• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Ryzen 3900X thread

mikeym said:
Did you have a go with CTR? I read that it is basically automated undervolting / all core OC tool. Therefore you lose single thread performance....

I do a lot of VR sim racing, including RaceRoom and Assetto Corsa, so single thread performance is important for me.

Will likely upgrade to 5900x if leaks/rumours are realised - may wait for a while though...
Click to expand...

I've been playing about with it but for the life of me can't get it to complete fully with out it hard power resetting the PC.
Diag completes ok, gives me what looks like decent figures and a lot of people do seem to be running it fine with good results.
Probably worth a go though as the lower volts and temp may still help you out.

People seem to be getting higher multi thread scores with Cinebench at lower voltages/heat levels. Though I guess, as you say, it probably affects single core a bit.

tbh I'd say just give it a go if you're interested and see what results you get but I do notice you have an Aorus mobo - I have the Pro version - and it seems that a lot of people are having issues with the board, though for some it seems fine.

Also if you've not already tried it, give the DRAM Calculator a go by the same guy as using that to lock in my mem timings gave me a huge uplift in mem performance which helps all round.

I've only just this week built this system so still playing about with it and need to do a fresh install of Windows at some point - just moved the old install over for now - but do have an og HTC Vive that I need to test out on the new setup.
 
Does anyone known what ideal idle temps in bios should be? Since I built the system 4 months ago, I was always seeing temps in the mid 40s in bios, and as high as mid 50s/60s under low/moderate load in windows. Had a friend look at it this past Monday, he said the chip is faulty (he was getting temps in the 20s), but I can’t remember if he said he used a 3900x or not, or an older chip. So when I get a replacement (probably won’t wait for amd RMA to go through, they haven’t gotten back to me yet asking for the current chip) what temps should ideally be looking for?
 
AnthonyA85 said:
Had a friend look at it this past Monday, he said the chip is faulty (he was getting temps in the 20s),
Click to expand...

your friend is most likely jealous and trying to throw you off the track and ending you selling the CPU at a discount thinking it is broken.

no way a CPU can idle at 20c while under conventional cooling. That’s lower than ambient or case temperature so you need some kind of refrigeration or liquid nitrogen or phase change system to get CPU for temp lower than ambient.

You temps are completely normal assuming you got a fairly decent air cooler or something like a 240 AIO. However if you are running the wraith cooler and getting those temps then it is pretty awesome.
 
pc-guy said:
your friend is most likely jealous and trying to throw you off the track and ending you selling the CPU at a discount thinking it is broken.

no way a CPU can idle at 20c while under conventional cooling. That’s lower than ambient or case temperature so you need some kind of refrigeration or liquid nitrogen or phase change system to get CPU for temp lower than ambient.

You temps are completely normal assuming you got a fairly decent air cooler or something like a 240 AIO. However if you are running the wraith cooler and getting those temps then it is pretty awesome.
Click to expand...

No, i mistyped my post (i was on my iPhone at the time), he tested my chip in one of his boards, with my AIO, and my chip usually has temps in the mid 40s while in BIOS, and I’ve seen it reach temps in the 50s/60s under low/moderate load in Windows. He said he was getting temps of 25 with a chip HE used, one of his own.

He also said the ssd, which was brand new, had a defective cluster on it which can’t hold data, and that is a brand new drive, he was able to install windows on on twice, but he had to do a long wipe to make sure it was completely clean. He used to build and sell computers for a living, before he had to close up shop.

It literally CAN NOT be anything else other than the CPU at this point. We’ve replaced EVERY. THING. in the system, except the GPU (which is working fine) and motherboard. The Motherboard was RMA’d to overclockers and they couldn’t find anything wrong with it, sadly i didn’t buy the cpu from them or i would have sent that too. That’s why i asked here what ideal temps i should be seeing. I’m using a H100i AIO. We’ve already sent the ssd back to amazon (a second Sebrant, I’m beginning to think SSDs with Phison memory controllers are just defective out of the box, that’s the second sebrent drive we sent back)

SOMETHING is stopping me from installing windows, i get a WHEA UNCORRECTABLE ERROR EVERY SINGLE TIME I TRY

So at this point it has to be the cpu, so, back to my original question: What are the ideal temps i should be seeing with an AIO cooler on one of these?
 
Mine idles around 36 to 38 in windows with spikes into the 50s in normal use. Gaming mid 50s with a 360m aio . Cb20 it hits 70 degrees just about.

Auto voltage it draws 142w performs worse and hits 80 degrees after sustained cinebench runs. I run 1.28175v fixed on my asrock taichi x470
 
@AnthonyA85 - Do you have another HDD /SSD you can try and install on - if nothing else may rule out your SSD as the culprit for windows failing to install.

Which Mobo is it ? BIOS up to date ? Have you checked your PSU is OK ?
 
No spare drives to test, only one i have is a 3.5 mech drive in a sealed enclosure, and tearing that apart to get the drive out might damage the drive.

First nvme i bought for this system we sent back, believing it was faulty, now that i think about it, it might not have been, but i did get a few errors when trying to load parted magic couple of weeks ago before we sent it back and got a replacement, which apparently had a defective cluster, so that went back too.

The old psu which I transplanted out of my 5yo intel system was a cooler master silent hybrid, 850 wats gold cert, but it was 8 years old, the new one is a Corsair RMi 850 watts, had it a week or two. The old ram from my intel system was not on my AMD board’s QVL list, it was rated at 2666HMz XMP, and Memtest DID show two errors with that ram in this system, after the initial crash, so that was replaced with some QVL approved ram, I ran memtest on that twice, both tests passed all stages with no errors. The ram is fine.

RMA’d the motherboard to Overclockers and they tested both with a 2.5 inch SSD and nvme m.2, both tests passed, they were able to install windows and run multiple stress tests, so the board is fine.

We’ve replaced just about everything that could be causing the crashes, except the cpu and gpu. The GPU is a 2070 Super, and if that was faulty, i likely wouldn’t get any display out of it. I do have a 7950, that’s about 7/8 years old, but it failed last year, hence why i got the 2070.

Unfortunately i did not buy the cpu from OCUK, or i would have already sent it back to them. I got it from amazon, and i can’t return it (doesn’t give me the option), and although i did contact the seller, haven’t heard anything back (and probably won’t). I sent AMD an RMA request directly, but have yet to hear from them. I’ll give them until monday at the latest, and if i don’t hear from them, I’ll just order a new chip.

My board is an asus x570 I Gaming, BIOS version 2606, latest version. I can not roll it back, I’ve tried, there is no BIOS flashback feature on this board, even though the manual says there is. There is also no Q CODE LED display, which probably would have helped me fix this issue weeks ago if my board had one. Seriously, those things should be standard on ALL form factors. I’d be more than happy to give up the onboard HDMI/Display Ports (which i don’t use) in exchange for a Q Code LED.
 
@AnthonyA85 Seems normal. On a 3950x here but it behaces the same either under a D15 or or a 360 AIO. BIOS temps will be higher because it won't fully downclock. I found applying a manual overclock in BIOS or with CTR evens the temp spikes out but you lose single and lightly threaded performance so I just leave at stock with auto volts and set the fan curve so they don't ramp until +60C. Board/VRM temps are low so nothing is cooking.

Oh the 1usmus power plan and suggested BIOS settings in the install instructions helped too.
 
Well, I’m using a 240 AIO. Wondering if i should replace the paste on it when i get the new chip (I’ll also be getting a new nvme, gonna go samsung this time, different memory controller), unfortunately, whole lot is gonna cost me about £528, with an OEM chip (don’t need the wraith cooler)

My specs are in my signature.
 
Could I get a sense check please chaps?

When it comes to daily driver settings for my 3900x, in the BIOS I change:
  • RAM speed from 2400 to 3600
  • RAM timings from Auto to 14, 15, 14, 28
  • RAM voltage from Auto to 1.5

Literally everything else is left Auto to save myself the hassle.
Should I be considering an all core manual overclock?

O4HS3lt.png
 
Last edited:
ceaton88 said:
Could I get a sense check please chaps?

Should I be considering an all core manual overclock?

Click to expand...
I can recommend going to Techpowerup!'s website and reading the guide on Ryzen memory overlocking. Its written by the guy who made Dram calulator for Ryzen and Ryzen clock tool. Its the order in which to tweak which is completely missing in Ryzen Dram calculator. ProcODT (processor On die Termination) should be tested thouroughly. I went from 53 to 40 and it got rid of the 11 errors I got with running tm5 2 hour test.
 
megatron said:
I can recommend going to Techpowerup!'s website and reading the guide on Ryzen memory overlocking. Its written by the guy who made Dram calulator for Ryzen and Ryzen clock tool. Its the order in which to tweak which is completely missing in Ryzen Dram calculator. ProcODT (processor On die Termination) should be tested thouroughly. I went from 53 to 40 and it got rid of the 11 errors I got with running tm5 2 hour test.
Click to expand...

Cool thank you will check it out although my memory is stable. I picked up RAM testing software from Karhu and its amazing, also been running these settings for a while now without issue so confident I am stable tbh.
 
megatron said:
Did you test during the summer heatwave? 1.5v is a lot, and you may get thermal runaway (in the sence of stability not killing the chips).
Click to expand...

I can't remember if it was heatwave but was in June this year, I realise 1.5v is probably the maximum but only way I can run those timings @ 3600 (I couldn't even get past 3200 on my previous 1600x chip)
 
megatron said:
I can recommend going to Techpowerup!'s website and reading the guide on Ryzen memory overlocking. Its written by the guy who made Dram calulator for Ryzen and Ryzen clock tool. Its the order in which to tweak which is completely missing in Ryzen Dram calculator. ProcODT (processor On die Termination) should be tested thouroughly. I went from 53 to 40 and it got rid of the 11 errors I got with running tm5 2 hour test.
Click to expand...
that process he described in there is so convoluted, no chance average person or even an expereinced overclocker is going to do it.

going into bios and change one of the 50 settings at a time until you get it working and each time you have do CMOS clear is just a no goer. thats why I am not even bothering with ram overclocking anymore. it is just not worth the time and flaff...
 
pc-guy said:
that process he described in there is so convoluted, no chance average person or even an expereinced overclocker is going to do it.

going into bios and change one of the 50 settings at a time until you get it working and each time you have do CMOS clear is just a no goer. thats why I am not even bothering with ram overclocking anymore. it is just not worth the time and flaff...
Click to expand...

I've come to the same conclusion, very time consuming. Really could do with some software that can test automatically and tell you what to set it at with no user input.
 
You must log in or register to reply here.
Back
Top Bottom