Threadripper 3960X RAM overclocking

I don't like this video at all.

Too many I dont knows, or I am not sure, or I guess, or I dont have experience with comments.

To me that shows no authoritative knowledge or confidence in your understanding of memory tech.

Thus, I submit I am not subscribed to the channel.

I will tighten my ram later and see what kind of changes it makes.

I have 32 GB of B Die unmatched so It might hit 14 stable or it might have to settle at 15 cl.

I am going to sell the ram soon and get a matched 4x4 or whatever 64gb kit.

Gaming is the only thing I can imagine. Not many performance synthetics care about ram timings.

Its really hard to quantify latency changes in gaming when your CPU is at the pinnacle of performance and your GPU is a 2080ti. Where you would see the huge difference in the low end or middle of the road, like APUs or Athlons or lower end GPUs or I3/I5 cpus etc....

I am not equipped to do such benches at this time.
 
Last edited:
  • Like
Reactions: erek
like this
So I'm at the RAM tweaking point in my 3960X threadripper build and so I'm looking for any lessons learned so far and found this thread. My main components include a Zenith II Extreme, a Deepcool Castle 360EX AIO (the easy route), and 128 GB Corsair Dominator Platinum @ 3600 C18 (the only 3600 128 GB kit in the QVL cuz I wasn't rolling dice on that much money), and I recycled my GTX Titan (Maxwell) and two 1080Ti's into the build (I'll see what this year holds in GPUs and maybe upgrade). I couldn't get all 128 GB to boot at the 3600, even with the D.O.C.P. 18-19-19-19-39 1.35V settings... I did finally manage to make it work at a stable 3400 with 14-15-15-15-30 settings at 1.34V and 1T. The downside is that my CPU idle temps are in the mid 50's with that, at least it feels like a downside as without the RAM overclock the temps were in the mid 40's. I suspect the increased SOC voltages or memory controllers are contributing quite a bit to those temps. I also realize the cooler I'm using is nowhere near what most other folks are using here with custom loops, so that plays a factor and at some point I'm going to need to upgrade to a better cooler. That all being said, I just have to wonder whether the timings are all that important in the end as every time I run Cinebench R20 with different frequencies and timings the scores barely move staying between 13100 and 13500... even with the default 2133 system boots. While Cinebench is perhaps more a CPU test, I would have thought that the increased frequencies would have made the infinity fabric run faster and hence the CPU. I don't game a whole lot and not trying to get the highest frame rates anyway, but I just have to wonder if the added temps are worth the higher frequencies and tighter timings...
 
I wish I remember what youtube channel it was but they were testing ram timings on ryzen in games and other benchmarks.. there was a big difference in fps between timings in games.. for instance 3600 ram at stock timings got x fps.. ram at lower speeds and real tight timings got like 10 percent more fps or something like that.. was pretty crazy how much a difference it did make. It may of been a jayztwocents video.. I watched it a month or two ago. I need to try to find it again as I want to watch it again anyway.
 
I couldn't get all 128 GB to boot at the 3600, even with the D.O.C.P. 18-19-19-19-39 1.35V settings... I did finally manage to make it work at a stable 3400 with 14-15-15-15-30 settings at 1.34V and 1T.

8x16gb is officially only supported at 2667mhz. I would consider it lucky to get 8 sticks at 3400mhz. Temp wise you kind of need more cooling imo.
 
Yep, that's pretty much where my thoughts are too given the temps... better cooling would help. Outside that, the only other thing I can do is lower my memory frequencies and timings to take some of the load off the memory controller or lower my SOC voltages (@1.05V with the 3400 settings), and I guess I'm still on the fence as to how hard that would hit my render times. Looks like I'll have to try a few renders at each setting and see how things pan out. Maybe next week... if I have to look at that BIOS one more time this week trying to pin down a decent memory setting I might snap... :wacky:
 
Alrighty, this niggled at me all night, so ran some tests today just using cinebench R20 and adjusting my RAM frequencies from 3400 down to the default 2933 to see if that would help my idle temps. It didn't. And suprisingly, it didn't even affect my cinebench score which remained virtually constant through all the tests so I guess that test does not involve infinity fabric too much. That pretty much leaves my cooling solution as the likely weakest link. It is worth noting too that closing Corsair's iCUE and rainmeter along with HWinfo background tasks gave me back almost 10 degrees which I'll take for now as I ponder a better cooler.
 
Alrighty, this niggled at me all night, so ran some tests today just using cinebench R20 and adjusting my RAM frequencies from 3400 down to the default 2933 to see if that would help my idle temps. It didn't. And suprisingly, it didn't even affect my cinebench score which remained virtually constant through all the tests so I guess that test does not involve infinity fabric too much. That pretty much leaves my cooling solution as the likely weakest link. It is worth noting too that closing Corsair's iCUE and rainmeter along with HWinfo background tasks gave me back almost 10 degrees which I'll take for now as I ponder a better cooler.
I pick up like 800 points form stock xmp to one notch under 3900 at 14-14-14-14 and I forget the rest. I get like 15100 r20 all cores at 4.45

Try a bit higher soc to like 1.2 and see if that helps
 
Last edited:
Back
Top