My g.skill ddr4 3600 4x32gb 1.45v modules are dropping like flies

[xorbe]

X570 + 5800X3D + g.skill Trident Z Neo

Bought 4x32 ~18 months ago. Could never get it all stable and working at once.
Gave up and ran 2x32 at 3200. Turns out one of the sticks was bad (fails even at 2666, any slot).
Explains sporadic software issues I've had the last several months.

Got the other 2 sticks out that were not used, and everything was great! For 11 days no software crashes.
Then one of the sticks up and died today, prime95 errors within seconds again even at 2666.

So now I have 2 remaining sticks that still work. Was there a bad batch of g.skill ram that I missed?

 

[VuurVOS]

X570 + 5800X3D + g.skill Trident Z Neo

Bought 4x32 ~18 months ago. Could never get it all stable and working at once.
Gave up and ran 2x32 at 3200. Turns out one of the sticks was bad (fails even at 2666, any slot).
Explains sporadic software issues I've had the last several months.

Got the other 2 sticks out that were not used, and everything was great! For 11 days no software crashes.
Then one of the sticks up and died today, prime95 errors within seconds again even at 2666.

So now I have 2 remaining sticks that still work. Was there a bad batch of g.skill ram that I missed?

Did you check the serial number if all the sticks contains the same memory chips?

 

[xorbe]

Rats, my two kits came with different chips.

The failed sticks are S821M ("Hynix MJR").
The two that (still seem to) work are S820M (also "Hynix MJR").

32GB 3600 16-22-22-22-42-64 1.45V 2T GDM-off

 

[Assimilator]

DDR4 baseline voltage is 1.2V, 1.45V is more than 20% overvoltage. That's murderous for DDR4 DRAM, no matter what G.Skill tries to sell you.

 

[3x0]

DDR4 baseline voltage is 1.2V, 1.45V is more than 20% overvoltage. That's murderous for DDR4 DRAM, no matter what G.Skill tries to sell you.

Only Samsung C-die is sensitive to voltages above 1.35v.

 

[xorbe]

If the 2x32GB Hynix S820M 3600 also fails, I have both 2x16GB 3200 and 3600 Samsung B-die (8810B) here which either would surely work.

 

[VuurVOS]

32GB 3600 16-22-22-22-42-64 1.45V 2T GDM-off

The failed sticks are S821M ("Hynix MJR").
The two that (still seem to) work are S820M (also "Hynix MJR").

Did you buy two dual channel kits?

If it is a dual channel kit, then I would RMA the faulty kit if it does give you issues when that kit is only installed.

 

[xorbe]

Hey is there something about 3600 memory speed that can cause USB to misbehave? Anything in the BIOS that I should have been adjusting? (X570 + 5800X3D)

 

[iameatingjam]

DDR4 baseline voltage is 1.2V, 1.45V is more than 20% overvoltage. That's murderous for DDR4 DRAM, no matter what G.Skill tries to sell you.

Oh geeze you are making me nervous about my 1.5v oc, but its ddr5... I'm guessing thats not better... its hynix m if that matters?

Not only that but this is the second time I have heard of ram dying from too high voltage today. The other time was on fb and the voltage was only 1.5v.

--

OP... best to buy kits made to be run together. You can always sell your old ones. Ram usually sells pretty well on the used market I find. When in working condition that is.

 

[ir_cow]

Depends on the ICs what the max safe voltage. 1.6v will be okay for Hynix DJR but you do that to Samsung C die and it will be dead probably within minutes

 

[tabascosauz]

Hey is there something about 3600 memory speed that can cause USB to misbehave? Anything in the BIOS that I should have been adjusting? (X570 + 5800X3D)

USB issues can come from a mix of higher Fabric speeds / not enough VSOC for that speed / not enough/bad VDDG_IOD configs for that speed

Or it could also just be a board/CPU problem, which is pretty common and unfortunately not really solvable if you landed on an unfortunate sample.

I dunno why you ran 1.45V through your DDR4. Only B-die, Rev.E/B, and DJR actually benefit in any way from higher volts and are also safe to use at higher volts. Even 8Gb CJR (Hynix' second best) was not recommended to run at >1.4V, if that, and until DJR came along Hynix didn't have a single high VDIMM tolerant IC. Amongst the vast remainder of the mediocre trash on the market (of which MJR/JJR are definitely candidates), it's hard to say which ones will die from high VDIMM but not unreasonable to assume and act like they all do, as a rule.

For G.skill Hynix, fourth digit being 0 or 1 doesn't indicate different chips. The rest of the code still points to 16Gb MJR, only possibly a newer revision, maybe by G.skill internally. Hynix still only has 1 real high end DDR4 IC that is both voltage-scaling and voltage-tolerant - DJR - and this ain't it. All new Hynix in the past 2 years that isn't DJR has behaved about the same - poor timings performance (especially at sub-4000 speeds) that does not improve with voltage, and zero to minimal freq headroom.

If you are using P95 as a memory test, it sounds like you are also long overdue with acquainting yourself with either TM5 or HCI. P95 large FFT has its uses but it is not a first line test.

 

[A&P211]

Oh geeze you are making me nervous about my 1.5v oc, but its ddr5... I'm guessing thats not better... its hynix m if that matters?

Not only that but this is the second time I have heard of ram dying from too high voltage today. The other time was on fb and the voltage was only 1.5v.

--

OP... best to buy kits made to be run together. You can always sell your old ones. Ram usually sells pretty well on the used market I find. When in working condition that is.

Whats the speed that your ram runs at, and why 1.5v. Doesnt DDR5 normally run at 1.1v.

 

[xorbe]

USB issues can come from a mix of higher Fabric speeds / not enough VSOC for that speed / not enough/bad VDDG_IOD configs for that speed

Okay I found some suggestions on reddit about vsoc (1.1) / vddp (0.90) / ccd (0.95) / iod (1.05) / pll (1.85), and will see how that goes. Otherwise I guess my chip will need to drop from 3600 to 3200 ram.

 

[tabascosauz]

Okay I found some suggestions on reddit about vsoc (1.1) / vddp (0.90) / ccd (0.95) / iod (1.05) / pll (1.85), and will see how that goes. Otherwise I guess my chip will need to drop from 3600 to 3200 ram.

1.1V for 3600 is middle of the road - however, your setup is quad rank if you run all 4 sticks of MJR. Much more UMC load than most people are running, not so much about Fabric. I'd allow it up to 1.15V if needed, and adjust IOD accordingly (-0.05V from VSOC is fine). PLL usually does not help, no point trying it if you haven't done the other more obvious steps.

You really should be running something proper like TM5 to make sure the profile you are running is actually stable.

Memory Testing with TestMem5 TM5 with custom configs

Hello everybody I am just making a very light tutorial with a collection of custom config files and a DOWNLOAD LINK for TM5 v0.12 anta777 absolut config Official Intel DDR4 24/7 Memory Stability Thread None of the work is mine but it seems like a pretty good and fast testing app Update...

www.overclock.net

 

[xorbe]

No I'm only running 2x32GB (also 2 of the 4 sticks died so I only have 2!) I will look into TM5 and HCI.

 

[iameatingjam]

Whats the speed that your ram runs at, and why 1.5v. Doesnt DDR5 normally run at 1.1v.

5600 cl36>6400 cl30. Some knowledgeable people on this forum seemed to think it was okay, except for it potentially shortening lifespan. And... well, it has been okay so far at least, but I've had this config for quite a while now.

Edit: That voltage has been like that since June 1st. Last week one of the rgb lights kept going on and off. Thought it might have been related to voltage. But ever since reseating its been fine.

 

[TheoneandonlyMrK]

DDR4 baseline voltage is 1.2V, 1.45V is more than 20% overvoltage. That's murderous for DDR4 DRAM, no matter what G.Skill tries to sell you.

First thing I thought wtaf 1.45.

My G skull sit at 1.38 max at 3800 and I have seen no ddr4 kit use or ask for 1.45 also 1.4 is the max the imc likes to deal with.

 

[basco]

over 200 kits ddr4 and 5 with 1,45v to 1,60volt

Arbeitsspeicher (RAM) Speicher mit Übertragungsrate ab 3200MT/s, Betriebsspannung laut Spezifikation: 1.45V/1.5V/1.55V/1.6V Preisvergleich Geizhals Österreich

Preisvergleich und Bewertungen für Arbeitsspeicher (RAM) Speicher mit Übertragungsrate ab 3200MT/s, Betriebsspannung laut Spezifikation: 1.45V/1.5V/1.55V/1.6V

geizhals.at

 

[GerKNG]

DDR4s JEDEC specification is 1.2V
Samsung own specifications are up to 1.5V for b-die as an example.

 

[mechtech]

X570 + 5800X3D + g.skill Trident Z Neo

Bought 4x32 ~18 months ago. Could never get it all stable and working at once.
Gave up and ran 2x32 at 3200. Turns out one of the sticks was bad (fails even at 2666, any slot).
Explains sporadic software issues I've had the last several months.

Got the other 2 sticks out that were not used, and everything was great! For 11 days no software crashes.
Then one of the sticks up and died today, prime95 errors within seconds again even at 2666.

So now I have 2 remaining sticks that still work. Was there a bad batch of g.skill ram that I missed?

Interesting.

I would expect that 3600 could be done with 1.30V, depending on the timings that is. My 3200 runs at 1.20V albeit at jedec timings.

 

[ThrashZone]

Hi,
I'm guessing this is your memory

https://www.amazon.com/G-Skill-Trident-F4-3600C16D-64GTZN-2x32GB-3600MHz/dp/B08MFMXVQL

1.45v is pretty wild for 3600c16 on wildly loose timings but 32gb sticks hynix land

Voltage can be bad long term but dimm temps is why they start to show errors and or die prematurely
Think error zone starts at 45c +-

Most of my ddr4 is 8gb sticks at 3600c16 1.35v but b-die with tight timings
I have one set of 4000c16 at 4000 at 1.4v 2x16gb also b-die again tight timings 16-16-16-36

No luck on rma ?

 

[xorbe]

I haven't tried to RMA the 2 failed sticks yet, using the 2 remaining "good" ones. Edit: Just put in for an RMA request.

My current issue which is new, is after flashing the motherboard bios to the latest version + latest AMD drivers, is that either at boot-up w/drive already plugged, or at the Windows desktop plugging in my USB drive, can possibly result in its access light endlessly blinking, and the computer doesn't boot until USB is unplugged, or the Windows desktop goes catatonic until the USB is unplugged. Nothing is logged in the Event Viewer so it seems to be a low level issue. Otherwise everything else is working great, which is an improvement over last month.

I need to pop in some 3200 2x16 and see if this (hopefully) final problem disappears. Maybe 3600 2x32 is a no-go with the hardware I have. Not a big deal, but you know how it is.

Edit: Oh there's even a newer bios w/new agesa for my Asus Dark Hero as of Nov 1st, off to flash again and start over!

Edit: Already gave up because it didn't post once, it's just not happy with 3600 2x32.

Edit: 3600 2x16 also triggered the overnight cold start USB endless blink. Suggesting the problem is likely 3600, and not 2x32. Next up, 3200 2x16 Samsung B-die.

 

[xorbe]

Okay, so it seems I did solve the original ram and USB issues. I didn't detail the USB woes here, but I was having USB go catatonic in Windows. I flashed the motherboard from Mar'22 to Jul'23 bios (then to Oct'23), and installed the latest AMD drivers, and those particular USB problems immediately went away.

What appears to be happening now is the USB sometimes hangs on initial handshake (endless blinking) which is a new problem after the bios update. It happens with 2x32 3600, 2x16 3200, and JEDEC Crucial 2x8 2133. I think we can rule out memory / high clocks as the cause of the blinking USB.

Additionally, when I turn off the PC, the USB keeps blinking until the PSU caps stop supplying 5V! Maybe I can work around the issue with a different SSD enclosure. Currently it's a UGreen USB-C "3.1 Gen 2".

Anyway, G.Skill approved my RMA request for the 2 failing ram sticks. Seems like my remaining USB issue isn't memory related.

================================================================

edit: Solutions I've concluded wrt my USB problem for my SSD enclosure:

1) Avoid SuperSpeed10 USB ports (what I've done) at observed 20% bandwidth loss
2) Replace my Gen2 enclosure with a Gen1
3) Buy add-in USB card
4) Upgrade to 7000 platform

So I moved all my slow USB peripherals (scanner, printer, mouse, keyboard) to Gen2 ports, leaving the four Gen1 ports for fast USB drives. Ironic.

================================================================

Now that I know the USB issue is not related to memory, I found I can simply run:
3600 2x16 1.45V CL14, and soc lowered to 0.95V! The other voltages auto to 0.90-0.95V per ZenTimings.

 

[xorbe]

HOLD UP

Why do memory sticks have SPD/XMP info if the motherboard doesn't apply it?

I manually entered their Trrds / Trrdl / Faw / Trfc{1,2,4} and now the "bad sticks"
seem to work fine.

My guess is that the S821M sticks are more picky about tRFC versus S820M, so
the 820 squeaked by, but the 821 didn't. Both have the same SPD numbers.

4 x 32GB drops to 2666 though, I didn't catch that before.
No can do 3600 w/4x32GB sticks it seems.

 

[ir_cow]

HOLD UP

Why do memory sticks have SPD/XMP info if the motherboard doesn't apply it?

It does apply it. JEDEC specifications that is. XMP is a extension of JEDEC and was developed by Intel. While motherboard vendors could enable this by default, the results would be none stop RMAs and customer support tickets.