8 PACK MEMORY RANGE GROWING: SAY HELLO TO 8 PACK RIPPED EDITION & 32GB KITS!!!
- Thread starter Gibbo
- Start date
Soldato
Hah, post 1.7.0 the calculator suggests c16 for my RAM when previously it suggested c14! Though if I just chuck in the calculator timings 1.7.0 give me it was unstable. I thought my RAM was faulty as I couldn't get it stable even at 2933 with the calculator settings. I ended up learning what each timing does and which should be adjusted together, and tuned them myself. The timings which contribute to single errors after hours of testing, at least for me, are
tFAW
tWR
tRTP
Though from looking at the results some of you are getting on the MEMbench, some of you appear to have these timings set at board auto, yet have quicker completion times than those who have set them lower.
I would definitely suggest to those here who have dialed in low timings across the board, to test at least overnight, as I've managed to get an error at 15000% in Karhu before now, which some would be happy with, but an error is an error, and that could mean corruption down the line!
You will also find that setting some secondary and tertiary timings too low can negatively affect performance. Lower isn't always better
I'm a noob relatively speaking at this... but my thoughts are any instability you just bump the voltage 0.01v until it disappears. If I got an error at 15000% I'd bump the voltage.
At the moment I'm game stable. If I get any instability, I'll bump the voltage. Fingers crossed that sorts things out. It has in the short term.
Also your find your instability far sooner than waiting for 15000% to pass by using a real world load.
There's a good explanation of this on the HCI webpage. It's call transient errors. You can test and test and test until the cows come home but you're not necessarily get an error until you do.
This is also why I don't waste my time with endless hours of stress tests when I could stress test whilst playing games. I get initial stability with stress tests obviously then it's time to test game stability. At the end of the day that is the load you are going to be putting on the system in real time so your better of being stress test and game stable than just relying on a thousand hour stress test stable.
Heat, power consumption and compute are different when you game plus the CPU is using the ram randomly (reading and writing). Whilst when you stress test there is a consistent stream of load on the ram plus the system isn't being loaded up with power and heat by the GPU so I think real world usage puts different stress than just stress testing programs.
Sure you might be able to change the timings to improve error rates but I not gonna spend my time to learn what all the 60 odd timings do that's what I use the calculator for. I got to trust that it is right. My only tool to improve stability is that dram voltage.
Last edited:
Man of Honour
Associate
With my current timings, I didn't touch my computer for a few days due to crappy work pattern. Come back to see it still testing without error, so I'm pretty certain my settings are stable
I do agree with you on this. I just didn't personally like knowing that there was an error, even at 15000%. I could have tested to 6400% (Which is 99.41% confidence according to Karhu FAQ) and been happy! This takes just a few hours if you Enable CPU cache rather than leave at default in Karhu. Working nights allowed me the time to go to the trouble of getting rid of these single errors as I wasn't using the computer anyway. Though of course, test long enough and you will likely get an error due to transients, cosmic rays etc!
Absolutely, though while I was testing, the RAM test alone didn't create enough heat to spin my chassis fans higher, pushing the DIMM temps up higher than when I use my system for gaming. They got as high as 55c while testing, though I only ever see them get into the low 40's while gaming or if I also load up the CPU and GPU, or force my chassis fans to full speed.
The calculator changes what it suggests almost every time it updates recently. So the timings you've input now may not match what it suggests next time it updates.
There are 'just' 26 timings displayed in the calculator, though it is only worth tuning 17 of them (the primaries and secondaries), and most of these work in pairs. The tertiaries will likely be set at or close to those suggested by the calculator, automatically by your motherboard.
My RAM wasn't stable with settings copy pasta from the calculator, which led me to learning about the timings and which work together, to whittle down which timings, on my system, were causing errors reasonably early into testing. If your system is stable with the timings given on the version of the calculator you are using at this moment, then you are lucky and have saved yourself a week or two of testing different settings!
Glad you're happy with your RAM overclock though
Last edited:
I think the editor is broken, at least in firefox, it's not adding the closing tag and typing it in manually isn't working either.I'm really thick can't work it out, but thanks for the help
You really shouldn't be using Firefox.
Hey mate,
Is this what you are after?
Taiphoon XMP HTML profile if possible. e.g. follow from 1:56 to 3:46 in this vid. The exported HTML file is what is useful: https://youtu.be/Bwr7sl5eZUM?t=116
Lucky you checked cause I think someone set your XMP wrong. CAS# Latency should be at 14....
Man of Honour
That is strange, its the only screen that shows C15, all others show 14, same as my BIOS and Ryzen Master
I'll grab the screenie for you soon, just off to have some lunch
Associate
@Space Monkey off topic slightly, but would you recommend your Sabrent Rocket 1TB? Over other slightly more expensive NVME like the Corsair MP510?
Man of Honour
Yes mate, I'm currently running the Sabrent Rocket 512GB & 1TB NVMe drives, running sweet so far, and IIRC they are based on the same controllers as the Corsair anyway?
Man of Honour
Man of Honour
Thanks for that and this is good confirmation that the XMP is fine. I'd really like to know why CPU-Z is showing 15 though for curiosity sake. The only thing I can think of is having 4 dimms of this is somehow confusing CPU-Z.
If anyone is running 2 dimms of the 3600C14 stuff I'd like to see what your CPU-Z says.
If anyone is running 2 dimms of the 3600C14 stuff I'd like to see what your CPU-Z says.
Associate
Yeah get your pics out for the lads.
Associate
Correct, they are. Though in the thread I saw you had a Rocket, it was mentioned that the MP510 has the full amount of DRAM whereas the Rocket has half. You running two NVMe in one system? Don't think I can do that on X470 without a PCIe card, though I think it would drop my GPU down to 8x
@GregI I've found any program that is reading SPD values, CPU-Z, Thaiphoon etc can read them wrong or not read entire values at all. Refreshing it sometimes works. This also appears to be the case for programs that can read timings live, like Ryzen Timing Checker and Ryzen DRAM Calculator.
Associate
Look at the CAS Latencies Supported field under the XMP section. It doesn't list C14...which could be why CPU-Z is reading it wrong, or at least displaying it incorrectly as the readout is showing that the XMP profile works with C15 to C19
Man of Honour
What conditions does that make a difference though? (Question as I don't know lol) For general PC use and gaming it runs really well and super snappy
Associate
Not a clue mate! I doubt it makes a difference, I believe the DRAM is used mainly for mapping etc on NVMe storage. Do you have them both in one system? I'm thinking of getting the 1TB and putting my most played games on that, along with the OS.
I'm guessing that's what people are doing now, especially with the upcoming consoles having NVMe?
I'm guessing that's what people are doing now, especially with the upcoming consoles having NVMe?
Man of Honour
Yeah I've got the 512GB as an OS drive and the 1TB as my Steam Games drive
