactly at the time of Zen 4 where AMD will also launch the DDR5 / PCIe5.0 (this PCIe is not confirmed, maybe the Zen 4 Raphael might not have but a Zen 4+ or etc, totally my own guess) so they duke out with one advantage, platform maturity and marketshare.

Third, Datacenter - Xeon SPR will launch in 2022, on 10nmESF with PCIe5.0 and DDR5 taking all those HUGE orders of Intel's real deal of money minting, Datacenter. Thus edging out AMD's EPYC, basically TSMC is the Achilles heel of AMD thanks to crApple junk for social media toys and luxury BGA products choking even more, TSMC 5N is already ready and only Apple is making them, M1Pro/Max, A15.

Basically with these 3 steps Intel got -- Client sales from E/P BS on Lapjoke BGA Turdbook market AND Win11 M$ Mafia deal + Desktop LGA1700 sales trying to take back the AMD marketshare and mindshare + SPR XEON over EPYC-X (Milan 3D-V Cache) I bet this is why AMD stopped Threadripper 5000, they will eat up into EPYC bins as TR die's are top of the stack for Client.

Now for the big part, performance lol, these are baby cores no matter what as you said bro Papu are never going to be real cores. So Intel is not going to get all good pass against Ryzen and since 3D-V Cache refresh on X570 comes and close in the short ST lead that Intel gets with ADL. Many people already who have machines and due to the dGPU BS Intel has to deal with all the competition pressure from AMD on the side, but these are small since BGA turds, Xeon sales trump more than LGA1700 so they went with this plan.

Now that's out. I hate how Intel is releasing so fast, this ADL will be another CPU like RKL which will be having shortest life (Don't forget CML 10th gen, it's still super competitive and will be just like 9900K I still wish Intel released damn 10900K at 10nm on PCIe4.0 instead of RKL 14nm backport), on top we do not know how long the chipset will work plus will there be another damn Z790 or not when AM5 hits or what will Raptor Lake have DDR4 or DDR5 only...get ready to be screwed if bought DDR4 Z690 basically AND why refresh the Desktop parts so fast, AMD's Zen 3 is going to be in the market for more than a year just like GPUs it makes life easier, not like smartphone BS where people shell money for every year mediocrity.

 

I won't know until I have time to fart around with it and see if there is anything I can do to correct it, but it seems that either the liquid metal was not correctly applied or delidding has ruined my 5950X. Temperatures are absolutely horrible. At least 20°C hotter than before delid. I don't know how the test results James sent me were not more horrible than they appeared to be. I thought they didn't look right as soon as I saw them. They didn't send me back the stock IHS so I had to email and ask him to mail it to me. It may be the copper IHS doesn't fit properly. Hopefully, they didn't do something regrettable, like throw away the stock IHS. I wondered if maybe a change in z-height made a fit problem for the monoblock, so I put back the stock VRM cooler and switched back to the Optimus CPU block and it made no difference whatsoever. Equally horrible. Note what an utterly absurd variance between core temperatures it has now, and how much severely hotter CCD1 is compared to CCD2. It was only a few degrees variance between cores and CCDs before it was delidded. That is the only thing that gives me some level of hope that incorrect liquid metal application is the reason for it. I'm not very happy right now.

 

Outch. How is the coolest cores vs. your previous tests? And rip off that lid and do some pressure testing between lid and dies.

 

Was their testing done before they resealed it? Could be the culprit.

 

Ouch.....

Sometimes we get so used to Intel being subpar with their IHS and STIM/TIM application that maybe AMD is more exacting than we realized? Some more testing and adjustments will see where it takes you but these initial results are pretty painful. Hopefully it comes down to a liquid metal issue.

 

Well, I wish I could say that I think it is a good product but I believe the exact opposite to be true. I am not sure that "exacting" is accurate because the product has plenty of engineering defects and undesirable limitations even when it hasn't been delidded.

The delid process basically ruined it rather than improving its flaws. I disassembled and cleaned it up, then reapplied new liquid metal and the results are the same. I am not sure what kind of glue they used. It was not SuperGlue and it took several hours of soaking in acetone and razor blade nudging to get the IHS off.

The CPU is no longer usable unless I run it stock and live with high temperatures. I don't know whether to blow $700 on another POS 5750X and live with the platforms glaring defects or just be done with AMD and go back to hating everything about them again like I have for the past decade. Like all the rest of my experiences with their trash, this has been truly regrettable by nearly every measurement. I honestly don't understand what the fanboys find great about Ryzen.

I just cancelled the OptimusPC GPU block. I am not sure what I am going to do at this point. Maybe I will just part it out and be done with all modern tech trash. I'll sleep on it a few times and decide.

Just looking how this abortion is designed under the hood should have been a clue to me that it was trash. Very expensive demonstration of poor judgment on my part even considering it.

 

Here's a volcanic CPU-Z benchmark with BIOS defaults. Nice, huh?

Maybe the Ryzen delid service being taken down is a result of how "successful" the idea was with this $750 piece of crap.

 

Unfortunate to hear that really.

I'm no expert or even have any intermediate level of information in this but here's what I think. Looking at Intel's monolithic application of the LM/Indium with gold coating they are having varying thicknesses across the board since 9th and 10th,11th gens now 12th but since it's a single silicon block the method is straightforward and applies uniform pressure on all of the die area.

Whereas with Ryzen MCM design, the CCDs and the I/O blocks have a lot of distance between them plus different lithography with respect to heat density so AMD has to engineer a lot probably which makes them to apply STIM in a different way. Also note the LM coat on top of Gold thickness vs the IHS thickness plus the each silicon die thicknesses.

Maybe I think you should ask them the original IHS and measure both of these properly for any thickness differences of the IHS and as Papu suggested do a pressure test.

 

I suspect it's just a sloppy design where each of the dies have minor variance in z-height and AMD filled it up with solder to even things out. Liquid metal can't deal with fit problems like that. It might even run a little less hot with normal thermal paste under the IHS if that is the problem. There is so much crap around the dies that the idea of using liquid metal is pretty risky when you stop and look at it. I am waiting for Rockit Cool to reply to my email and confirm they are sending back the stock IHS and haven't done something silly like thrown it away already.

 

I expect they use the product they sell https://rockitcool.myshopify.com/collections/clean-and-prep/products/loctite-super-glue-gel-control

They see more cores at ok price point. But not for the 5000 series if you want their +12 cores chips. And many of the Ryzen boys run it stock. Even with all problems for the platform they defend the products.

Have you tested with ICD? This to see if the new copper Lid is the problem.

More Golds.
https://hwbot.org/submission/484313...lti_core_with_benchmate_core_i3_3120m_1267_cb


https://hwbot.org/submission/484313...ngle_core_with_benchmate_core_i3_3120m_499_cb

 

damn brother hate this has happened. I’ll sell you a brand new sealed z590 dark and sealed 11900k for $1k lol

 

That is very tempting at this point. If you're not in a rush, let me think about it a day or two. That might be better than wasting $750 on another 5950X. I can sell this mobo, EK monoblock, OptimusPC CPU block and 2600X to help offset the cost.

 

Nah no hurry. It’s been sitting here on my desk for 2 months tempting me. Don’t have the time right now and it’ll be probably a few more months before I’m ready to build anything. Hope the amd works out for you though, hate when this crap happens!

 

WoW.....

wow wow wow

I am so sorry it went from bad to confirmed bad brother.

If you're going to change directions, I would spend the next few weeks trying to minimize the furnace you have right now and figure out if there is a foxfix or work around or solution to get back to where you were at least if not better and wait for Alder lake results to drop before making any type of in the moment, clearly perturbed, "into the bonfire you go!" frame of mind.

I do know you would have a blast with 11th gen as it is made for overclocking and the cooling is better. Alder lake leaks seem to indicate it responds well too but we will see.

That's a good call to sleep on it too and start stemming the flow atm.

 

I've finally had a chance to play around meaningfully with the X170SM-G:

Memory:

Using my last pair (2x8GB) of Corsair C18 3800 sticks confirmed and tested on my P870TM1 @ 3466:

Memory slots under the keyboard are problematic >2933 on their own to the point the system will sometimes hang on reboot/boot.

Memory slots on the bottom isolated work fine up to 3200 but I started getting errors in TM5 @ 3333 with known good timings on my pair of 3800 sticks. I'll go in and work with VCCIO/SA later. Upping the Vmem past 1.35 actually produces more errors and the memory slots being RIGHT NEXT to the CPU and heatsink wasn't a good call for heat.

I have a pair of these on order because they have great timings at 3200 and dual rank: https://www.amazon.com/gp/product/B08W224DXL/ref=ppx_yo_dt_b_asin_title_o00_s00?ie=UTF8&psc=1

CPU:

No throttling at the stock settings 4.9 all core under CB23 with the 10900k even pulling 194w. Good start but see if I can shave down that 94c. I'll get in there over the next week or so and go over everything meticulously as always.

GPU:

Runs pretty much cool as can be. Time for some OC magic. Ill get in there over the next week or so.

 

As I am crunching the numbers, I think I am going to most likely take up brother @pathfindercod on his offer. I am going to try to get the temps down to where the CPU is usable, but I am thinking that now is the time to kick AMD to the curb. I can spend $750 for another 5950X and hope/pray that it is not garbage and doesn't have USB drop-out and WHEA errors like many do. Or, I can sell the mobo, monoblock, this CPU (with disclosure, of course), the Optimus Signature block and 2600X and cover probably close to half that total cost. I just have to figure out how I am going to pay the $1K in a way that works. The cost is not an issue, but how to put the money in his hands (in one chunk) while allowing me to float a couple of payments while I sell these parts and pay off the balance is what I need to figure out. I have used Affirm this way in the past, but I don't think I can use that option for a private party purchase.

 

Tertiary timings might need to be loosened up a bit, think they are more sensitive and CPU/IMC dependant?

 

This is really the Ultimate combo for pushing the XOC 3090 Kingpin Hydro Copper.


PS: Even my Debug LED is coming back
to life! Which it has been broken for a couple weeks now, and showed no signs of ever working properly. It has somehow managed to sort its self.

$1,000 is an amazing deal, I love mine!

 

Brother @pathfindercod and I am working on the details to see if I can make it work for both of us. It's not going to do anything for me on HWBOT (and may actually lose ground) but at least it will be fun to play with, reliable and not be an AMD hot mess that is riddled with random the malfunction and bugs they are famous for providing, LOL.

 

I am not sure and they are not great at communicating in real time. The initial report on the temps before they shipped it seemed problematic to me at first blush. I was instantly worried before it even left their shop.

In follow up to your thoughts (via rep comments) about lapping, all of the components surrounding the 3 dies are just barely shorter than the top of the dies and I fear that doing that would be the final nail in the coffin for it because those components would then be affected. There is also no air space between the copper IHS and the PCB. So, if I shorten the z-height I think it will produce an air gap that would require shaving on the IHS as well, so like a vicious circle of chasing one problem while simultaneously creating a new one.

I am going to tinker a little more and see if I can at least make it sellable with temps that are fine without overclocking. Maybe fill up the interior of the IHS with normal thermal paste or something like that and see if the temps improve or worsens things. If I can get the temps to be decent stock, some people don't care about overclocking.

 

More gold for the team boys

I'm going to do the CBR11.5 Benchmate record later this weekend as I'll have more time. The Benchmate record may be a bit harder and require some more tweaking to beat, but I think I'll be able to get it. If thermals are the issue, I've always got the nighttime to eliminate that issue

 

I would take a glass PC panel and cover it with sand paper, then get it wet with water and lap the bottom of the IHS to make it shorter. Then you’d have that air gap between the substrate and the IHS edges. This should provide better contact with the die, and the inside of the IHS.

By lapping, I am referring to this part of the IHS that is marked in RED.

Their Copper IHS may be too thick creating a gap on the inside.

Or pack it with quality thermal paste on the inside. For normal stock operation only.

 

I might give that a try. It could work. It might even be a flaw in the design of the IHS. I am almost at the point of not caring any more. If I can work something out with @pathfindercod I may just blow it off and sell the other parts. I just want to be done with AMD's trash at this point. Shame on me for even going there. I feel very stupid for not listening to my inner voice, LOL. I never should have given Team Red the time of day. An expensive lesson learned, and one I hopefully won't forget. The lessons of the past were also expensive and my real fault was being optimistic that they don't suck as bad as they used to.

 

I checked my notebook where I jotted down all my timings and used those but that sounds about right and will be the next step to go in and start to loosen up everything but hopefully I won't reach a point where very loose timings negate any frequency gains but we'll see. All I know is at best I will be running 2x8GB on the C18s for >2933 timings so I ordered the other 2x16GB sticks to give them a try @3200 as the timings were actually decent and dual rank for a little extra punch.

 

Well, if you're going to do it, you might as well join the team....

 

I think that is what is going to go down. I put the 2600X turd CPU back in the Crosshair.

I may just grab a low buck PSU and remove the 3090 and sell the rest as a "bring your own GPU" package on Craigslist. I've got extra RAM, even an extra PSU that I am using to run all 3 pumps and 18 fans off the Lamptron CF525 hub. I have a Corsair H100i collecting dust that I can slap on the CPU. It would make a fine gaming system that looks great for a kid without a lot of money, or an adult that doesn't care about face-melting performance.

Then put my current RAM, PSU and new parts on the Praxis Wetbench. We'll see... still need to sleep on it.

 

Money sent. Thanks, brother.

 

Got it brother Fox! Sent you email with shipping schedule etc.

 

That's kind of what I'm going to do. When I get around to installing this Z590 Dark, I'm going to take the GigaFail motherboard and build out in my wife's Corsair case from earlier this year since I have parts galore laying all over the place and list it. I did it with her last collection of parts too and fused them with some I had on the shelf.

 

You’ve got one too? When did you get a Z590 Dark? Dang!! Nice lol. And here I thought I was the only one running one on here.

And nice deal you got on that 3080 Strix!

 

That’s awesome!

PS: I’d go straight for 1.550V on the VCCIO voltage, and VCCIO AUX. I’ve been running that daily for over a month now. No issues at all. I run 3,882Mhz Gear 1 CL14-14-34.

You might just get lucky and get 4,000 Gear (1)

 

You're scaring me already, LOL. Coming from the Z490 Dark and running 4600-4800 CL17, "maybe" is not what I wanted to hear on the heels of this AMD garbage nightmare.

Is the voltage that high on VCCIO and VCCIO Aux normal? That's way higher than what you and I ran with the 10900K on the Z490 Dark.

Hopefully, OptimusPC will process my cancellation/refund on the Kingpin GPU block. That, together with what I get for the AMD parts, should pretty well cover this with maybe enough to also cover the OptimusPC CPU block that I just ordered so I would have a way of cooling the 11900K when it gets here.

The good thing is that if the 11900K isn't enough CPU to make me happy, my Strix Z490-H supports it and the Z590 Dark supports the 10900KF I am currently running in the Strix. So, I've got options.

 

(I was leery at first too lol, especially since the 10900K died on the Z490 Dark) but, I don’t think that was what killed that last one 10900K. I think it was just a coincidence. (CPU’s die sometimes)


I can confirm it has not hurt my CPU at all. I will go back and forward gear 1 ultra low latency to gear 2 high speeds low latency.

 

I'm going to stick with Socket 1200 till at least January (or beyond) and this GigaFail motherboard is going into a sell build while the dust settles on Alder Lake and I wait to see the AMD refresh. Somehow the entire abode is now Socket 1200.

I'll give the Strix 3080 a whirl in the next few days and see if I'm going to jettison the 3070ti.

New 16GB sticks for the X170SM-G will be here tomorrow from da 'zon but I'm going to finish relaxing the terts and bump IO/SA since I know the sticks work fine at the speeds and settings tried all the way up to 3466 so it's the system limitations.

 

Man I really wish the Clevo X170SM-G had a giant unified vapor chamber heatsink. The CPU really needs a vapor chamber on top of it rather than just a coldplate in order to really spread its wings.

I am able to do 5.3 GHz all core in this laptop for a short time before thermal throttling kicks in. I can only do this for short bursts. I think I'll need to wait till a cold day in December to see just how long I can sustain this in cold weather. I want to push higher, but this laptop's cooling system can't really handle sustained loads above 5.1 GHz all core, which is still incredibly impressive. I'm also impressed I can do 5.3 GHz all core in this laptop at all, even if it's only for a short burst.

At 5.3 GHz all core, my 10900K requires 1.48v and it draws about 275 watts of power. In terms of chip quality, is this bad, average, or good?

I'm also surprised at how little power the 10900K draws while idle. It only draws 5-7 watts of power while idle, which is the least amount of power I've ever seen a DTR's CPU draw while idle. This is also without any undervolting, so I could probably push the power draw even lower. I could never get my Alien's CPU to draw less than 9 watts while idle, even with undervolting.

Since I've only been playing with pushing performance, now it's time to go the other direction and push efficiency. Time to see how low I can drop the voltage and power consumption. That'll allow me to establish daily driver voltages and maximize this CPU's lifespan.

The 10900K is a very fun CPU to play around with!