5950x and PBO overheated something or busted custom loop?

[Ferrum Master]

Ah, number chasing is OK. But I do it only when chasing BSOD and doing curve UV with PBO. At least then the time ain't wasted totally. You should better start to invest time in that.

For 10minute run having +2.7C rise is a bit fast in my books. But my loop has more capacity, I am running two 360ties + 240 at the bottom, I couldn't fit my 280 at the bottom as O11D has a funny design flaw with that, that none of the so called self proclaimed expert reviewers noticed. Also keep in mind that EK block you have leaves much to be desired, I have a similar design implemented in my Monoblock.

 

[A Computer Guy]

Ah, number chasing is OK. But I do it only when chasing BSOD and doing curve UV with PBO. At least then the time ain't wasted totally. You should better start to invest time in that.

Well rather than watching TV this is my entertainment for this month.

For 10minute run having +2.7C rise is a bit fast in my books. But my loop has more capacity, I am running two 360ties + 240 at the bottom,

I take it that I did a pretty good job with my thermal paste then, thermal grizzly kryonaut performance confirmed? Isn't the blocks performance kind of moot since my loop under full load currently can only dissipate heat at half the rate it absorbs for this rather short burst of activity? You have a bit over twice the heat dissipation potential of my setup so I'd think your absorption/dissipation ratio should be much closer to 1:1 than mine.

I couldn't fit my 280 at the bottom as O11D has a funny design flaw with that, that none of the so called self proclaimed expert reviewers noticed.

I'm curious to what is your rad configuration? Having the front EK distro plate like I have unfortunately blocks the idea of the side mounted rad (or fans) in the front and a rad on the bottom becomes a bit problematic with clearance and access to connectors of a full ATX motherboard (d-rgb cable connector is too tall) and whether or not you have access to the bottom most PCIe slot. I was considering if I could fit another 360 in the side mount rear but the clearance would be really tight with the rear panel when regular 120mm thickness fans are installed. I remember from reading reviews EK CoolStream 360 SE won't fit back there because of front panel wiring issues (rad is too wide) so I would need some other kind of slim RAD that isn't as wide and also thin enough to fit back there. It seems to me a 3 rad config isn't really ideal/optimal/practical concept for the original 011D and I've pretty much given up on the idea of adding more rads in the case because of these constraints. My only practical option is to go external with more rads for improved heat dissipation.

Also keep in mind that EK block you have leaves much to be desired, I have a similar design implemented in my Monoblock.

I'm not quite sure what you mean? The performance of the block isn't really earth shattering but doesn't really seem bad either.

EK-Quantum Magnitude CPU Water Block Review

EK Water Blocks finally launches the EK-Quantum Magnitude, a halo CPU block with customization cues from the EK-Supremacy EVO from yesteryear that kicks things up a notch. With a clear focus on design, material choice, performance, and customization, it aims to offer a lasting CPU cooling...

www.techpowerup.com

 

[Ferrum Master]

Well rather than watching TV this is my entertainment for this month.

I have a overclocked GPU in the loop spitting 300W of energy in the loop also. Top internal side rads are 360ties and bottom 240. If I put the 280 all my bottom motherboard I/O for USB switches etc are blocked and headers have no space to be attached. It is a serious flaw in the case design.

I had an opportunity to compare blocks when building loop for a mate using my older block and thus I made a conclusion on my own, EK sucks about for ~5C. I have an older pic of the build. There are things to improve. I can attach another 360 sandwiched in the internal side wall. Well maybe if those new gen 600W GPU's it would become a need.

 

[A Computer Guy]

Didn't have time yet to run more PBO tests but got these puppies to help complete the aesthetic of my build since my ECC ram is naked without heat spreader.
It was a bit of a PITA to install but it cools just fine with a bit of air flow. (I'm NOT getting the water block for this.)
Probably going to post some pictures tonight of the assembly and workaround of the installation issues. The design is somewhat terrible. Not EK's finest product.
Currently waiting for replacement PrimoFlex tubes to try and reconfigure loop to improve flow and reduce pump noise.
I finally got my ARGB extension cables so I can hook my distroplate rgb back up again since the motherboard swap. When I light it up the ram should reflect ARGB greatness.

Heat spreader installation.

Before. Micron OTE75-D9ZFV, I think these are Micron E-die (https://www.micron.com/products/dram/ddr4-sdram/part-catalog/mt40a2g8jc-062e)

Ohhh, shiny

The pads only came for 8 chips per side so I had to improvise

Sizing up the ram

Need a common point of alignment so I can get both sticks symmetrical

After assembly there was a noticeable size difference at the end of the module. The mounting point to the end was off by about a 1mm.
( sorry I don't have a picture of that, just imagine picking up a ham sandwich and squeezing one end of the bread )

Got to do some measurements. First time using this caliper, just opened it from the package. Was sitting in my tool drawer for about 7 months unused until now.

First the joined edges on both sides. Looks pretty good.
(yes I blurred out my fingerprint in the photo)

Now the other side

Not liking this much. Going to try to reduce the tilt on the chips to ensure a more level compression.

These are the smallest washers I have on hand. I cut them to fit as spacers.

Reassembled, lets take some more measurements. Looks like from certain angles I can see the machining lines.

Ok ends looking a bit more consistent. The middle is a little more compressed.

back to the joined edges

So the mount is still tilted a bit but a significant improvement. I need to find a way to make up 0.22 mm.
Ideally I'd like to get rid of the spacers and perhaps find some thinner thermal pads but that's a project for another day.

One module reinstalled. On to testing and comparison.

Heat absorption by the heat spreaders confirmed but testing with OCCT confirmed I'm going to need some direct airflow.
Noctua 40mm fan to the rescue.

Not sure where the DIMM temp sensors are but DIMM 1 always runs a little hotter about 1 degree.
Since DIMM 2 is on the out edge it does get some reflective airflow from the side fans

I lost my OCCT screen shots comparing DIMM 1 (with heatspreader) & 2 (bare) thermal results so I will have to redo my testing some day.

Cold Start Test and some TPU web browsing

After 1 hr OCCT ram test. Ram stayed under 45 degrees. (previous peak temp was 51)

 

[Ferrum Master]

Another sanity call.

Well... a heatsink without ribs... an additional surface area... ain't a heatsink, it is more a heat spreader.

Basically those EK are just for the looks, nothing else. EK has many things to improve on.

 

[Valantar]

Another sanity call.

Well... a heatsink without ribs... an additional surface area... ain't a heatsink, it is more a heat spreader.

Basically those EK are just for the looks, nothing else. EK has many things to improve on.

There isn't much room for fins or ribs on a RAM heatspreader though - the best you can realistically do is some more thermal mass and a larger area to spread the heat across. Those heatsinks with fins at the top are pretty much useless, as thermal conductivity into those fins will be next to nothing, rendering their marginal extra surface area useless. And, tbh, a simple heatspreader is all RAM needs - if it needs anything at all. Sure, some RAM (like Samsung B-die DDR4) needs keeping somewhat cool, but the heat output just isn't enough to warrant any serious cooling.

 

[Ferrum Master]

There isn't much room for fins or ribs on a RAM heatspreader though - the best you can realistically do is some more thermal mass and a larger area to spread the heat across. Those heatsinks with fins at the top are pretty much useless, as thermal conductivity into those fins will be next to nothing, rendering their marginal extra surface area useless. And, tbh, a simple heatspreader is all RAM needs - if it needs anything at all. Sure, some RAM (like Samsung B-die DDR4) needs keeping somewhat cool, but the heat output just isn't enough to warrant any serious cooling.

You can do better by simply putting holes or making it with bumps, no need for Dominator approach. Other than that, this is just a fancy sock, or maybe because of the price a piece of exclusive lingerie lol.

 

[Calenhad]

Well... a heatsink without ribs... an additional surface area... ain't a heatsink, it is more a heat spreader.

Any piece of metal you sink heat into is by definition a heatsink (it is literally in the name). It can be ribbed for your pleasure, and more surface area.

 

[A Computer Guy]

There isn't much room for fins or ribs on a RAM heatspreader though - the best you can realistically do is some more thermal mass and a larger area to spread the heat across.

The EK spreader has a fairly large mass and the total surface area to dissipate heat ends up tripling (sorry didn't calc the measurements exactly) on on the surface of the modules alone. I don't have a picture if it but also there is quite a bit of room inside the spreader as well. Forcing air through that cavity again increases the surface area again maybe by another 3/5ths and when shoving airflow through both sides of the spreader there is enough surface area where I can reach parity with the motherboard temp during the entirety of the OCCT 1hr test.

Caveats of the EK heat spreader

• Cost - the cost is totally unrealistic of course.
• Assembly is problematic. Alignment of the modules takes some work. Fitment can be a problem. My double sided modules were a bit too thick and I had to compensate by adding the spacers to the clamping area to get a more level fit. The 0.5mm pads were too thick for my application and needed maybe 0.2mm to 0.25mm on each side instead.
• The heat spreader is not perfectly flat so a pad is required to make up the difference.
• The heat spreader is not symmetrical. So one side gets slightly better spread than the other.
• Without airflow the spread makes temps worse
• There is no surface that isn't a fingerprint magnet

Those heatsinks with fins at the top are pretty much useless, as thermal conductivity into those fins will be next to nothing, rendering their marginal extra surface area useless. And, tbh, a simple heatspreader is all RAM needs - if it needs anything at all. Sure, some RAM (like Samsung B-die DDR4) needs keeping somewhat cool, but the heat output just isn't enough to warrant any serious cooling.

Like with my EK NVMe heatsink it ended up with higher temps without airflow but with airflow it ended up with cooler temps. The same thing happened here with these ram heat spreaders. In my case I ended up shaving off -5c from my previous max temp running OCCT with just some light airflow from my 40mm fan.

( additional update 9/2 )

So I got my new ARGB extension cables so I can rehook up the distroplate RGB. I took a whopping 2hrs because it turns out he cable I tried to use was bad and I didn't know it until after I got all my zip ties in place to hold the connectors together. After some troubleshooting and finally swapping out the bad cable (and trying for like 15 minutes to get the darn thing plugged into the mainboard again) it finally worked and I had to redo all my zip tie work.

The black PrimoFlex tubes won't arrive until next week should be an interesting effect with the lighting.

 

[Ferrum Master]

It looks expensive. Good job

I am not a fan of distro plates... too much tubes.

 

[A Computer Guy]

It looks expensive. Good job

I've been piecemealing the improvements over the years.

I am not a fan of distro plates... too much tubes.

If I just would have stuck with the pump/tube combo I would have had simply 4 tubes + drain and life would have soo much simpler.

I have a overclocked GPU in the loop spitting 300W of energy in the loop also. Top internal side rads are 360ties and bottom 240. If I put the 280 all my bottom motherboard I/O for USB switches etc are blocked and headers have no space to be attached. It is a serious flaw in the case design.

I had an opportunity to compare blocks when building loop for a mate using my older block and thus I made a conclusion on my own, EK sucks about for ~5C. I have an older pic of the build. There are things to improve. I can attach another 360 sandwiched in the internal side wall. Well maybe if those new gen 600W GPU's it would become a need.

View attachment 260223

Is the top rad a cross flow? Are they all xspc rads? Is the side rad ports on bottom or top?

 

[Ferrum Master]

I've been piecemealing the improvements over the years.

If I just would have stuck with the pump/tube combo I would have had simply 4 tubes + drain and life would have soo much simpler.


Is the top rad a cross flow? Are they all xspc rads? Is the side rad ports on bottom or top?

No, only the top one is crossflow, it really helps with the tube management as the CPU exit tube looks always ugly if there are two of them. The exit of the crossflow goes directly into the tank and is crewed on with extender, so the tank does not need any clips, it just stays where it should be. I got it years ago also. The internal side is a Bykski 360mm/38mm thick low density fin one, ports are at the bottom, all connected with fittings and extenders, basically tubeless. Actually I am planning to do the there the stacked variant. You can fit a slim 360, so rads on both sides of that bracket. You can screw them on by taking out and slide them opposite ways, top screws should be visible, thus you can manage to tighten them enough when sliding back and then put in the case, the fan pressure should be enough, but it is just all about heat capacity. The bottom one 240 is my oldest rad, I have even repainted it twice. Shame 280 didn't fit there, it caused me an unwelcome surprise as found out about only when screwing everything together. And the manometer there ain't for the looks also. It is a 1bar one and it works really well as a loop internal pressure gauge. You can test leaks with it also. If the pump ramps up, needle goes up. Also I changed the CPU tubes to 10/16. Thus now I don't need a crouch to support the GPU, it stays firm in the slot.

But yea... with that distro plate you caused more clutter than use versus doing the traditional way. But much more easy, as you do not need to cherry pick the fittings.

 

[A Computer Guy]

Got the PrimoFlex today

I finally refitted the flow but my pump noise still remains - no improvement. Either I simply got a noisy pump, or the distroplate 90's are the issue, our my case resonance with the EK D5 pump/distroplate combo is just noisy. It looks better though.

Before

After new tubes
- had to relocate my warm temp sensor up top with my 4 way splitter.
- reconfigure drain value and since I had a ball value laying around I also installed the drain on the other side of the distroplate.

The bottom loop was a bit challenging to make it small enough but not kink. To help prevent kinking I used some zip ties.

 

[ThrashZone]

Hi,
Still should lay your case on it's back so distro plate is facing the ceiling
Might help get any air out of the pump if any exist
Then bop the case to release any bubbles.
Repeat tilting in all directions bop again.

 

[A Computer Guy]

Hi,
Still should lay your case on it's back so distro plate is facing the ceiling
Might help get any air out of the pump if any exist
Then bop the case to release any bubbles.
Repeat tilting in all directions bop again.

I'll try that. Also the distroplate flow indicator does have improved spin at low RPM so I can see it is flowing better.

 

[tabascosauz]

The EK spreader has a fairly large mass and the total surface area to dissipate heat ends up tripling (sorry didn't calc the measurements exactly) on on the surface of the modules alone. I don't have a picture if it but also there is quite a bit of room inside the spreader as well. Forcing air through that cavity again increases the surface area again maybe by another 3/5ths and when shoving airflow through both sides of the spreader there is enough surface area where I can reach parity with the motherboard temp during the entirety of the OCCT 1hr test.

Caveats of the EK heat spreader

• Cost - the cost is totally unrealistic of course.
• Assembly is problematic. Alignment of the modules takes some work. Fitment can be a problem. My double sided modules were a bit too thick and I had to compensate by adding the spacers to the clamping area to get a more level fit. The 0.5mm pads were too thick for my application and needed maybe 0.2mm to 0.25mm on each side instead.
• The heat spreader is not perfectly flat so a pad is required to make up the difference.
• The heat spreader is not symmetrical. So one side gets slightly better spread than the other.
• Without airflow the spread makes temps worse
• There is no surface that isn't a fingerprint magnet

Like with my EK NVMe heatsink it ended up with higher temps without airflow but with airflow it ended up with cooler temps. The same thing happened here with these ram heat spreaders. In my case I ended up shaving off -5c from my previous max temp running OCCT with just some light airflow from my 40mm fan.

( additional update 9/2 )

So I got my new ARGB extension cables so I can rehook up the distroplate RGB. I took a whopping 2hrs because it turns out he cable I tried to use was bad and I didn't know it until after I got all my zip ties in place to hold the connectors together. After some troubleshooting and finally swapping out the bad cable (and trying for like 15 minutes to get the darn thing plugged into the mainboard again) it finally worked and I had to redo all my zip tie work.

The black PrimoFlex tubes won't arrive until next week should be an interesting effect with the lighting.

Am I missing something here...? The EK kit is not a heatspreader, it's an adapter for their RAM waterblock (same with Barrow). A lot of flat slab third party RAM spreaders are not designed for air, instead intended as either waterblock adapter or XOC use (LN2). For air you need surface area and airflow, flat slabs don't provide much. You wouldn't be too far off just using the 40mm fan on naked DIMMs.

Case in point - the FMMJ kit looks really nice but that's really all it has going for it, for above reason

XTIA & HAO FMMJ RAM Jacket V3

The United States will impose a unified tariff of 100-200 US dollars on all small packages after May 2 (ignoring the declared amount and declared type). There are no special changes to the policies of other countries and regions for the time being. If you are unwilling to pay this tariff, it is...

xtia.design

 

[A Computer Guy]

Am I missing something here...? The EK kit is not a heatspreader, it's an adapter for their RAM waterblock (same with Barrow). A lot of flat slab third party RAM spreaders are not designed for air, instead intended as either waterblock adapter or XOC use (LN2). For air you need surface area and airflow, flat slabs don't provide much. You wouldn't be too far off just using the 40mm fan on naked DIMMs.

This was primarily for aesthetics although it does increase the surface area and spread heat too. With some air flow it is mildly effective at managing heat.

 

[A Computer Guy]

Check this out. I had this in my storage and decided to see if it still works.

TPU's Nostalgic Hardware Club

Have a set of Geils just like those - Same thing in fact and they will do DDR500..... And not an inch more. Not really happy with them but they were cheap at least when I got them.

www.techpowerup.com

RAM DISK configured in dual SATA mode via jumper. (will present itself as two disks, because two banks of ram)

Turning on the Windows 10 software RAID 0. (Hardware RAID in UEFI/BIOS is off) SATA mode is ACHI, so 2 disks detected by OS and joined as a raid 0 volume in Windows.

Turning on the RAID in the UEFI/BIOS and configuring the RAM Drive as an actual RAID 0

Here is a comparison against a real RAM disk carved out of my actual RAM

 

[Mussels]

"- So EDC past 145A kills boost clocks so that is a no go."


This is a BIOS bug that's cropped up, i don't know specifics but it's been mentioned a few times

 

[Ferrum Master]

This is a BIOS bug that's cropped up, i don't know specifics but it's been mentioned a few times

I do not think it is a bug. It is intended by design. AGESA brang it, there's no blame on motherboard vendors.

 

[Mussels]

I do not think it is a bug. It is intended by design. AGESA brang it, there's no blame on motherboard vendors.

Not the first time an overclocking feature has gone awry and been fixed in updates.

Some of those changes came in for the 5800x3D and got applied to the regular 5800x, as an example.

 

[Ferrum Master]

Not the first time an overclocking feature has gone awry and been fixed in updates.

Some of those changes came in for the 5800x3D and got applied to the regular 5800x, as an example.

They came at least one official AGESA before the 5800x3D, so it may not be the culprit.

 

[Zach_01]

"- So EDC past 145A kills boost clocks so that is a no go."


This is a BIOS bug that's cropped up, i don't know specifics but it's been mentioned a few times

EDC has a default limit of 140A for the 105W TDP SKUs anyway.
Most users tend to extend this limit (EDC) along with PPT and think they are giving the CPU headroom to boost more.
So the board sees the extended limit and without further ado cranks everything up... Its just how vendors have them.
The reality is that the CPU silicon FITness (health) management (which is more intelligent than the board, but still its not AI) sees the increased current(A) and cuts down boost, because that is its main job.

Its "common sense" or it should be by now (for those who understand a couple of things about silicon degradation) that within the same power envelope and same temperature, higher current(A) can start killing the silicon little by little through EMI (migration). If you keep steady the power and temp, and start bringing down current(A) the clocks will go up (hence the same W and T), to the point that voltage is enough for sustained boost. If you restrict EDC too much clocks will go down again.
Curve Optimizer (negatives) plays a major role into this. It helps you bring down EDC more to the point that (again) voltage is enough for clocks to go up (but) with stability.
*If you can cool down silicon further with a better cooler (lets say from 85C to 75C), then FIT will let not only clocks but current(A) too to go up. Power will go up as well as a result of this.
*Its too difficult though to keep temp under control on these chips mostly because of the heat density.

The only real bypass of FIT is the CPU PBO Scalar (xX). The more times you set it (x2, x3, x4...) the bigger the bypass.
I've been saying this for about 2.5years now, and finally, at least Gigabyte has a description about it in BIOS.

Here:



EDIT:
Typo(s)

And some addition (*)

 

[Ferrum Master]

Gigabyte has a description about it BIOS.

I refuse to believe my eyes seeing that something is done right in Gigabyte Bios.

 

[Mussels]

I believe that's why TDC exists

Thermal Design Current (“TDC”): The maximum current (amps) that can be delivered by a specific motherboard’s voltage regulator configuration in thermally-constrained scenarios.

Much later i've come to the conclusion that that TDC only applies when the CPU is overheating, and doesnt get any sort of feedback from the VRM's - because we've never seen any motherboard actually apply it, they all just overheat.

TDC is for something like a 2700x on the wraith prism, where it lowers the amperage enough to function without overheating the CPU - it's not aware of VRM temperatures, and assumes the motherboard will control those by limiting itself to values it can handle.