Memory ASRock DeskMini X300 - RAM question

[HydrAxx747]

Those timings are quite impressive 3800... cl16 and 54ns is extremely good

I hope as much as possible that the kit can hold @4400MHz (IF 1:1 of course) with latencies 18-18-18-... or 19-19-19-... but maybe I asking too much (especially with only 1.35v vDIMM), but if only it could hold 4000MHz in CL17-17-17-... or 4266MHz in CL18-18-18-.../CL19-19-19-... that would be really awesome!

 

[msystems]

Any progress with it?

 

[HydrAxx747]

Any progress with it?

Sorry, for the late response I was on the weekend away from home ^^

And no, no progress because the sender of the package forgot to write the name of my city so the package has to go around in circles for a week because in the department where I live (Vendée in France), no less than 9 cities have a street with the same name as mine. But hey, now the problem is solved with UPS and I receive the package without fail tomorrow afternoon, I can't wait to see what they will give!!

On the other hand, I just noticed on the screenshot that I posted a little earlier, the user managed to lower the values of Trfc to 304/192/132, I had never seen such low and stable values in DDR4 SO-DIMM on the Deskmini A300/X300, even with my Corsair Vengeance 16GB(2x8) 4000MHz CL19 kit... These are values worthy of Samsung B-Die kit in standard UDIMM format, what do you think?

 

[msystems]

Only Bdie can do this low trfc, its lower than mine also (320).

 

[HydrAxx747]

Only Bdie can do this low trfc, its lower than mine also (320).

Hello, finally some news for you about the Corsair Vengeance DDR4 SO-DIMM 32GB(2x16GB) 3000MHz CL16 kit, I was waiting for the arrival of 4 heatsinks (Those of 0.5mm copper with graphene coating) for these ram sticks, because during my first tests at 3800MHz and original XMP timings (CL16-18-18-39-1T) under 1.35V voltage, I already reached almost 60°C on the stick closest to the CPU) after 3 passes of TestMem5 passed (stable), so I preferred to wait for the delivery of the heatsinks before pushing further.

I had forgotten to take into account the fact that there would be twice as many B-Die chips per side(compared to the Vengeance 4000MHz CL19 of 8GB per stick) but the RAM stick temperature sensors quickly reminded me of this ;-)
So I received the heatsinks today and installed them this evening, so I will do more tests in the days to come, but I still have a few photos before the heatsinks were installed for your curiosity.

From the serial numbers of the two modules, we can deduce via their first four digits, their datecodes "2006" (and "1934" on the PCBs of the two RAM modules), which means that they were produced and / or leaving the factory around mid-February 2020, so what surprises me is that practically no DeskMini A300 or X300 user has paid attention to this kit, because if we take a closer look at the specifications of origin of this kit, we realize that despite its frequency of 3000MHz, which is not excessively high, but to hold this frequency under relatively tight latencies of 16-18-18-39-1T and especially under a voltage of only 1.20v (standard JEDEC voltage), which no other DDR4 SO-DIMM kit is capable of apart from other high-end kits equipped with Samsung B-Die chips such as your Corsair Vengeance 3800MHz CL18 kit and the mine at 4000MHz CL19.

Regarding the RAM modules, these are "E0" PCBs as for your "Vengeance 3800MHz CL18" and my "Vengeance 4000MHz CL19", I would have liked it to be rather equipped with the revised "E1" PCB (Equivalent to "A2/B1" PCB compared to "A0/B0" PCBs in RAM Desktop) but hey, it's already good for a kit (in new condition) that I paid around 100usd and which seems quite promising in terms of overclocking. To be continued^^

 

[msystems]

Interesting, excellent find and information. Can't wait to see the results.

 

[HydrAxx747]

Interesting, excellent find and information. Can't wait to see the results.

So I finally had time for myself to carry out my first tests (finally a vacation, I haven't had one since last September, thank you for the Christmas vacation).

On the other hand, I performed these tests on the Beta "L1.79" BIOS and not with the latest "L1.80A" despite the fact that I had downloaded it almost when it was published, and since then I had completely forget so I would do other tests in the days to come.

So, for the overclocking results of this Corsair Vengeance SO-DIMM 32GB(2x16) 3000MHz CL16 @1.2v kit, they are really good because at the moment I managed to overclock them starting from the settings of the other overclocker which has the same kit of which I had published one of its screenshots earlier in this thread, i.e. 3800MHz@1.35v CL16-16-16-16-32-48-1T-etc. and I therefore managed to climb in a stable way (verified with TestMem5 and other stress tools) up to 3933MHz and this with exactly the same latencies as those visible on the screenshot in question, which is really impressive but it's after that things go wrong.

So I wanted to continue on this good momentum which took me a lot of hours to check the perfect stability, and it's when I want to go to 4000MHz or more, that there "blackscreen" and impossible to start , and which forces me to "Clear CMOS" manually by shunting the 2 dedicated pins (luckily I installed a quick and easy-to-access switch for this), and so the first time this problem happened I thought say that it was the latencies that were very tight, so I relaxed them to 18-18-18-18-36-54-... as well as the other secondary timings and the different tRFC values too, results: nothing to do "blackscreen" no boot and force ClearCMOS again, so I relaxed even more by completely resetting the XMP default values of the secondary timings and even trying to leave most of the timings settings on "AUTO" so that the system itself chooses the settings to increase the chances of starting, but nothing 'there it stubbornly refuses to start (blackscreen) as soon as I set to 4000MHz (2000MHz in the bios of course) or more.

And so for the moment I have not tested further and so I will try other things to adjust tomorrow to try to unblock the situation, for example I have not yet tried to increase the value of the VSOC voltage and others VDDx voltages which could be related, after that it's weird anyway because I have the impression of having a kind of invisible wall from 4000MHz, whereas with my other kit Corsair Vengeance SO-DIMM 16GB (2x8) 4000MHz CL19 @1.35v at Stock, I had no problems at 4000MHz and I could even push up to 4333MHz and it still started but it was unstable though. Afterwards I say to myself: maybe it is the memory controller of my Ryzen 7 5700G which is much more stressed by the overclocking of RAM modules having 8 chips per rank instead of 4 chips per rank that there is on the Corsair 16GB 4000MHz kit in 2x8GB?

If you ever have any ideas or advice to give me, I'm all ears ;-) of course.
But unfortunately tired of this succession of failures, so I stopped to resume tomorrow as I said rather but I still forgot to make a copy of the overclock and stability validation screenshots and I unplugged everything and put away the x300 to reconnect and go back to my main config, but I promise the screenshots will be there from my next post and without fail ;-)

Well after a STABLE 3933MHz in CL16-16-16-16-32-48-1T and other secondary timings (as on the other guy's screenshot), it's still not too bad for a first 32GB SO-DIMM kit overclocking session, I just hope to be able to go up to 4000MHz and a little more (4266MHz would be really great already!)

To be continued... ;-)​

 

[msystems]

Hmm, that is unfortunate it will not do 4000, but fortunately 3800 cl16+ is already quite good value. And 4000 cl18 is not much better anyway even if it was successful. Perhaps the memory controller can't handle it. I assume the issue isn't a temperature problem since it won't post.