Overclock.net banner

1 - 20 of 41 Posts

·
Registered
Joined
·
59 Posts
Discussion Starter · #1 ·
Hi guys,

just want to share how I was able to get my FCLK to 1900 mhz 100% stable on a Crosshair VIII impact ( with G.SKILL 4000 mhz CL19 32GB ( 2x16 )

My system is 100% stable at 1866 FCLK and 7333 memory, but it would not boot at 1900.

So, at 1900 FCLK the board would not even post. With everything on auto except for memory at 3800 and FCLK at 1900. The one thing that made it post was to disable SB spread spectrum. That single setting that was on auto, was the only thing I needed for the board to POST. The system was almost stable. I never had a crash when running stress tests or gaming and the only thing that didn't work properly was that my oculus rift would lose tracking from time to time ( tracker is connected via USB 3.0 ). I confirmed this to be from the FCLK by going down to 1866 and the problem going away.

So, to fix the USB thing I simply changed the VDDSOC load line calibration from auto to 5 ( the maximum ). The USB problem went away.

I
After deliding the CPU and using bare core on block ( I am using a heavily modified enermax liqtech II 240 ) I wasn't able to go further than 1900 FCLK. Temps are 18 degrees lower after deliding and using the bare core on the block.

For testing stability I ran 12h memtest86 plus 12h of prime95 ( latest version, with the default setting for testing memory ) and again 12h of prime95 with the max heat output setting. After that, one week of regular usage ( gaming/ browsing ).

What are you guys doing to get this going even higher? I had this wet dream of getting it to 2000 mhz but I'm not seeing how. I tried to change most settings but the asus bios ( 1209, latest ) is so buggy that it is a pain in the ass to try to do something meaningful with it.
 

·
Registered
Joined
·
1,178 Posts
Hi guys,

just want to share how I was able to get my FCLK to 1900 mhz 100% stable on a Crosshair VIII impact ( with G.SKILL 4000 mhz CL19 32GB ( 2x16 )

My system is 100% stable at 1866 FCLK and 7333 memory, but it would not boot at 1900.

So, at 1900 FCLK the board would not even post. With everything on auto except for memory at 3800 and FCLK at 1900. The one thing that made it post was to disable SB spread spectrum. That single setting that was on auto, was the only thing I needed for the board to POST. The system was almost stable. I never had a crash when running stress tests or gaming and the only thing that didn't work properly was that my oculus rift would lose tracking from time to time ( tracker is connected via USB 3.0 ). I confirmed this to be from the FCLK by going down to 1866 and the problem going away.

So, to fix the USB thing I simply changed the VDDSOC load line calibration from auto to 5 ( the maximum ). The USB problem went away.

I
After deliding the CPU and using bare core on block ( I am using a heavily modified enermax liqtech II 240 ) I wasn't able to go further than 1900 FCLK. Temps are 18 degrees lower after deliding and using the bare core on the block.

For testing stability I ran 12h memtest86 plus 12h of prime95 ( latest version, with the default setting for testing memory ) and again 12h of prime95 with the max heat output setting. After that, one week of regular usage ( gaming/ browsing ).

What are you guys doing to get this going even higher? I had this wet dream of getting it to 2000 mhz but I'm not seeing how. I tried to change most settings but the asus bios ( 1209, latest ) is so buggy that it is a pain in the ass to try to do something meaningful with it.
I'll try your finding. With me previous 3900X and 3950X I had no problem running 1900 FCLK, but with the new 3950X I just cannot get it to post, so I'll give your finding a go to see if it works.
 

·
Registered
Joined
·
2,202 Posts
what exactly is a "heavily modified liqtech II"? At the end of the day, its still just a AIO, why not build a real loop?
 

·
Registered
Joined
·
59 Posts
Discussion Starter · #5 ·
Basically, I bought a LiqTech II 240, machined the CPU block to be able to clear the socket when using the bare cores on the block, and replaced the radiator for an all copper 240mm/45mm radiator from alphacool. I also replaced the fans with two noctua industrial ppc 2000.

The reason is that my case is a ITX Ncase M1 and the space is quite tight, so I needed a cpu block with some very specific characteristics to be able fit a 45 tall radiator, namelly, the cpu block height must be small, must have an integrated water pump and the hoses must come from the side instead of from the top of the block. The one from LiqTech II has all this plus a 450l/h rating ( that I found out is false, the true rating is 60l/h )

I made some videos of everything, including the deliding, but I am struggling to get time to actually edit all the stuff and put it on youtube. Anyway, the deliding and bare core on block did bring my cpu 18 degrees celcius down but I'm still unsure if it is worth the effort, or even the risk.
 

·
Registered
Joined
·
9 Posts
I disable SB Spread spectrum by default anymore (it gives a true 100mhz FSB) and have had no problems running my 3950x at 1900FCLK with the SoC set to 1.1 in bios (1.08 actual)

I have developed an annoying click in my USB DAC and I'm curious if the VDDSOC load line will fix it, as it has resolved your USB related issues.

Very curious about the deliding and your setup!
 

·
Registered
Joined
·
59 Posts
Discussion Starter · #7 ·
I disable SB Spread spectrum by default anymore (it gives a true 100mhz FSB) and have had no problems running my 3950x at 1900FCLK with the SoC set to 1.1 in bios (1.08 actual)

I have developed an annoying click in my USB DAC and I'm curious if the VDDSOC load line will fix it, as it has resolved your USB related issues.

Very curious about the deliding and your setup!

I think the load line @5 only fixed it because it upped the VDDSOC a bit from what it was @auto. I think at auto it displays 1.084v in bios and at 5 it displays 1.096v. But essencially, those where the only things I had to change from a completly default setting in bios to get to 1900 fclk. After that I messed with the memory timings and PBO and tightening the memory didn't take the stability away. I actually tested getting the VDDSOC again at auto with the current tight memory timmings I am running and the problem came back. Same with the SB spectrum. I turn it on and it doesn't POST.

I will try to find time this weekend to post the videos, or at least the tastiest one from the delid. It took me almost an hour to get the heat spreader off, and there were times where I really thought that I had killed the thing. You can't imagine the relief when tried it and it booted!! pheeww..
 

·
Registered
Joined
·
2,202 Posts
Basically, I bought a LiqTech II 240, machined the CPU block to be able to clear the socket when using the bare cores on the block, and replaced the radiator for an all copper 240mm/45mm radiator from alphacool. I also replaced the fans with two noctua industrial ppc 2000.

The reason is that my case is a ITX Ncase M1 and the space is quite tight, so I needed a cpu block with some very specific characteristics to be able fit a 45 tall radiator, namelly, the cpu block height must be small, must have an integrated water pump and the hoses must come from the side instead of from the top of the block. The one from LiqTech II has all this plus a 450l/h rating ( that I found out is false, the true rating is 60l/h )

I made some videos of everything, including the deliding, but I am struggling to get time to actually edit all the stuff and put it on youtube. Anyway, the deliding and bare core on block did bring my cpu 18 degrees celcius down but I'm still unsure if it is worth the effort, or even the risk.
good stuff man, honor your ability to delid that cpu!
 

·
Registered
Joined
·
59 Posts
Discussion Starter · #9 · (Edited)
I posted the delid video, will try to publish some others about the temps ( before and after ) and also the build process and the Liqtech modding this weekend/ next week.

The last AMD CPU I owned was a phenom II ( also delided ). Since the FX fiasco that I ditched AMD for Intel until now. But things have clearly changed regarding overclocking with AMD and I am struggling the accept some things, like having to choose between an overclock that allows me more freedom ( manual ) but denies me power savings and power savings without any real overclock ( PBO ). Also, I think that with this AGESA thing, AMD is essentially doing that Nvidia has been doing for the past couple of generations. Artificially limit the GPU potential by implementing hardware and software limits ( imposing temp, voltage and power envelopes ).

 

·
Registered
Joined
·
9 Posts
That is awesome! Props for having the will power to go through with that! What compound did you use to polish the silicon dies?
 

·
Registered
Joined
·
59 Posts
Discussion Starter · #11 · (Edited)
That is awesome! Props for having the will power to go through with that! What compound did you use to polish the silicon dies?
I used the blue polishing paste from wolfcraft. there are no specs for it that I could find.

http://www.wolfcraft.com/en/products/p/accessories-10/2_polishing_pastes/s/p/index.html

What the polishing procedure does is remove the very thin coat of indium solder that is left on the cores after you scrap them with a blade, exposing the naked silicon.
The silicon itself is very scratch resistant and doesn't scratch at all. The only thing one needs to take care of is actually breaking it. The corners are very sensitive and they can chip if hit with blunt impact.
Actually, this is the exact reason heat spreaders started to be used by Intel back in the day. The warranty claims started escalating around the pentium III time because of cracked cores from improper heatsink installation ( the heatsink would be installed at a slight angle, chipping the core and thus damaging it). This is actually very easy and happened to me with a couple of bare core cpu's. The solution was the heatspreader. It's name is misleading because it actually hampers heat transfer if compared with a properly installed heatsink on top of a core.
 

·
Meddling user
Joined
·
7,423 Posts
I posted the delid video, will try to publish some others about the temps ( before and after ) and also the build process and the Liqtech modding this weekend/ next week.

The last AMD CPU I owned was a phenom II ( also delided ). Since the FX fiasco that I ditched AMD for Intel until now. But things have clearly changed regarding overclocking with AMD and I am struggling the accept some things, like having to choose between an overclock that allows me more freedom ( manual ) but denies me power savings and power savings without any real overclock ( PBO ). Also, I think that with this AGESA thing, AMD is essentially doing that Nvidia has been doing for the past couple of generations. Artificially limit the GPU potential by implementing hardware and software limits ( imposing temp, voltage and power envelopes ).
+rep for video share. Surprised you so such a gain, as IIRC from all of der8auer's videos on this aspect he didn't see the gains you did.
 

·
Registered
Joined
·
59 Posts
Discussion Starter · #13 ·
+rep for video share. Surprised you so such a gain, as IIRC from all of der8auer's videos on this aspect he didn't see the gains you did.
I think that has to do because in all of der8auer's delid videos ( except the last one ) he keeps using the heat spreader, so he only replaces the indium solder with the liquid metal compound. In my case I don't use the heat spreader anymore and the cores touch the water block directly. This removes two layers from the heat path from the core to the air, one layer of thermal compound and one layer of metal ( the heat spreader ).

This 15 to 20 degree improvement is in line with all the delids I've made trough the years where I never used the heat spreader, since my phenom II times and then a bunch of subsequent intel cpu's that I owned before the 3950x.

There is one extra very interesting thing regarding using the bare cores onto the cooler block: it is block pressure and evenness. In the last delid video from der8aurer, on his first power on try he has quite bad results, very high temps, and he them reapplies liquid metal and tries again with better results. It didn't happen to me, but I did optimize my block pressure and evenness in a very easy and surprising way: I set the CPU to 2500mhz and locked voltage to 1v, and in prime95 I set a fixed fft and run at 32 threads. This made the heat output from the cpu very very constant. I let the system run until the temperature no longer raised and I then launched hwinfo.
hwinfo gives you per ccx temperatures with very fine grain in real time and I used this information to tune in the four screws of my cpu block. the result was surprising. I was able to gain 4 extra degrees ( from 14 to 18 ) just by adjusting slightly each of the four screws and it was quite funny as you could see temps change when rotating the screws.
 

·
Meddling user
Joined
·
7,423 Posts
Ahh yes it slipped my mind on his use of IHS with LM and overlooked you stated you've gone bare die.

Again video was very interesting :) , thank you again for share and clarification :) .
 

·
Registered
Joined
·
59 Posts
Discussion Starter · #15 · (Edited)
I had quite a joy building up my rig but that joy has converted to a kind of state of depression. I was really put down by the liqtech and their false claims ( 450l/h pump versus the 60l/h pump that I measured ) and also the very hard choice I have to make for my next step, and that is actually overclocking the damm thing. AMD has gone the path of nvidia putting limiters everywhere. We are now limited by whatever agesa says we can do or not with our chip. The fact that my crosshair VIII impact bios is buggy as hell didn't help either.

Anyway, I just ordered a just released ekwb AIO that fits my case. I will disasemble it as soon as I get it and test the pump performance and will use the best one on my rig. Regarding overclocking, I may start this weekend to try out per ccx overclock but the fact that I must give away clock and voltage scaling is really killing me.

On another note, I just received my 8mohm resistors to raise the power limit of my 2080ti blower edition, so I will also make a how to video about that, and the performance before and after. Maybe that will lift my mood a bit :)
 

·
Registered
Joined
·
233 Posts
I posted the delid video, will try to publish some others about the temps ( before and after ) and also the build process and the Liqtech modding this weekend/ next week.

The last AMD CPU I owned was a phenom II ( also delided ). Since the FX fiasco that I ditched AMD for Intel until now. But things have clearly changed regarding overclocking with AMD and I am struggling the accept some things, like having to choose between an overclock that allows me more freedom ( manual ) but denies me power savings and power savings without any real overclock ( PBO ). Also, I think that with this AGESA thing, AMD is essentially doing that Nvidia has been doing for the past couple of generations. Artificially limit the GPU potential by implementing hardware and software limits ( imposing temp, voltage and power envelopes ).

https://www.youtube.com/watch?v=qu839b_VleE&t=47s
So like...
What are those?
 

Attachments

  • Rep+
Reactions: finas

·
Registered
Joined
·
59 Posts
Discussion Starter · #17 ·
So like...
What are those?
wow! they sure seem like two capacitors!!!!
I didn't notice them when I did the delid, and the only plausible explanation is that they came off during the delid process.

Having said that, I haven't noticed absolutely nothing strange with my system ( it's been on 24/7 since I finished assembling it ) and I have been doing benchmarks, gaming with maxed out PBO without any stange thing happening. The system is ultra smooth and stable.

I tried to dig some info about it's use and I came up with this post just now:

https://electronics.stackexchange.c...ted-capacitors-located-directly-on-top/273360

So, if that information is correct it seems they are used to make sure power is feed to the cpu during transient ( load ) events.

Anyway, excellent catch, I will update the video description with a note and warning.
 

·
Overclocker in training
Joined
·
14,315 Posts
wow! they sure seem like two capacitors!!!!
I didn't notice them when I did the delid, and the only plausible explanation is that they came off during the delid process.

Having said that, I haven't noticed absolutely nothing strange with my system ( it's been on 24/7 since I finished assembling it ) and I have been doing benchmarks, gaming with maxed out PBO without any stange thing happening. The system is ultra smooth and stable.

I tried to dig some info about it's use and I came up with this post just now:

https://electronics.stackexchange.c...ted-capacitors-located-directly-on-top/273360

So, if that information is correct it seems they are used to make sure power is feed to the cpu during transient ( load ) events.

Anyway, excellent catch, I will update the video description with a note and warning.
Hi,
Yeah I imagine it was your razor blade that took those off.
 

·
Registered
Joined
·
31 Posts
How did you mount the cooling block on the die? Did ya use a special frame or something created by yourself?
 
1 - 20 of 41 Posts
Top