That would be because the manufacturer didn't care to reprogram the SPD chips after switching from working with B-Die chips to working with C-Die, or it's just a simple oversight. That SPD chip doesn't actually read info about the memory, it is only there to be read.
With my modules, voltage at 1.35 was actually counterproductive, I had better luck staying as low as possible. I've also read an account of a user killing multiple C-Die DIMMs after half a year at 1.35V, so it's absolutely a good idea to keep voltage low.
I spent a few hours yesterday trying to reach a stable clock at a lower voltage, and I was unable to improve or even keep the same clocks I have now. With the same timings, it doesn't post at 1.25 V, and fails memory tests horribly at 1.30 V (with some uncorrectable errors on top of correctable ones). I'm not sure what to make of that. Do you think I should loosen the timings to run at a lower voltage?
Edit: Actually, my sticks are running at 1.392 V according to HW info. Apparently the setting on the motherboard is just a suggestion. In a similar vein, I also have to set the cas latency to 17 in order to set it to 18, since 18 sets it to 19. I'm not entirely sure what's going on with my motherboard firmware/uefi.
Ah...I'm planning on using the Black Ridge with a downfiring 120mm fan (something I expect a lot of people here are aiming for, or something similar given the relatively limited need for something smaller than an IO shield), so I imagine temps would stay relatively cool as long as the CPU can also stay cool, since air will constantly be moving over the RAM modules.
My sticks are running just above 40 C under my black ridge (in a sentry 2.0). I've seen them reach 50 C after an extended time under CPU/GPU load, which is probably a bit iffy for ram stability, but I have not ever seen any memory errors (no ECC corrections or hard crashes).
Edit: I just ran a test with the large ffts on prime95. My ram hit 55 C, which is higher than I've seen it (probably because I just swapped the CPU fan from exhaust to intake). Supposedly instability can start to creep in after about 50 C, but I guess it hugely depends on your sticks, your clocks and all sorts of other factors.
Ah...I'm planning on using the Black Ridge with a downfiring 120mm fan (something I expect a lot of people here are aiming for, or something similar given the relatively limited need for something smaller than an IO shield), so I imagine temps would stay relatively cool as long as the CPU can also stay cool, since air will constantly be moving over the RAM modules.
Oh yeah I have that setup, in a K39. I've just checked it out because it interested me and according to the on-DIMM sensor, with a silent setup of 900RPM (NF-A12x15), ram temps went into the high 60's at 1.40V VDIMM (69°C @ 22°C ambient, during Karhu which is gonna be one of the hottest tasks for memory). That's pretty hot. At max RPM it stuck to 55°C.
I figure that during summer, this could legitimately lead to instability. I'm gonna decide to run higher fan speeds then.
I spent a few hours yesterday trying to reach a stable clock at a lower voltage, and I was unable to improve or even keep the same clocks I have now. With the same timings, it doesn't post at 1.25 V, and fails memory tests horribly at 1.30 V (with some uncorrectable errors on top of correctable ones). I'm not sure what to make of that. Do you think I should loosen the timings to run at a lower voltage?
I'd say that depends on how much of a real performance loss you'll see, but if those were my only sticks I'd personally pick a (presumably) longer lifespan over tight timings. But I know this takes a ton of effort yet again, which obviously sucks
anything wrong with running the kingston KVR26N19S8L/8 with a ryzen 3600 and the black ridge with a nh a12x15 fan on intake? want to see if everything runs cool enough and still be able to play games and not be limited by ram stock timings
Oh yeah I have that setup, in a K39. I've just checked it out because it interested me and according to the on-DIMM sensor, with a silent setup of 900RPM (NF-A12x15), ram temps went into the high 60's at 1.40V VDIMM (69°C @ 22°C ambient, during Karhu which is gonna be one of the hottest tasks for memory). That's pretty hot. At max RPM it stuck to 55°C.
I wonder if it would be worth it to get some of those little heatsinks and stick them onto the RAM modules, or get something like the VLP heatspreaders that you see on some RAM sticks to help with the more bursty RAM loads that might be more common in something like games. I didn't realize RAM temps that low could cause instability, but I guess I was working under the assumption that GDDR# RAM temps would apply to normal system RAM.
I mean, it might even be worth just buying some cheap registered VLP DDR3 and just taking the heatsinks off of that, because IIRC RAM is all essentially the same length, just with different pins and notches, right?
Find many great new & used options and get the best deals for SAMSUNG / IBM VLP SERVER RAM 4GB 2RX4 PC3-10600R DDR3 44T1498 43X5052 at the best online prices at eBay! Free shipping for many products!
I mean, it might even be worth just buying some cheap registered VLP DDR3 and just taking the heatsinks off of that, because IIRC RAM is all essentially the same length, just with different pins and notches, right?
Find many great new & used options and get the best deals for SAMSUNG / IBM VLP SERVER RAM 4GB 2RX4 PC3-10600R DDR3 44T1498 43X5052 at the best online prices at eBay! Free shipping for many products!
Yeah I've been thinking about the same thing! Haven't seen those used server modules under 11€ per stick around here tho, that wasn't worth it imo. They should fit tho, DDR3 and DDR4 sticks have the same length and VLP is obviously a standard height.
From what I've read so far, common advice seems to be to keep your modules under 60°C, with B-Die allegedly even causing issues above 50°C when you're running high VDIMM. At least when it comes to tight overclocks where you're already intentionally pushing it close to the edge of instability
anything wrong with running the kingston KVR26N19S8L/8 with a ryzen 3600 and the black ridge with a nh a12x15 fan on intake? want to see if everything runs cool enough and still be able to play games and not be limited by ram stock timings
Hey everyone, I've been testing several VLP kits lately on the search for the best option for my system (and out of fascination, they're cool and niche). I figured I should tell this thread about my experience.
My kits:
Kingston KVR26N19S8L/8 2x8GB 2666C19
V-Color TE416G26D819V-VKC 2x16GB 2666C19 ECC
Innodisk M4C0-AGS1TCIK 2x16GB 2666C19 ECC
MB: Asus Rog Strix B450-I Gaming
CPU: R5 3600 (previously an Athlon 200GE)
The Athlon should approximately represent your average Zen1 and Zen+ memory controller and its limits, whereas the 3600 is obviously much more capable in every regard and can unlock the true limits of these ram sticks. The Strix B450-I motherboard is generally regarded one of the best options for memory overclocking on Ryzen and shouldn't be the limiting factor in most cases.
The modules use memory chips by Nanya Technology, specifically their first 8Gbit die revision, A-Die. Nanya only entered the DDR4 market in 2019 and reviews of this die are scarce and hard to find. Everything I've found points towards respectable overclocking headroom, around the 3400-3800MT/s mark. Thats sufficient for Ryzen.
On my 200GE, I ended up getting these up to
Frequency
Voltage
tCL-tRCDRD-tRCDWR-tRP-tRAS
note
3666MT/s
1.45V
16-17-18-21-36
3600MT/s
1.35V
16-17-18-21-36
3466MT/s
1.28V
16-17-18-21-36
3200MT/s
1.20V
16-17-18-21-36
There might have been more potential in those timings at lower frequency, but I stuck to 3600MT/s so I didnt really care about that.
On my Ryzen 3600, I curiously had to change the tRCDRD timing to achieve stability
Frequency
Voltage
tCL-tRCDRD-tRCDWR-tRP-tRAS
note
3600MT/s
1.35V
16-19-10-21-37
3800MT/s
1.35V
18-19-20-15-38
Higher frequencies did boot but threw errors quickly. Not sure if I'd be able to stabilize them.
Overall I'm happy with the results of this memory kit, I ended up recommending my 3600MT/s settings to someone else on Discord and they seemed to work there as well. Can recommend.
These modules uses memory chips by SK Hynix, namely their second latest 8Gbit die revision, commonly referred to as CJR. They're great ICs, can be found on many 3600MHz 16-19-19-39 or 3600MHz 18-22-22-42 retail kits. They may have overclocking headroom between 3600CL16 and >4000MT/s CL18, from what I've seen.
They're also the same ICs as those found on ADATA VLP modules recently, from what I've gathered.
A thing to note here is that -AURMEND- got this same kit with Hynix AFR ICs. Those are inferior dies on average, usually reaching between 3200 and 3600MT/s at most. The amazon seller that I got these from also sells a 2400MT/s module, which according to the low-quality product picture and amazon user reviews contains Samsung dies (B-Die, it seems).
On my 200GE, this kit capped out at around 3333MHz. Their dual rank nature is simply too taxing for weak memory controllers.
On my Ryzen 3600, they got a lot further
Frequency
Voltage
tCL-tRCDRD-tRCDWR-tRP-tRAS
note
4000MT/s
1.40V
18-19-10-21-21
3800MT/s
1.45V
16-19-10-21-21
3600MT/s
1.40V
16-17-10-19-21
I'm highly impressed with these modules, I think I got lucky. The ECC functionality doesnt appear to reduce performance and can be disabled in the BIOS.
These EUDIMMs contain Samsung C-Die chips (detected via their ...C-BCTD code on the ICs). Those are known to be some pretty bad ICs. They have low voltage tolerance and scale negatively with voltage past a certain point. They can be found on recent Corsair v4.32 3200C16 and I believe 3600C19 kits. Some reports suggest they can clock rather high with loose timings, but only at low-ish voltage.
On my 200GE, they achieved no greater than 3266MT/s at lame timings (16-17-10-20-40 IIRC)
On my Ryzen 3600, I did get to push them further
Frequency
Voltage
tCL-tRCDRD-tRCDWR-tRP-tRAS
note
3200MT/s
1.24V
16-17-10-18-33
3600MT/s
1.28V
18-21-21-22-42
3800MT/s
1.24V
20-21-21-22-42
-boot issues-
Anything between 3666 and 3800 might be stable, but demand higher voltage for tCL 16 so I stopped. 3800 had issues where it oftentimes didn't boot successfully, which could probably be stabilized with more effort, but its timings are super loose already so I won't bother.
These sticks are the worst performing ones out of the three. Based on this individual experience I wouldn't recommend them in case there are other options available.
Conclusion:
Kingston OC: 3600C16 or 3800C18
V-Color OC: 3600C16, 3800C16 or 4000C18
Innodisk OC: 3600C18
The VLP DDR4 DRAM options on the market right now appear to have respectable potential. Western users are no longer restricted to uncertain purchases from South East Asia, like it used to be until mid 2019. I wouldn't hesistate to pick V-Color modules or Crucial's CT16G4XFD8266 over the Innodisk modules.
I hope this will be helpful to some. I know my testing wasn't perfectly consistent or perfectly thorough (you'd typically want to run an overnight ram test for every setting), but this should be fine in terms of giving some form of indication.
I have also decided to compile this spreadsheet for information about VLP DDR4, with reported overclocks I could find on here, on Reddit and on Discord.
I think I'm gonna sell my Innodisk modules, I'll post a trade thread some time.
Funnily enough, I was running some calculations for work for the first time since I swapped my fan orientation, and I started seeing corrected error warnings at a rate of about 4 per hour, which means instability for sure. This was on the timings I had checked with memtest for about 12 hours a few months ago, and had given no such warnings before now (I was beginning to wonder if windows was even reporting ecc warnings correctly).
Unless my ram has already degraded, I suspect this is evidence of thermal instability, since they are running about 5 C warmer due to the fan set as intake (up to 55 C under load). In the end, I slackened my timings back to 3600 18-19-19-43 @ 1.35 V, which seems to be stable now. I still could not achieve anything even remotely stable at 3600 MT/s (and reasonable timings) at lower voltages though.
The more I read about Samsung c-die, the less sure I am about anything. There seem to be a lot of conflicting reports.
Yeah I've been thinking about the same thing! Haven't seen those used server modules under 11€ per stick around here tho, that wasn't worth it imo. They should fit tho, DDR3 and DDR4 sticks have the same length and VLP is obviously a standard height.
From what I've read so far, common advice seems to be to keep your modules under 60°C, with B-Die allegedly even causing issues above 50°C when you're running high VDIMM. At least when it comes to tight overclocks where you're already intentionally pushing it close to the edge of instability
I've looked into aftermarket heat sinks for vlp and the only place that has them is a company called memorypack. There isn't a lot of information how to order them and I don't know if you have to buy them by the carton (375 sets).
While looking for a more reliable source for heat spreaders I did fall across ram from Sunelectronics who use black pcbs for vlp ram. Looks like you would have to contact the supplier for pricing.
While looking for a more reliable source for heat spreaders I did fall across ram from Sunelectronics who use black pcbs for vlp ram. Looks like you would have to contact the supplier for pricing.
that's a picture of a ddr3 rdimm. tigo also produces vlp ddr4 with a black pcb. i talked to them last year and they weren't interested in producing it without someone purchasing a batch of 4000 dimms. i'm assuming sun electronics is probably the same kind of producer. you might want to contact them to get an MOQ.
My DDR3 ECC VLP RAM (purchased for the heatspreaders) came in today....man I did not realized just how small VLP RAM is. I'm still waiting for Amazon to ship the rest of my build out. For ~$5/stick the heatsink wasn't a bad price although I wish I'd known about the AliExpress option before I bought these RDIMMs.
I'm gonna test out the RAM OC without the heatspreaders first just to see what my temps look like, and then test them out with them on to see how much of a difference it makes. I am assuming that the DDR4 comes with easily exposed temperature sensors. On my current X58 build the temps aren't that easy to read/find/decipher, and the voltages and power readings are very suspect.
My DDR3 ECC VLP RAM (purchased for the heatspreaders) came in today....man I did not realized just how small VLP RAM is. I'm still waiting for Amazon to ship the rest of my build out. For ~$5/stick the heatsink wasn't a bad price although I wish I'd known about the AliExpress option before I bought these RDIMMs.
I'm gonna test out the RAM OC without the heatspreaders first just to see what my temps look like, and then test them out with them on to see how much of a difference it makes. I am assuming that the DDR4 comes with easily exposed temperature sensors. On my current X58 build the temps aren't that easy to read/find/decipher, and the voltages and power readings are very suspect.
