Separate names with a comma.
Discussion in 'Small Form Factor Systems' started by SaperPL, Apr 16, 2017.
Any result of the dynatron on ryzen yet ? I'm eager to know
sadly no, my memory is still at RMA...
Could you please post me the AM4 hole spacing and NH-L9i hole spacing? I decide to make a custom plate, my Sentry has lain idle almost for a month and a half....
AM4 is 90 x 54 mm, NH-L9i is 51 x 22.5 mm and it's plate is 41 x 40 mm.
Very detailed! Thank you!
can you please let me know what screw thread the backplate of your Biostar AM4 boatd needs? M3 or M4? Furthermore I like to know the height from MB to installed CPU topside.
As I know there are some standoffs on the backplate that come through the mounting holes. What is the diameter and how far stick they out?
Sorry for late reply, didn't had time at the moment and forgot about this afterwards.
The screws in the backplate that are holding the plastic brackets are M3.5x10.
The pcb-to-IHS-top height is 7.5~7.75 mm like I've written before - I think that the socket might not be moulded evenly as I've tried to re-seat the CPU during those measurements with same results.
Meanwhile I got my memory sticks back from RMA, looks like they work properly on JEDEC settings at least, but I won't be trying XMP until biostar releases the agesa update for this board.
T318 works, can't say anything concrete about it right now, I'll have to spend some time with it since Ryzen is pretty tricky in terms of clocks and temps.
Rough temps currently: ~39 degrees on idle in quiet mode, spiking to ~45 degrees when some load comes up and ~51 degrees under heaven load, spiking to 56 degrees
SaperPL, would Noctua's NF-A12x15 fit in Sentry with T318?
Vertically yes, total height will be 42 mm so that's okay. Horizontally though you'll have to fit this between the RAM and IO - on gigabyte boards this should be doable as the ram slots are on the edge of the board. On biostar you'd have something like 115 mm of space between io and ram while you need 120 mm. On asrock it might fit above and between the IO as it is quite unique
Thank You, good thing I just bought the Gigabyte board because... well, it was aviable. I guess constantly refreshing store pages paid off.
Hey SaperPL, did you have a chance to test the T318 some more?
I remember in the Sentry topic you were toying with the idea of making AM4 brackets for the T318, is this something you are at all interested in doing anymore?
Thanks for the updates so far, great stuff!
I'm currently testing it and getting used to it, sort-of. Ryzen is pretty weird in terms of temps and also I'd like to test it with memory running at their full potential since running on JEDEC 2133 settings might mean the CPU itself not running at full *thermal* capacity.
I've asked G-Skill for some feedback on how to setup the memory on Ryzen and they've responded requesting additional info, so at least there's some chance to get it going.
As for the T318 - for games without OC it looks to be running properly.
As for the bracket - I'm not really keen on making a custom bracket if you can mount it with simple elements from every hardware store. Additionally I've realised that there may be different mounting positions with each motherboard depending on VRM height/location and additional components obstructing the mount. Note that some of the screws in my previous post here are more angled then the others because of the collision with VRM.
Just got the feedback from G.Skill and it looks like I've been missing one, not that obvious thing. You can actually customise XMP profile on Ryzen since enabling XMP profile doesn't disable custom memory settings (anymore? since agesa 18.104.22.168 maybe).
While inputting the XMP settings manually to run lower clock speed doesn't work/doesn't boot, overriding the clock speed on XMP profile does boot and seems to be working fine on 2933 which is reasonable boost over 2133.
I'll try to run some tests after confirming this is indeed stable.
SaperPL, I was able to manually overclock my G.Skill TridentZ 3200CL16 2x16Gb to 2933 CL15, just fyi. Although, what this is doing in this thread?!
People were asking when will I be able to test out the T318, I think I should be ready now.
As for the TridenZ - those are certified for Ryzen, the RipJaws aint...
I don't think TridentZ are certified for Ryzen. I haven't seen any mentions of Ryzen on my kit.
It's FlareX that are certified
I got a reply from G.Skill earlier:
I believe FlareX are the ones that are "designed for" Ryzen, but TridentZ were the ones that were working properly earlier than the FlareX were made. Anyway RipJaws aren't best choice for Ryzen.
I would also report that I have
[Trident Z] F4-3000C14D-16GTZ(R) and I'm getting 2933 just fine with Ryzen. Yay Samsung B-Die!
So what CPU cooler I should buy for my Ryzen rig? SaperPL?
I think that L9i with AM4 is what you should look at if you want out-of-the-box equipment. It was covered here:
T318 seems to have few the advantages of vapor chamber, but all in all I'm not sure if it's that much better than L9i when mounted in Sentry.
I'll have to run some extensive tests on both coolers to test it out and compare, and I simply don't have enough time for this now.
I think the difference between T318 and NH-L9i inside Sentry is similar to comparing AIO water cooler to similarily sized tower cooler - the temps feel more stable on T318 and tend to raise more slowly than on NH-L9i, but all in all I think that during the workload it's the fan and the environment that limits how much heat it can dissipate.
I could not resist in trying this:
The one extra hole was a test drill and also I had not cut threads manually for a long time...
I am using a Ryzen 1600X and this helped to avoid throttling in prime. It is stable with stock 3.6 GHz & -50mV offset. Effective clock is between 3600 and 3650.
I also exchanged the LP53 orginal fan with the Noctua NF-A9x14 PWM from NH-L9a.
I am tempted to get a Dynatron T318 and also cut holes in it to fit the AM4 backplate.
I also used the NF-a9x14 with my LP53. Out of curiosity, how did you mount the fan to the cooler? Were you able to mount the LP53 with the fins parallel to the ram? Or could it only be mounted perpendicular? Currently my LP53 is on a delidded 4790k with the fins running parallel to the ram for better air flow. I'm really tempted to switch to Ryzen if/when there's a refresh or ryzen 2 with slightly faster clocks and a more mature platform. Now that I know it's feasible to use the venerable LP53 with am4, that's one less barrier to entry for me.
Unfortunately, judging by the adapter it is perpendicular.
RosaJ is right, parallel was not possible because of the adapter.
I used cable ties to connect the fan to the cooler. This was a bit of fiddeling.
Looks great! Do you happen to have some temps for us?
Nice try, but better will be symmetrical position (without losing cover surface).
What test would you like to see?
Sweet, if you could grab some temps from prime95 v26.6 on both LP53 and NH-L9a to compare and any other test you may want to do (what program are you using for temps?). I have neither, but would love to see how well the LP53 does on Ryzen. I will be doing a comparison with my C7 and T318 with a few different fans once my NF-A12x15 PMW comes in, hopefully before this weekend.
why prime v26.6? something special there?
I have v29.1.
may take a while to collect this data.. I am not promising but I will try to do it over the next week.
I would record temps, etc with HWinfo.
Thanks, I look forward to your results. I've been using 26.6 as it was for my old sandy bridge CPU, but I now realize I have a new CPU so I'll be using 29.1 as well thanks for the double check!
After some further testing, I have to correct my earlier statement - that the LP53 can prevent throtteling the 1600X
Just some basic information on the Ryzen 5 - Normal clock is 3600Mhz, XFR boost to 3700Mhz, precision boost up to 4000Mhz, depending on CPU temp (has to be low). I only reach the 4000Mhz in idle or when I am just starting an application.
In all described below T(Die) was around 74.5°C when the CPU was under load. T(CPU) is the temperature of the motherboard CPU sensor. T ambient was 23°C.
Neither the NH-L9a nor the LP53 (both with NF-A9x14) could cool the Ryzen5 1600X to prevent throttling under extreme load. Running prime95 (small FFT) a steady state of ~3500MHz was achieved in a ventilated case (1x 140mm in, 1x140mm out, both at 500-700rpm).
The CPU fan ran at max speed (100% 2600rpm), which is noisy. CPU temp 64.5°C
Interestingly, at 87% fan speed the same result (~3500Mhz) was achieved. CPU temp 65.5°C
Running the fan at 73% resulted in stronger throttling (~3400MhZ). CPU temp 67.5°C
54% 3200MHz and 69°C
Running a GPU benchmark (RX580) causing 100% GPU utilization in parallel increases the throttling for another 50-100Mhz
In comparison the NH-L9a did behave very similar, if not exactly the same, reaching the same Mhz at the same temperatures
I have put the above in a XLSX /ZIP with the HWinfo data I recorded… feel free to manipulate the data further..
In normal gaming scenarios, if the CPU is not at 100% load but well below, I have not seen a throttling
I also got a Dynatron T-319, but it does not fit my motherboard (Asrock B350 gaming), so no chance right now I can easily modify this (written this I could add a small copper plate between IHS and T-319 to lift the cooler some mm over the obstacles... mhh...)
I think you choose wrong position U heatpipes:
I cant turn the cooler by 90°C, as the mounting does not fit that way.
Unfortunatly, then Noctua NH-L9a has the same issue.
I will try to put the U down
Done - no noticeable difference in temperature or throtteling.
Maybe with bracket similar to Silverstone AM4 type C...
Has any of this info changed or improved? Did you find a way to mount the fan to the heatsink?
Any progress with the AM4 LP53 mounting?
Few posts below the one you have quoted me on I have attached photos of the fan ziptied to the T318.
Both NH-L9i and T318 work similarly in Sentry because there's not much airflow that could improve the performance of T318 which was designed for rack use.
I didn't have time to make selection of tests with swapping coolers in and out, but I've used NH-L9i for two weeks or so and T318 since the change.
R7 1700 on both had temps like 35~40 on idle/browsing, spiking up to ~50 in gaming and heating up to ~60 in some normal multithreaded workloads, not synthetics.
R7 1700 is also pretty weird CPU to benchmark because the turbo clock on single loaded thread heats whole CPU up faster than if all cores are fully loaded...
Changing the memory speed between 2133 MHz and 2933 MHz doesn't make a big difference in thermals as well. I couldn't run 3200 MHz on my sticks though.
Finally, I've already had issues with T318 base being blocked by VRMs on X370GTN so it's off center and I think only gigabyte's boards may not be affected by this problem.