Overclock.net banner

5641 - 5660 of 6341 Posts

·
Registered
Joined
·
6 Posts
I need input on AGESA 1.2.0.0 vs 1.1.9.0

With AGESA 1.2.0.0, no WHEA errors with this settings:
VSOC 1.15V
VDDP: auto
VDDG: auto
VDIMM: 1.47V
[email protected], aida64, R20, Games (only whea errors in prime95 blender)

If i manualy set vddp and vddg to the exact same settings i get whea just by launching windows WUT?

2nd part:
With AGESA 1.1.9.0 i had higher READ, COPY and L3 Cache by alot as you can see in screenshots below with almost the same settings and i was running it for like 10days without any game crashing and 0 whea errors.

PS.
1. FCLK1933 is not stable with this settings IDK?
2. FCLK2033 is stable with this settings
3. FCLK2066 whea errors in window does not crash (performance degradation)
4. FCLK2100 sound errors, windows crashes after 2-3min.
5. AGESA 1.1.9.0 maximum windows FCLK was 2066
5. Should i return to 1.1.9.0?
6. Should i bother with further sub timings ?


AGESA 1.2.0.0
2475079


AGESA 1.1.9.0 (asus pulled this bios after 24h coz of broken usb/M2)
2475080
 

Attachments

·
Registered
Joined
·
6 Posts
Also there is, a hidden limit for UCLK even if you set DIV1 MCLK=UCLK it does not work...
I did MCLK: 2300 mhz FCLK: 2000 mhz but uclk was at 1150 mhz...
[email protected] 4600mhz/cl19 resulted in 62ns latency not bad ?
If i could set MCLK=UCLK 1:1 it would prolly drop to 57ns w/o FCLK 1:1:1
Wonder what, a 2000fclk and mclk=uclk 5200mhz.
 

·
Registered
Joined
·
58 Posts
Does anyone with fast RAM have any benchmarks in a game that likes fast memory with zen3? I am interested to see the real world performance benefit between the usual 3600c16/3200c14 and the very high 3800c14+
 

·
Registered
Joined
·
63 Posts
Does anyone with fast RAM have any benchmarks in a game that likes fast memory with zen3? I am interested to see the real world performance benefit between the usual 3600c16/3200c14 and the very high 3800c14+
short answer, only benefit with high end gpu, search reviews from hardrware unboxed,
 

·
Registered
Joined
·
6 Posts
Does anyone with fast RAM have any benchmarks in a game that likes fast memory with zen3? I am interested to see the real world performance benefit between the usual 3600c16/3200c14 and the very high 3800c14+
Or high fps with medium quality.

3600x (4.4ghz 3733cl15) overwatch 240-360fps, horizon zero dawn 78fps
5600x (4.7ghz 4000cl16) overwatch 399-400fps , horizon zero dawn 84 fps
 

·
Registered
Joined
·
16 Posts
Hello guys! I've recently swapped 3600 CL18 kit with 3800 CL14 kit and I don't see any latency improvements.
My setup:
  • 5900x
  • MSI B550 Gaming Edge
  • current ram: 2x16 dual rank 3800Mhz CL14 G.Skill F4-3800C14D-32GTZN
  • previous ram: 2x16 single rank 3600Mhz CL18 Kingston.
G.Skill has an even worse latency result.
I tried fast timings from Ryzen Dram Calculator without touching the voltages and anything else, but it gave me ~5 ns improvement: from 67 down to 62, still bad.
DRAM voltage is 1.5V. What should I do next? Try to tighten timings further?
Also I'll definitely run memtest on G.Skill, maybe something wrong with the kit itself.
2475256
16097586756572.png
 

·
Registered
Joined
·
167 Posts
@Manuru
RAM is fine. It's all in the subtimings. Post your Zentimings screen so that we can have a look.

Btw, if u don't chase the last 1 percentage, it's usually enough to get the cheapest B-die RAM u can find and with some more voltage and tuning it will perform very very close to the more expensive bins, especially on Ryzen because u can't clock as high as on Intel. I have G-Skill 3600CL16 DR 16GB RAM and I use it 24h with 3800CL14 tRFC 252 with 53ns AIDA latency. It also worked @ 4066CL16 with my 5800X. Both on 1.5v.

As u experience yourself, for optimal performance u have to finetune anyway. 3800CL14 with loose subtimings will perform worse than hand optimized 3600CL16. High frequency and tight main timings are primarily for selling the kits at a higher price. Sub-timings are kept loose for stability and best compatibility since they don't want uninformed people to return the RAM when the XMP setting fails.
 

·
Registered
Joined
·
25 Posts
Hello,

I recently upgraded my main rig from a 3900X to a 5900X. Been playing around with getting my RAM OC stable but I feel like I'm making no progress so I could use some advice here.

Here are my specs
CPU: Ryzen 9 5900X
CPU cooler: NH-D15
RAM: G Skill Trident Z 32GB (4x8GB) 3600MHz CL15 (two of these kits F4-3600C15D-16GTZ)
Motherboard: Gigabyte X570 Aorus Master Rev 1.0 (using F32 bios Agesa 1.1.0.0D)
SSD: Samsung 970 Evo Plus 1TB
GPU: RTX 2080
PSU: EVGA 750G2

At first I tried just the XMP and that works successfully but performance wasn't where I would have liked it to be and latency was quite high around 60ns due to really loose auto timings.
Then I tried manual tight timings at 3600MHz and that works as well but I feel like I can do better.
Tried going balls to the wall with 3800MHz 14-14-14-28 and it wasn't successful. After playing around with it some more no matter what I did, 3800MHz at CL14 just simply wont boot for me.
3733MHz with those timings does boot and I can get into windows but errors out a couple mins into Karhu.
I said screw it, turn everything back to auto except mem frequency and left it at auto and just wanted to see what the mobo would try. It set the primary timings at 16-15-15-35 which isn't terrible but the subtimings weren't great and latency was back at around 58-60Ns

I then just tried lowering my TRC and TRFC and ended up with this
2475297

I feel like I can take this somewhere but its not stable. PC restarts 5-15mins into Karhu ram test.
I tried upping Dram voltage to 1.5V wasn't stable
Tried VDDG voltages at 1050mV not stable

You guys got any suggestions?
 

·
Registered
Joined
·
18 Posts
Would anyone please give me some advice?

I was offered a very good deal for 128GB kit, 4x32GB 3600 C18.

At first I installed the kit and was not able to get over 3200 Mhz. I read online that it can be of a benefit to check the sticks individually and fit into the A2/B2 the worst performing sticks and A1/B1 the better sticks.

I have done this and now the system will boot XMP and seems stable.

2475302


I have made some very slight adjustments to sub timings which has reduced latency by approx 5ms and improved read by around 4GB/S but I was hoping with the right settings I could achieve better performance.

I have no idea what to try with chipset/IMC voltages or DrvStr for this memory. Would anyone have any ideas? The only thing I have done to start is set SOC to 1.15v and CDDG/CLDO to 0.9v. I have not made any other adjustments yet.

I am sure the limit is the IMC as the worst performing stick was able to run 3866mhz on its own with the same subtimings.

If I am able to find settings that make things easier for the IMC I am hoping I can either tighten the timings a little or increase the clock speed.

It is very appealing to have 128GB but the latency is nearly 10ms more than my 32GB E die memory so I am hoping I can be closer than this.

Thank you in advance for any advice.

PS These memories have lights on them. I have never seen this before! Very fancy.
 

·
Registered
Joined
·
16 Posts
@Manuru
RAM is fine. It's all in the subtimings. Post your Zentimings screen so that we can have a look.

Btw, if u don't chase the last 1 percentage, it's usually enough to get the cheapest B-die RAM u can find and with some more voltage and tuning it will perform very very close to the more expensive bins, especially on Ryzen because u can't clock as high as on Intel. I have G-Skill 3600CL16 DR 16GB RAM and I use it 24h with 3800CL14 tRFC 252 with 53ns AIDA latency. It also worked @ 4066CL16 with my 5800X. Both on 1.5v.

As u experience yourself, for optimal performance u have to finetune anyway. 3800CL14 with loose subtimings will perform worse than hand optimized 3600CL16. High frequency and tight main timings are primarily for selling the kits at a higher price. Sub-timings are kept loose for stability and best compatibility since they don't want uninformed people to return the RAM when the XMP setting fails.
Thank you for the explanation.
You were right, RAM is fine. I double-checked everything and found out that I had to manually force 1:1:1 ratio in BIOS. After I did that, AIDA64 gave me expected latency below 60ns.

I decided to get a pricey kit because I expected it to be easier to overclock, but maybe it was too naive. I tried safe and fast presets from Ryzen Dram Calculator, but both of them led to errors in TM5. Stock XMP works fine. I think I'll wait AGESA 1.2.0.0 for my motherboard first.
2475313
2475314
 

·
Registered
Joined
·
167 Posts
Dont have much time. For better latency, u can reduce trc, tfaw and trfc.
for 1900 fclk u should be able to use lower voltages, esp vddp. Vddp 0.9, vddg ccd 0.95, vddg iod 1.0
Here is my 3609cl16 b die kit @ 2033/4066, gaming stable but ofc with whea errors. Its an older try and now i would probably use somewhat lower secondary voltages, esp vddp
2475320
 

·
Registered
Joined
·
218 Posts
Hey guys, been testing out 3800 all today and was wondering if you guys had some advice on tightening the timings at all? Dram 1.45v, Soc 1.125, VDDG's 1.050, and CLDO .950 atm.. Passed 400% Single on the Dram Calculator memtest.
2475332
 

·
Registered
Joined
·
61 Posts
Hi everyone, I was hoping someone could help me understand what the DOCP profile of this equivalent ram would be if it was the AMD NEO version, not the Intel XMP version that I ended up buying. I have a 5800x that I'm pairing it with.

I attached the Zentimings photo of the XMP profile of the F4-3600C14-16GTZR TridentZ RGB. I'll upload a better quality screenshot in the morning... I have it running decently with modifications for AMD at 3600 14 but it really seems finicky and often won't post with changes that I think are okay, or at least were easier on the Crucial sticks they are replacing.

Does anyone have the DOCP profile of the 3600C14 TridentZ NEO, or equivalent, they could post please? Or any other brand's 3600C14 B-die?
2475353


EDIT: Replaced my smart phone image of zentimings with the actual screen shot of zen timings. This is the original Intel XMP before I changed the settings to better match AMD.

Sent from my SM-G973U using Tapatalk
 

·
Registered
Joined
·
103 Posts
Hi everyone, I was hoping someone could help me understand what the DOCP profile of this equivalent ram would be if it was the AMD NEO version, not the Intel XMP version that I ended up buying. I have a 5800x that I'm pairing it with.

I attached the Zentimings photo of the XMP profile of the F4-3600C14-16GTZR TridentZ RGB. I'll upload a better quality screenshot in the morning... I have it running decently with modifications for AMD at 3600 14 but it really seems finicky and often won't post with changes that I think are okay, or at least were easier on the Crucial sticks they are replacing.

Does anyone have the DOCP profile of the 3600C14 TridentZ NEO, or equivalent, they could post please? Or any other brand's 3600C14 B-die?


Sent from my SM-G973U using Tapatalk
You shouldn't really be relying on DOCP much, as it only sets the Primary timings. What voltage are you running for RAM? You are also running with GDM off 1T, which is harder to do than GDM on (but that's a good thing for performance).
 

·
Registered
Joined
·
25 Posts
Alright follow up to my last post

All I did was slightly reduce Vsoc and lowered frequency and IF clock to run 3733MHz and ended up with this.
2475352


No errors.

So now I think that my 5900X might just be garbage bin quality. It doesn't score very high on single threaded tests, and I wasn't able to stabilize it at 3800MHz mem. Or the latest bios F32 for the X570 Aorus master is holding it back. Hoping its the latter. But I'm going to call it here for now and start tuning the CPU itself.
 

·
Registered
Joined
·
61 Posts
You shouldn't really be relying on DOCP much, as it only sets the Primary timings. What voltage are you running for RAM? You are also running with GDM off 1T, which is harder to do than GDM on (but that's a good thing for performance).
I tried the Intel XMP settings (posted above) out of curiosity when I started them up for the first time but the results weren't great compared to my 3600-18 Crucial LPX 16GB that I had running at 3733-16 and 3800-18 with better results.
2475354


After that I came up with these timings and they run pretty decent and appear to be stable. I've also tried 3600-14-14-15-14-28-42 with the all the other timings the same and it also seems to work well and increases the bandwidth result a little. I'm running at 1.45V. I was hoping to compare with NEO XMP to see how close I am to what AMD puts out but I realize I can do without them.
2475363



I have also tried this but didn't run OCCT on it yet but the results are encouraging. Maybe that's what I'll do now actually.
2475364


Any thoughts on ways to improve? BTW, I'm running 3402 AGESA 1.2.0.0 on x570 Gaming Tuf Plus Wifi Board.
 

·
Registered
Joined
·
5 Posts
What do you guys think of karhu test nowadays? My rig:
B450-F gaming latest 4202 bios
5900x
4x8gb ripjaws V 3200cl14

I used this combo for a whole month without issues, then 2 days ago I decided to test my dram with karhu and I got a reboot with whea error (machine check exception, with cache hierarchy error with variable ID numbers for APIC cores). My previous ryzen 3600 passed 12 hours without issues at the same config.

The thing is, under normal daily usage my pc is completely fine, no crashes on hours of cb23, prime95, small, smallest or large and memtest86 as well. Tried relaxed timings and raised VSoC 1.1 and 1.45 for dram without success, best I got was a karhu run for 2 hours and then the pc rebooted. With docp I can open warzone, cyberpunk, dbd, lol, csgo, eso, borderlands 3, chrome and discord at the same time without errors. Do you guys think I should consider my cpu+Mobo+ram stable???
 
5641 - 5660 of 6341 Posts
Top