[eighty20]

@eighty20
Why using this awfull image hosting? I can't see any of your screenshots and even my "AdBlocker Ultimate" think that it's trash, lol
There's more adblock-friendly imgbb.com (example below)
No stupid frames and galleries, just copy direct link and paste it via forum image link tool, if you absolutely need to use these adware-bloated image hosters.
Although, I think the forum tool is much more convenient

Well i didnt want to flood 1 post w too many images but sure will fix it later.

And it works just fine on my phone, using firefox w ublock origin.

 

[PJVol]

Well i didnt want to flood 1 post w too many images

You don't have to as there's a spoiler tag for this purpose.
Besides, on the PC side firefox ublock doesn't work as good as before.

 

[eighty20]

You don't have to as there's a spoiler tag for this purpose.
Besides, on the PC side firefox ublock doesn't work as good as before.

Ok will fix it later.
Mine still works just fine on both pc n phone. I have most traffic rdr to adguard home on opnsense then back to Zenarmor and unbound DoT using private dns as upstream so that maybe the reason i didn't notice any bloatware on..pretty much any pages i visit for awhile.

 

[eighty20]

Oh..Today I got a new board to play w for a couple hours. Turns out my cpu can do 106bclk on a MEG x570 unify, didnt touch anything besides changing bclk. Able to play some matches n crashed but yeah, it's kinda cool seeing how far I can boost this locked cpu.

Is it where eclk is?

 

[kawzir]

Hello, my CPU fan will stop spinning once every 2-3 days, until the CPU overheat and shutdown. Can it be related to unstable OC? But my PC is stable when I game hours on 90 degree CPU temp, 60 degree ram temp. Also pass Extreme1 TestMem5 with 3 cycles. It actually happened more when the PC is on low-load, just browsing internet or watching youtube.

It's an AMD Wraith Prism came with 2700X and it worked fine for 4 years. Recently I switched to 5800X3D and still using the Prism. It happened while gaming, idle or just watching Youtube.

Things I did: PBO tuner -30 all cores, PPT TDC EDC @ 100 70 100, ram runs 1.35V@3200CL14 (XMP profile but tighten timings a bit), global c-state on

 

[bmtmis]

Anyone have an issue with Win11 task manager only showing base clock speed of 3.4 GHz? It's running fine as represented properly with HWInfo etc but am seeing this after I dialed in all my bios settings prior to a fresh install.

 

[PJVol]

Right away, the main issue is Win11 itself.

 

[T[]RK]

the main issue is Win11 itself.

It is true. It's have it's own strange settings which look's like... not finished or heavy protected by someone in Microsoft. It's intentionally don't allow to do things... enable settings, disable them. It may show setting, but don't allow to change it, or allow change it, but not here, but there instead. And if it even enabled - don't show it.

I have to check again Windows Security settings after Cinebench R20 benchmark run (on my system it look a bit lower). And by checking i mean look in RegEdit and verify that DWORD values are correct and same values are correct in Group Policy (GPEdit.msc). So i have to:

1. Enable back Virtualization;
2. Visually check in Windows Security that:
2.1. Memory Integrity is disabled (and Verify this in RegEdit);
2.2. Kernel-mode Hardware-enforced Stack Protection is disabled (and verify this in RegEdit);
2.3. Local Security Authority (LSA) Protection is disabled (but since Microsoft is so protective - they HIDE it and i have to verify this in both: RegEdit AND Group Policy);
3. Enable (!) LSA again to correctly disable it later (to verify DWORD values in RegEdit);
3.1. Verify in Event viewer (as Microsoft suggest!) that LSA is enabled (it was);
3.2. Verify in RegEdit that LSA got both DWORD with same value (IT WASN'T! Fixed it);
4. Runs few tests with Virtualization Enabled + LSA Enabled;
5. Disable LSA (verify it in Event viewer, Group Policy and RegEdit. It forget to set one value AGAIN! Fixed it!);
6. Run few test with Virtualization Enabled + LSA Disabled;
7. Finally Disable Virtualization;
8. Run few tests with Virtualization Disable + LSA Disabled.

Results:

First i run benchmark in CPU-Z (Single Thread): 628

Actually 628 for both Single and Multi thread. Before tuning value might go as low as 626, 625. Sometime i see how it go up to 630, but then stay to 627-628. Probably hit limits and roll back.

If run benchmark in default mode results will be next:

Single thread: 627,5
Multi thread: 6296.7

Cinebench R20:

CPU: 5684 pts
CPU (Single Core): 583 pts

Before that as you can see there was result lower on exactly 100 pts.

I also run benchmark (multi) with HWiNFO64 at background and rusult was exatcly the same 5684 pts!

I check how stable frequency with Boost Tester and i can see result there too - almost on each core 4849 MHz effective clock!

So, there really was some kinde of tiny interference in system.

I also got less points in Cinebench R23 and 3DMark CPU Profile because Cores a bit hotter now. Results should be better with Curve Optimizer now. Didn't tested it yet.

 

[eighty20]

I noticed one thing abt 5800X3D or maybe AMD cpu overall. Since I have some thermal headroom at stock bclk -30CO typical stuff etc.. so i decided to increase offset +0.025 ~ 0.05vcore until all cores reach 4450MHz under ycruncher & linpackxtreme bm. However cb23 result is decreased a bit, weird. Some ppl said -25CO on both best cores yield better performance, would it be because there wasn't enough vcore, add a bit positive offset help to maintain 4450MHz under extreme load would be more ideal if thermal is not limited? Thoughts everyone?

@Blameless Do you think 1.313v observed in Hwinfo - CPU Core Voltage (SVI2 TFN) is a bit too much? I have tested 103.68bclk -25CO but needed +0.56vcore to get stable during corecycle prime95 AVX2 All. It only reaches more than 1.3V during ycruncher benchmark.

 

[Blameless]

@Blameless Do you think 1.313v observed in Hwinfo - CPU Core Voltage (SVI2 TFN) is a bit too much? I have tested 103.68bclk -25CO but needed +0.56vcore to get stable during corecycle prime95 AVX2 All. It only reaches more than 1.3V during ycruncher benchmark.

That's definitely on the high side for a 5800X3D. I wouldn't worry about it, unless you're using the LCLK bug to break temp limits.

 

[eighty20]

That's definitely on the high side for a 5800X3D. I wouldn't worry about it, unless you're using the LCLK bug to break temp limits.

Great. Thanks. LCLK disbled doesn't seem doing anything on my board so i have it enabled to get better score.

 

[eighty20]

lol Ycruncher crashed at 103.68bclk -2520CO +0.056vcore after adding a bit more liquid metal today. First time seeing a core temperature dropping this low, room temperature is ~16-17C. Can't wait to play w the AMD supercool direct die.

 

[T[]RK]

First time seeing a core temperature dropping this low

Only cores are main issue. So tiny and so high transistor density - cooler barely can hande heat. On my image you can see how small they are comparing to IHS.

Look on temperaturs of IOD or even L3 Cache. On my image L3 between cores array (two lines between L3).

https://www.overclock.net/threads/5800x3d-owners.1798046/page-387#post-29240664

 

[eighty20]

Only cores are main issue. So tiny and so high transistor density - cooler barely can hande heat. On my image you can see how small they are comparing to IHS.

Look on temperaturs of IOD or even L3 Cache. On my image L3 between cores array (two lines between L3).

https://www.overclock.net/threads/5800x3d-owners.1798046/page-387#post-29240664

Do you think the height of both die on 5800X3D are the same? I wanted to pop the lid off but I can't find any measurement besides some delid photos



I got a reply from the supercool guy. The block seems to be compatible with 5800X3D.

Anyway, i don't think I can get any colder without exotic stuff, gonna have to settle at 103.3-20CO all cores with offset +0.05vcore. Corecycler time

 

[T[]RK]

Do you think the height of both die on 5800X3D are the same?

It should be or it will not be compatible with IHS. AMD have to keep everything inside IHS exactly the same to be compatible with TIM and IHS. It was the reason they grounded down 3D V-Cache to same level as other silicon.

At the same time i agree, that information is... small. People which have got CPU - don't have electronic calipers or something similar. Even scaner to put delided CPU inside and scan it in high resolution.

I wanted to pop the lid off

Be VERY careful. I still looking all kind of information about process and i more like hot deliding (mainly because i have thermal camera for temperature control) instead of special tools like delider or vice.

P.S. I also saw funny method with iron. %) Fast and simple.

 

[eighty20]

Be VERY careful. I still looking all kind of information about process and i more like hot deliding (mainly because i have thermal camera for temperature control) instead of special tools like delider or vice.

P.S. I also saw funny method with iron. %) Fast and simple.

I don't know if I should do razor or iron method. Indium transfer heat pretty fast so I'm afraid too hot would kill the cpu, and I'm not sure if they use indium or something else nowadays. I probably gonna take my chance w the vise as usual then sand the plastic holder or whatever it called a bit then go direct die.

 

[Arlina]

Hi, I just bought a 5800x3d and boot windows everything ok but start linpack extreme 6gb it gives hwinfo whea errors at 1900 FCLK.

Can anyone give advice on how to make the 1900 FCLK stable?

SOC
VDDP
CCD
IOD

What needs to happen for it to be stable at 1900 FCLK?

 

[eighty20]

Hi, I just bought a 5800x3d and boot windows everything ok but start linpack extreme 6gb it gives hwinfo whea errors at 1900 FCLK.

Can anyone give advice on how to make the 1900 FCLK stable?

SOC
VDDP
CCD
IOD

What needs to happen for it to be stable at 1900 FCLK?

What i learned fm this thread is that VSOC > IOD > CCD, 50mv in between. Not sure abt VDDP but I have it at 0.9 for now.

This is mine currently, VDIMM 1.54v, auto LLC. Timing was the same from the one that passed a couple cycles of absolut ( forgot to including zentimings).

You can try with C16 flat first ( 16 16 16 16 32 48) 1.5V, GDM on 1T, tRFC just leave at 342 = 180ns. I'm using DR so not sure how it's like w SR for Rtt. For Setup and Drvstr just leave them auto. Also would be great if you give bit more detail of mb, ram etc..

 

[Blameless]

Do you think the height of both die on 5800X3D are the same?

The Vermeer-X CCD is exactly the same height as a standard Vermeer CCD and both are the same height as the IOD.

It's a common misconception that the V-cache die is some how dimensionally different, but this is false. ~95% of the thickness of a CPU die is empty silicon, so adding the cache die involves shaving off most of the die, attaching the cache stack, equalizing the z-height with oxide-bonded silicon support dies, then doing the same thing with a silicon cap to restore the standard height.

This is also why the 5800X3D runs hotter...there are two additional thermal interfaces inside the CCD between the cores and the IHS that the non v-cache parts do not have.

 

[eighty20]

Then it would be insane to lap the die since the silicon support is a lot thinner. Thanks @Blameless, I almost go w the idea of lapping the die a bit too. So most 5800X3D has the same layout w 5900x minus 1 CCD right? Do you know what underneath the glue/silicon in the photo, it's 5800X3D delid I found somewhere.