• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

Overclocked budget CJR with the least amount of VDIMM possible to a pedestrian 3600 CL16

Joined
Jan 5, 2024
Messages
37 (0.07/day)
Location
South East Asia
System Name Promontory's Last Hurrah
Processor Amd Ryzen 7 5700x
Motherboard Gigabyte X570 Aorus Ultra
Cooling Thermalright Frozen Warframe 360 SE
Memory Fangxiang UD11 (Sk Hynix CJR) DDR4-3600 CL(16-20-20-38) 32GB
Video Card(s) Emtek Turbojet RTX 3070
Storage 1TB Colorful CN700 , 1TB Ediloca EN855 , Toshiba MD04ABA400v 4TB hdd
Display(s) Acer SA22 100hz IPS
Case FSP CMT580B
Audio Device(s) Nicehck YD520/ Moondrop Space Travel
Power Supply Darkflash Perfectmost 750W
Mouse ML6 Pro
Keyboard GarudaGear GG100/ Langtu LT104
Software Windows 11 24H2
I overclocked my budget sticks of 8Gbit Hynix CJR to see what I can get by with as little as VDIMM added as possible compared to the XMP settings that it comes with default. The Thaiphoon burner screenshot is shown below for reference.
1750515842820.png

I tried setting it loose at 3600 16-22-22-42 and it booted, albeit with very loose TRFC and other secondary and tertiary subtimings as compared to 3200CL16 XMP.

I tried really aggressive settings afterwards, and my pc only booted once and ran Testmem5 with the Absolut preset loaded (And got an error #1) before I got constant winload.efi checksum mismatch errors at boot. Even raising VDIMM close to 1.4V didn't do anything as it was too tight for my CJR sticks. Too bad I didn't have screenshots of that ordeal as it was quite funny to behold.

I tried if I can get away with copying most of the 3200CL16 timings at 3600, and lo and behold it worked. I am pretty sure I can bring some of the timings a bit lower, but with only +.02 VDIMM voltage added to the stock 1.35v XMP voltage, I am quite glad it worked. I ran 2 cycles of Absolut on it and then let my sibling play Once Human to make sure it is still stable even if the GPU is heating up the PC Case interior.
And it is still stable until now with no stuttering or WHEA errors.

I must say, the older DDR4 topology on Gigabyte's 1st release X570 motherboards is not as good as their 2nd gen memory topology found in their Rev 1.1/2 boards in X570/B550 with using my B550 Aorus Pro as reference coupled with buildzoid's experience with the B550 Vision D. Even my comparatively cheaper and simpler B550m Pro4 did not immediately resulted in errors and corruption when I tried similar timings with the same memory die, and that was with a Ryzen 5 3600 which has a weaker IMC compared to my 5700x.

If you can suggest better timings feel free to reply on this post.
1750514416835.png
 
Last edited:
Hi, I have an older OC that was done on another forum. Please excuse the imperfections of this crude copy paste. I can straighten this out later.

This is a budget kit of PNY 2666mt/s kit which was actually an XMP profile. None of this is tuned for performance, just the possibilities of some DDR4 ram kits. Thought this was a fun experience, didn't have to go nuts with any votlages or anything, purely as for example OC only.

_________
Edited for fitment.
>>

Thaiphoon burner doesn't seem to read? (I'll show you here how to get the information. Be it complete or not....)
But just to let you know, this will be a fresh start OC ABOVE XMP rated frequency.

The Kit. MD16GK2D4266616XR

Speed Compatibility - 2666MHz, 2400MHz, 2133MHz - 1.2v

The Look - Not horrible, but actually reminds me of that Oloy memory if any of you are familiar with it....
Which this kit would be about the same quality if given a similar speed and timing set.
Does NOT need these heat spreaders. Totally just for looks.


1402244297_PNY2666.thumb.jpg.598a2a4370e11cab2f5d5c8c5c7846b3.jpg



The Thaiphoon Burner - Well as you will see here (first spoiler) Eeprom tells us nothing.
But if you hit that report button.(second spoiler below)...Which will also tell you the PCB revision of your memory kit as well.
As you'll notice, the memory kit is running XMP. I have used it for general purpose and F@H. So far so good. Stable. No heavy load testing yet.
(testing using OCCT memory linpack screen shots in the OC posts below)
Played some Need for speed at most for the gamers out there. But nothing heavier than that.




51240803_Thaiphoonburnermissingdieinfo.thumb.png.a3996cea6c6c63fe7de2e94fd02a580f.png

1181988725_ThaiphoonburnerDieandpcbinfo.thumb.png.d01afdb68b6174cd51b902d53d1a2c73.png


Adia64 cache and memory benchmark *Located in the tools tab at the top of AIDA64* (Our initial measurements)

XMP rated 2667 16-18-18-35 2T 1.2v
56ms latency but only 35k mb/s read writes.
Decent latency and poor bandwidth.

1537304232_2667mhzXMPAIDA64I.png.7694c2ae2ce6f8a994dd6c56bf9038b0.png


Check HWInfo64 for Dram temps.
=None.
Voltage steady good 1.2v
VCCIO 1.152v (bios set 1.1150v)
System Agent 1.148v (Bios set 1.1150v)
VCCIO and SA increased for overclock. Also used for daily on B-die, I did not change the values.
Are higher than input via bios, but that's how Asus boards roll. OV the shit out of everything.
With XMP, on this board, they will default over 1.4v - So users, keep these manually checked!


871807317_NoDRAMtemps!.png.7f43b1314a5bb1c6236592ae955d3933.png

First OC - Simple
XMP enabled. Adjusted manually Dram Frequency to 2700mt/s.
Set the Dimm voltage to 1.25v manually in bios. F10 save.

Cpu 8700K stock default cpu settings. (Testing memory only)
Quick OCCT linpack memory 20 minutes.
Looks pretty stable. Daily tasks, browsing, watching video and gaming is fine.

Memory latency lowered just a little, but the frequency change wasn't enough to boost the bandwidth in any significant way. On the right track so far with minimal changes. (timings on auto)



594666603_PNYOCI2700mhz.thumb.png.8123e500553d5e176f40c49907b8e0d8.png

The next available frequency in bios is 2800mt/s.
Will try this one Dram frequency change only, 1.25v
Figured I'd drop a validation this time. what the heck right?
https://valid.x86.fr/cj48r2



1449616687_PNYOCII2800mhz.thumb.png.9c0b6cdab999c09c7718cee91ce00cee.png

Testing again simple OC, timings auto.
Notice tRFC is getting looser. Results from timings on auto.

2900mt/s (1450mhz effective)16-18-18-35 1.25v -
Another Validation -
https://valid.x86.fr/vz5f3c
53ns latency, bandwidth creeping up with frequency

1500662341_PNYOCIII2900mhz.thumb.png.c40c0590e86d3c4ebdb8b27787a631a2.png


16-18-18-35 is only good for 2933mhz 1.25v
This timing set created errors at 3066mhz. 1.35v

.......
So changed the timings instead. Testing crudely.
18-20-20-40 1.35v (all other timings on auto xmp still enabled.)
3200mhz. And looks pretty good. Not the greatest results, but making some bandwidth.


PNY OC IV 3200mhz.png
3333mt/s (1666mhz) 18-20-20-40 1.35v
Passing 20 mins OCCT and feeling good about the latency here. Getting close to under 50ms, which is pretty decent for most of the timings set at auto.
541880907_PNYOCV3333mhz.thumb.png.aa136d9480daf26ab1eb0b6a76184388.png



3500mt/s 20-22-22-42 1.35v
20 minutes again. Rest on auto, cpu stock.

791337413_PNYOCVIII3500mhz.thumb.png.91dd4ff3bdab68f40e4cc095bece3973.png



3600mt/s 20-22-22-42 1.35v
Tested OCCT about 50 minutes (free version is only 60 minute restricted)

So here it is. I'll be pushing further though. No reason to stop yet. Not quite 1000mt/s OC yet.
But am shocked at the results so far.
930209662_PNYOCIX3600mhz.thumb.png.3c6691a14d33942800e41adb44ea19a7.png



3700mt/s - 20-22-22-42 1.35v
All other settings auto.

It likes that 20 cas!

797012957_PNYOCX3700mhz.thumb.png.ab55db9861007fe46e089384810c597c.png


3733 fails under a minute.
Example -
And will loosen Cas to 22 and leave the rest.
Next after this 22-22-22-42 and see if it will get up. 1.35v.

2000681604_PNYOCXI3733mhzERROR.thumb.png.67e5a2c723b4b02d626f8538d7229a52.png


Had to go looser.
The rest still on auto.
Now at 3733mt/s 22-24-24-44 1.35v
44696428_PNYOCXII3733mhz.thumb.png.1d3c119850d6deb822467723ae2f4f6f.png


THIRTY 8 HUNDRED -
22-24-24-44 1.35v
567mhz OC.
10531611_PNYOCXIII3800mhz.thumb.png.484199bcbb71db0e30b38fce54ef0e44.png


3866 is a NO GO.

Tried fail post count 5 and 6 tries. Standard is 3. This gives the PC extra chances, you never know it might post up on one of those attempts! So it's an old trick I use for adventures like these.
Tried memory retry button several times.
Tried 24-26-26-46
Tried 1.4v / 1.45v
NOTHING- all above just fail safe to defaults.

3800mt/s is good to go. Posts right up.
22-24-24-44 1.35v is just dandy.

Will tweak from here. Maybe improve the performance a touch, maybe not. So far more voltage does nothing.
Maybe even some 3D comparisons.

...Although Im not sure would be good for comparison. 3800 untweaked vs tweaked or just xmp vs 3800 tweaked??


No stability testing yet.

Base 2667mt/s XMP profile
16-18-18-35 1.20v

Hide contents
1491774050_2667mhzXMPAIDA64I.png.202b1dc5d61606f9e9148a7067c118f3.png


3800mt/s
22-24-24-44 1.35v
TRFC 665 to 605
Dram Write latency 19 to 16
(did not change average scores, could have left these on auto)
HT enabled

Hide contents
880993909_3800mhztestI.png.6a10c7b331eb86f990f1b2bb1a54ff8a.png


Same as above with HT disabled.

Hide contents
468424715_3800mhztestIINoHT.png.756e0bb7c718830fe762a54eda1346d9.png


This is a very considerable improvement over the 2667mt/s profile despite the Cas-22.
Not horrible for a pretty cheap under-dog kit imo.

Oh, the HT enabled/disabled is to demonstrate the L3 cache outcome, in case any where wondering.
 
Last edited:
If you want a straightforward comparison, just compare stock xmp to your new, tweaked settings.

I should really get a 10th/11th gen based system sometime, just to learn more about memory OC.
 
If you want a straightforward comparison, just compare stock xmp to your new, tweaked settings.

I should really get a 10th/11th gen based system sometime, just to learn more about memory OC.
The bottom 3 screen shots represent exactly that actually.

You can do DDR4 on AMD similar to Intel. Just depends on the kit and the board, which you did mention and know about.

Typically, in short, CJR clocks OK with looser timings and doesn't respond greatly to a lot of v-dimm. DJR might be a better option if you wanted to get like 4600mt/s on DDR4, then yeah an Intel platform might be better suited.
 
Good B-die is still expensive here (similarly priced to 32GB DDR5-6000 kits), and I'm pretty satisfied with what I got now. I'll just probably buy a cheap 9/10th gen 1dpc motherboard and the cheapest i7 i can get my hands on. Or an APU as APUs clock really high compared to their igpuless brethren.
 
Good B-die is still expensive here (similarly priced to 32GB DDR5-6000 kits), and I'm pretty satisfied with what I got now. I'll just probably buy a cheap 9/10th gen 1dpc motherboard and the cheapest i7 i can get my hands on. Or an APU as APUs clock really high compared to their igpuless brethren.
I will give a you a part number for a kit that I know does 4600mt/s CL20
Used it with 13700K / Mem 2266mhz CL20-24-24-44-68 1T / MSI Z690_A ddr4
CMW32GX4M2D3600C18
image_id_3069738.jpg
 
Back
Top