• Guest, The rules for the P & N subforum have been updated to prohibit "ad hominem" or personal attacks against other posters. See the full details in the post "Politics and News Rules & Guidelines."

AMD RYZEN Builders Thread

Page 199 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

DrMrLordX

Lifer
Apr 27, 2000
14,457
3,433
136
Hello old Ryzen builders thread! What good is a builders thread without some posts about teardowns? I just got finished decommissioning my R7 1800x to make ready for R9 3900x. Not much to report here (other than the fact that I lost my box for my BPX NVMe SSD, grr) until I got to the CPU. I had considered leaving the HSF mounted so I could just pop board + CPU + HSF with a known-good mount into another case someday for a new build. Now I'm glad I didn't.

This setup has been up and running mostly-untouched (only hauled it out to the garage for dusting 2 times) since late March 2017 when I got my eBay Taichi for $317 (ouch). It was my first time using Conductonaut. I was familiar enough with the Noctua mounting system than installing the NH-d15 was mostly painless.

I was expecting greyish crust all over the CPU and HSF when I unmounted the HSF ala CLU. Oh no, that's not what Conductonaut does.

It stays liquid.

Well, mostly. There was a cloudy streak down the middle of the IHS where apparently there was maximum contact. Or something. I saw the same streak reflected on the HSF base. It was nickel->nickel contact (no lapped surfaces) so aside from the streak there was no particular adhesion to the nickel surfaces. In retrospect I should have lapped CPU + HSF to at least see if Conductonaut would "corrupt" copper (and stick to it better).

The other thing is does is pump out.

I saw several (at least three) large dollops of Condoctonaut that had vacated the space between IHS and HSF base that rolled down the side of the IHS and just . . . stayed there, on the edge of the CPU pcb. Only a tiny bit got onto the board in an irrelevant location. Thankfully. I managed to sort-of roll some of the excess back onto the IHS with a coffee filter before removing the CPU from its socket. A tiny bit almost got onto the CPU pins, but I rolled it away with another coffee filter. Another tiny bit got onto the edge of the socket . . . but not into any of the pinouts.

At least theoretically board + CPU should work for someone else.

Kinda makes me wonder if I should use Conductonaut again. The contact pattern it exhibited matches the two-screw retention mechanism for the Noctua well-enough that I guess I shouldn't be surprised it behaved as it did. I have Kryonaut handy if I need it. The funny thing is the Conductonaut mount cooled so well for over two years, and yet it certainly looks like I got only partial contact between IHS and HSF for who-knows-how-long.

Anyway, farewell 1800x. You were a revolution two years ago. Hopefully I'll find a good home for you. Think I got a nice one where you can run stock for who knows how many years, without those loud fans attached. You get to stay with that top-bin b-die RAM too, and in your old Taichi motherboard. And maybe even the BPX but we'll see.
 

UsandThem

Elite Member
Super Moderator
May 4, 2000
12,336
2,815
146
Kinda makes me wonder if I should use Conductonaut again.
After watching various videos of that stuff, and seeing people talk about it on various forums over the years, I personally wouldn't touch that stuff with a 10 foot pole.

Give me some good paste, and a good cooler.
 
  • Like
Reactions: Makaveli

DrMrLordX

Lifer
Apr 27, 2000
14,457
3,433
136
After watching various videos of that stuff, and seeing people talk about it on various forums over the years, I personally wouldn't touch that stuff with a 10 foot pole.
As a post-post mortem, it looks like Conductonaut got on some CPU pins (I guess?) and shorted out/killed my board during a test run to see if I could flash the board to a newer UEFI to sell it to someone else. Woops. I think the moral of the story here is that you should have at least one copper surface to prevent Conductonaut from rolling around all over the place.
 
  • Wow
Reactions: Ajay

VirtualLarry

Lifer
Aug 25, 2001
46,870
4,554
126
As a post-post mortem, it looks like Conductonaut got on some CPU pins (I guess?) and shorted out/killed my board during a test run to see if I could flash the board to a newer UEFI to sell it to someone else. Woops. I think the moral of the story here is that you should have at least one copper surface to prevent Conductonaut from rolling around all over the place.
:( :( :( :(
 

nOOky

Golden Member
Aug 17, 2004
1,535
335
136
Dredging this up... is the Ryzen 7 2700X still a decent pick? The wife's FX6300 took a poop on her, so I'm upgrading the CPU, MOBO, SSD, and new memory. I think what I have picked is pretty standard stuff for the CPU i.e.:

G.SKILL Flare X Series 16GB (2 x 8GB) 288-Pin DDR4 SDRAM DDR4 3200 (PC4 25600) AMD X370 Memory Model F4-3200C14D-16GFX

ASUS ROG Strix B450-F Gaming Motherboard (ATX) AMD Ryzen 2 AM4 DDR4 DP HDMI M.2 USB 3.1 Gen2 B450


And a
Samsung 970 EVO SSD 500GB - M.2 NVMe Interface Internal Solid State Drive with V-NAND Technology (MZ-V7E500BW)

We will be keeping the existing case, fans, PSU, GPU, etc. and she runs a 4k monitor, but only lightly games older STEAM games etc. so the RX580 is staying.
 
Feb 4, 2009
24,397
5,290
136
Dredging this up... is the Ryzen 7 2700X still a decent pick? The wife's FX6300 took a poop on her, so I'm upgrading the CPU, MOBO, SSD, and new memory. I think what I have picked is pretty standard stuff for the CPU i.e.:

G.SKILL Flare X Series 16GB (2 x 8GB) 288-Pin DDR4 SDRAM DDR4 3200 (PC4 25600) AMD X370 Memory Model F4-3200C14D-16GFX

ASUS ROG Strix B450-F Gaming Motherboard (ATX) AMD Ryzen 2 AM4 DDR4 DP HDMI M.2 USB 3.1 Gen2 B450


And a
Samsung 970 EVO SSD 500GB - M.2 NVMe Interface Internal Solid State Drive with V-NAND Technology (MZ-V7E500BW)

We will be keeping the existing case, fans, PSU, GPU, etc. and she runs a 4k monitor, but only lightly games older STEAM games etc. so the RX580 is staying.
Per my shopping around last year I **think** there is something weird about flare x memory and Ryzen 2 chips. Cant remember exactly but I think it won’t run at it’s rated speed.
I could be wrong so check the vendors memory support list.
 

DrMrLordX

Lifer
Apr 27, 2000
14,457
3,433
136
FlareX is just b-die with profiles set up for AMD machines. The "XMP" profiles might not work on Pinnacle Ridge since I think they were set up for Summit Ridge. But you can hand-tune to get around that. That being said, I wouldn't flock to FlareX since you can get Micron e-die for less.

 

amrnuke

Senior member
Apr 24, 2019
497
449
96
This is such a noob question...

I have two front intake Corsair LL120s, one rear exhaust LL120, and one top exhaust (default thermaltake 120mm fan with the Versa H18 case).

I have a CM 212 Evo LED on the 3600.

I also have a reference blower 5700.

While running F@H on the GPU and WCG/R@H on BOINC on CPU, I wanted to try something, so I cracked the bottom edge of the glass side piece. I noticed that the system is sucking in air through the side of the case when it's cracked open. I guess I have some questions on what I should be doing. I know the blowers intake air from the case and toss it out the back, but I didn't expect the suction through this bottom edge of the case glass to be so prominent. It leaves me wondering if I shouldn't just leave the bottom edge of the case cracked open, or if this might be indicative of some larger issue with airflow in my case. How do y'all handle this? Is there any legitimate reason to seal up the case and rely on the fans alone? Or is liberating the airflow going to cause any real issues?

Even with it cracked, I'm a little dismayed to see my 3600 with an aftermarket cooler hitting 77+C on a frequent basis even with the case cracked. Did I just get a bad chip? Did I use too much thermal paste? Ambient is 70F right now.
 

DrMrLordX

Lifer
Apr 27, 2000
14,457
3,433
136
Looks like you're running a negative pressure case, which is (in some instances) a dust magnet. Not really a big problem otherwise so long as you are satisfied with component temperatures.
 

amrnuke

Senior member
Apr 24, 2019
497
449
96
Looks like you're running a negative pressure case, which is (in some instances) a dust magnet. Not really a big problem otherwise so long as you are satisfied with component temperatures.
I think (keyword) that 77C sustained with all 12 threads being used, at 3.9GHz, is fairly reasonable, but I'm a bit jealous of some of the other lower temps I've seen. I just wonder if I'm missing something here on how to cool more effectively. I'm considering tossing a 3900X in this bad boy, same cooler, and reusing the 3600 with stock cooler for a work PC upgrade (EMRs, dictation software, Office suites, music, imaging, etc). Just don't want to put the 3900X in a clearly poor situation, but if the 3600 just might not be a great bin, or if 77C is a solid temp for WCG/R@H 24/7, then I'll be content.
 

DrMrLordX

Lifer
Apr 27, 2000
14,457
3,433
136
I think (keyword) that 77C sustained with all 12 threads being used, at 3.9GHz, is fairly reasonable, but I'm a bit jealous of some of the other lower temps I've seen. I just wonder if I'm missing something here on how to cool more effectively. I'm considering tossing a 3900X in this bad boy, same cooler, and reusing the 3600 with stock cooler for a work PC upgrade (EMRs, dictation software, Office suites, music, imaging, etc). Just don't want to put the 3900X in a clearly poor situation, but if the 3600 just might not be a great bin, or if 77C is a solid temp for WCG/R@H 24/7, then I'll be content.
That's one of the more-extreme workloads you can throw at a CPU. I can't promise you that the same cooler will work on a 3900x (my guess is, you'll get lower clocks). But running a pressure-negative case isn't going to make things unworkable.
 

nOOky

Golden Member
Aug 17, 2004
1,535
335
136
FlareX is just b-die with profiles set up for AMD machines. The "XMP" profiles might not work on Pinnacle Ridge since I think they were set up for Summit Ridge. But you can hand-tune to get around that. That being said, I wouldn't flock to FlareX since you can get Micron e-die for less.

Hmmm I specifically picked the FlareX because of the Samsung chips, I don't really want to spend the extra $50 or so over the Crucial Ballistix, but I just want memory that runs stably at 3200 etc.
 

daveybrat

Elite Member
Super Moderator
Jan 31, 2000
5,223
547
126
Hmmm I specifically picked the FlareX because of the Samsung chips, I don't really want to spend the extra $50 or so over the Crucial Ballistix, but I just want memory that runs stably at 3200 etc.
How about just the regular FlareX CL16 kit, it works great with Ryzens and is cheaper than the B-Die CL15 ones.

 

amrnuke

Senior member
Apr 24, 2019
497
449
96
That's one of the more-extreme workloads you can throw at a CPU. I can't promise you that the same cooler will work on a 3900x (my guess is, you'll get lower clocks). But running a pressure-negative case isn't going to make things unworkable.
Thank you, I guess I shouldn't be doing apples-to-apples comparisons, but just seemed odd that a 3700X with an NH-U12 was getting 66C in Blender benchmark in Techpowerup's reviews, but that's of course without a hot 5700 reference card sitting right under the CPU tower cooler.
 
Feb 4, 2009
24,397
5,290
136
I ordered everything including the cheaper memory, hope the wife likes it!
I like my 3800x, I am confident she’ll like the 2700x
Only thing I don’t like about my Ryzen machine is the boot time isn’t super fast. Not real slow but not real fast either. Hard to describe.
 

Kenmitch

Diamond Member
Oct 10, 1999
7,660
757
126
Per my shopping around last year I **think** there is something weird about flare x memory and Ryzen 2 chips. Cant remember exactly but I think it won’t run at it’s rated speed.
I could be wrong so check the vendors memory support list.
Works fine on my rig. I did have a bad uEFI version or two that didn't want to play nicely with it.

I run mine at 3800 CL15 1T @1.42v's in 1:1 mode on my budget x570 MSI motherboard without any issues. Tested all the way up to 4333MHz, but the dang penalty hit going 1:2 mode kills the latency!

It's pricey for sure, but if you win the silicon lottery it makes it more bearable in the end. I ordered my kit before Zen2 was launched and was looking for something with good compatibility odds.
 
Feb 4, 2009
24,397
5,290
136
Works fine on my rig. I did have a bad uEFI version or two that didn't want to play nicely with it.

I run mine at 3800 CL15 1T @1.42v's in 1:1 mode on my budget x570 MSI motherboard without any issues. Tested all the way up to 4333MHz, but the dang penalty hit going 1:2 mode kills the latency!

It's pricey for sure, but if you win the silicon lottery it makes it more bearable in the end. I ordered my kit before Zen2 was launched and was looking for something with good compatibility odds.
Cool I wasn’t sure if I was correct
 

DrMrLordX

Lifer
Apr 27, 2000
14,457
3,433
136
Hmmm I specifically picked the FlareX because of the Samsung chips, I don't really want to spend the extra $50 or so over the Crucial Ballistix, but I just want memory that runs stably at 3200 etc.
b-die is awesome for what it is, though when you get the lower-speed bins like FlareX, it's a crapshoot as to how far you can push it . . . and on Pinnacle Ridge, it's not going to go faster than DDR4-3466 in most cases anyway (DDR4-3600 in edge cases). Micron e-die is a budget champion overclocker right now, routinely hitting DDR4-3600 CAS/CL16. You should be able to hit the same timings as the FlareX with e-die at DDR4-3200 pretty reliably. Just use Ryzen Memory Calculator and off you go.

Thank you, I guess I shouldn't be doing apples-to-apples comparisons, but just seemed odd that a 3700X with an NH-U12 was getting 66C in Blender benchmark in Techpowerup's reviews, but that's of course without a hot 5700 reference card sitting right under the CPU tower cooler.
Blender isn't as harsh as some of those DC workloads. That one that Larry had us run is . . . woof. Solid, non-stop AVX2 hammering your CPU from every angle. I can run Blender @ 4350 MHz with my 3900x, but I had to step back to something like 4150 MHz to run that PrimeGrid mess. And it still ran hotter.
 
  • Like
Reactions: VirtualLarry

amrnuke

Senior member
Apr 24, 2019
497
449
96
Blender isn't as harsh as some of those DC workloads. That one that Larry had us run is . . . woof. Solid, non-stop AVX2 hammering your CPU from every angle. I can run Blender @ 4350 MHz with my 3900x, but I had to step back to something like 4150 MHz to run that PrimeGrid mess. And it still ran hotter.
Good to know!

As I plan for a 3900X upgrade I'm wondering if I shouldn't go ahead and move to a Noctua tower cooler or even an AIO.

Also, man, there's Crucial Ballistix Elite BLE2K8G4D36BEEA 2x8GB 3600CL16 kit at MC for $120... or a Trident Z RGB Neo F4-3600C16D-16GTZNC same 3600CL16 open box for $91... I may be stopping by MC today.

Any recommendations? Was leaning Ballistix as I believe this is an E-die kit?
 

VirtualLarry

Lifer
Aug 25, 2001
46,870
4,554
126
If I had the money today (instead, gotta pay my rent), I would pick up a pair of those nice red Ballistix E-Die DDR4-3200 2x16GB kits, and run 64GB of RAM, hopefully @ 3600 CAS 16, on this R5 3600.

Just a note, after instituting an all-core OC of 4.0Ghz @ 1.3685V in BIOS, with AGESA 1.0.0.4 patch B (UEFI 3003 on Asus B450-F ROG STRIX ATX), my rig seemingly hasn't hard-hung or frozen (yet?) again.

Seems maybe that thread questioning the overall stability of Ryzen 3000-series CPU's auto-OCing features that constantly adjust freq. and voltage, in response to workload and temps, might have had some truth to it... at least in MY case. (I did run my R5 3600 @ 126C, stupidly, for like two hours at least, thinking my temp sensors were offset 20C for some reason.) So I may have "degraded" my chip, such that voltage/freq curves set at the factory may be offset by my degradation, and thus now unstable using the factory curve presets, and thus, my all-core OC provides a stable freq., and enough voltage buffer to prevent occasional rare crashes. Hey, it seems to work as a hypothesis.

Then again, I have yet, to try that PrimeGrid workload again. *shudder*
 

ASK THE COMMUNITY