Processors

[kb1972]

I run a small computer business.

I had a  Ryzen 7 3700X based machine come in recently where the machine would hang with the display going blank and the DRAM light on the motherboard (Asus Gaming TUF X570 Plus) constantly lit. Once in this state, the machine no longer POSTs. 

The problem has been narrowed down to a faulty CPU. Confirmed with a different CPU in the same board and this CPU in a different board.

In over 30 years of working with computers, I've only come across a dead CPU once. And that was caused by a customer taking the heatsink off and putting it back on with only some dried up thermal material covering only patches of the CPU surface.

I did suspect the cooler on this Ryzen machine straightaway from the dubious pattern left by the thermal paste when the cooler was removed. 

I have also seen numerous other posts in various forums regarding the DRAM light issue (different boards), with no solutions for most of them. 

Take a look at the attached pictures. In "Cooler-defect 1.jpg" you can see the straight edge only makes partial contact with 2 of the 4 heat pipes. The 1st heat pipe makes no contact whatsoever and the 4th one is barely touching. 

In "Cooler-defect 2.jpg" taken further along the heatsink surface, things are a bit better but still far from perfect.

It's almost as if the heatpipes are sunken in further than they ought to be on one side. 

This leaves some parts of the CPU cool while other bits cook themselves to death.

This has the potential to affect a lot of customers and obviously substantial financial harm to AMD if lots of CPUs start dying - all because of some shoddily made coolers. Yes the rest of the cooler looks perfect, but that hardly matters if the business end is f*cked. 

Has anyone else come across any dodgy bundled Ryzen 7 coolers? In this case, the cooler is the Wraith PRISM. 

I think AMD should investigate this and if some coolers are found to be sub-standard, it would still be cheaper for AMD to replace these than it will be to replace loads of dead CPUs.

[pokester]

I agree, that is as bad as I have ever seen. No amount of lapping will fix that. I hope it is a very isolated issue but obviously quality control failed you purchase.

You can RMA here. https://www.amd.com/en/support/kb/warranty-information/rma-form 

amdmatt  may have some questions about your situation and possibly could pass this along to someone who can evaluate. 

[pokester]

I think cooler master makes these for AMD. Maybe send those pictures to their support department too. 

[kb1972]

I would like to think it's an isolated case too, but there are numerous posts in other forums with people having the same symptoms - albeit none seem to have correlated it with a dodgy cooler. The problem is, most people check the system reported temperature and see nothing wrong.

Faulty cooling is easily fatal to chips. If the temperature diode is cool, the system and hence you, have no way of knowing another part of the chip is literally cooking itself to death.

My customer will RMA himself as he bought the parts and assembled the machine himself. He only came to me when he started having problems.

I make a habit of checking thermal paste distribution when fitting a cooler so would have picked this up before any damage was done if I'd assembled the machine. I'm fully aware that most people assume the cooler fits perfectly and don't bother checking other than perhaps running a temprature check after getting the machine up and running.

Hopefully someone from AMD can pick this up and look into it.

This guy had already RMA'd his motherboard and RAM before he came to me. And he still has problems believing it's the CPU as the motherboard tells him it's a DRAM issue and not a CPU issue.

The problem is, I can see people giving up on AMD and switching to Intel if they end up with problems they can't solve as they don't have the knowledge to look past what the lights are telling them.

And that would be a crying shame given all the good work AMD has done lately in delivering improved CPU performance.

[elstaci]

Can you take a photo of the CPU Cooler heat sink under pressure to the CPU ILS to see if the slight warped surface bends to make full contact?

[kb1972]

2 things:

1. This is not a slight warp. The heatpipes are sunk in too far into the copper block on one side
2. I'm no mechanical engineer, but I seriously doubt the block would bend from the force exerted by the retention clip. I suspect the plastic lugs that hold it in place would shear off before you could successfully "bend" the heatsink.

In any case, the customer has collected his machine already to RMA the CPU so I couldn't take any more pictures even if I wanted to. That's assuming a) I could somehow get a camera level with the top of the CPU surface into the narrow confines around the CPU socket where it's surrounded by other components and connectors and b) any gaps (or lack of) would even be discernable on such an image.

[elstaci]

True, I chose the wrong words when I mentioned about it being warped.  The image you posted gave me that impression. 

It would have been better, in my opinion, to place a thin straight edge like a metal Ruler or something similar right against the Heat sink plate surface then shine a bright light behind the plate and see where it isn't making contact with the surface.

Just because the copper tubing is not perfectly aligned to the surface at the edge doesn't mean it is not making contact in the middle of the plate.

I just saw a Youtube video on how CPU Cooler heat sinks are manufactured. Grooves are made on the heat sink in which the copper flat pipes are laid in: The production of be quiet! CPU coolers, part 2: punching and milling - YouTube 

This 3rd part of the youtube shows how the copper tubing is inserted and checked for alignment to the surface of the heat sink: The production of be quiet! CPU coolers, part 3: Assembly and QA - YouTube 

So most likely it was just a few that weren't perfectly aligned during manufacturing and QA didn't catch it. Otherwise AMD would be swamped with burnt out Ryzen CPUs. 

Here are all the bundled Ryzen 7 with the CPU Cooler that goes with it:

If what you said was true then it would also be affecting Ryzen 9 processors plus any Ryzen processors that uses the Prism or Spire CPU Coolers will be affected since the heat sinks are all the same whether it was RGB programable or not.

[kb1972]

That's exactly how I took those pictures. The straight edge you see is a high precision caliper.

The bright light behind the heatsink is the sun. Can't get much brighter than that easily.

[chad27]

I thought about one of these for my 3600 that's gets really warm with stock cooler, But maybe Ill look for another option now.

[kb1972]

Personally I find the bundled coolers, especially the lower end models a bit noisy. They emit a whirring sound which I find annoying. There's no way I would use one in my own machine.

For quiet budget builds the Artic Freezer 34 is a good cooler.

For high end air cooling I'm not sure you can beat a Noctua NH-D15. Temperatures pretty much on a par with water cooling without the risk of water inside your machine.