Overclock.net banner

141 - 160 of 185 Posts

·
Registered
Joined
·
590 Posts
I did use a negative offset lost single core performance as a result and in the end i reverted back to auto, I would of liked to disable llc but on this board its strangly worded such as normal, standard, low, meduim llc and so on. As for the motherboard its the Gigabyte x570 gaming x X570 GAMING X (rev. 1.0) | Motherboard - GIGABYTE Australia I wish i had paid an extra $100 for the better version thats the one most people seem to have and there are more guides for it. But its becoming more easy to cross reference overclocking information across different motherboards as I learn more. After trying to limit all background programs and setting cinebench on high in task manager I hit 7190 willing to bet the missing score is because of my security software running. As for voltages i left everything on auto expect my ram voltage. Also wanted to ask, the adia64 memory benchmark will it show better readings as a result of better ram timings or is it something that fluctuates?

Forgot to mention my temps, I dont know what to really say expect ryzen masters readings for my temps are a alot lower then hwinfo64 and speccy. I messed around with the amount of pressure I applied by hand on the water cooler screws till i saw lower temps. Basically made it hand tight without over applying pressure. At first I suspected I applied to much pressure and saw crazy spikes and that stopped after loosening the pressure i applied. I then reapplied thermal paste in a square fashion, I also keep both water cooler radiator fans at about 2100rpm. Which is surprisingly quite and I see a max of 71c underload heavy load in prime 95 compared to 89c when i applied to much pressure on the watercooler.
There's a limit for a negative offset, and depends on the LLC was well. Any negative offset with LLC5 will crash. LLC5 is the weakest one in the AsRock bios. LLC4 onwards is fine with any negative offset. Don't disable llc, leave on auto, or enabled, or normal, try them out. I have no konwledge about any other mobo besides the taichi x370, I can't help with that.
As you speak about security software running in the background, so then you might have even more. You are making too much fuzz and you know the cause of it. Fix it or accept it.

Aida64 bench are okay but SisSandra is better for benchs, comparisons, etc. Yes Aida will show any performance increase, or decrease you might face while messing with timings. And the results also fluctuate. I can do 5 runs and 3 of them will give me 63.9 ns and 2 will be unrealistic, 60+- ns.

Well, all I can say about your temps then is they are great. Few seconds on prime95 smal fft and the 3900x goes to 75C.

Download the ZenTimings and post a print of it. It will show the voltages/timings/etc it's a useful software, or even RM does the job. But I avoid having RM installed so I ZenTimings and HWiNFO64 to monitor sensors.

You could work your way to lower the latency. 69.7 is high.
 
  • Rep+
Reactions: Ark-07

·
Registered
Joined
·
236 Posts
There's a limit for a negative offset, and depends on the LLC was well. Any negative offset with LLC5 will crash. LLC5 is the weakest one in the AsRock bios. LLC4 onwards is fine with any negative offset. Don't disable llc, leave on auto, or enabled, or normal, try them out. I have no konwledge about any other mobo besides the taichi x370, I can't help with that.
As you speak about security software running in the background, so then you might have even more. You are making too much fuzz and you know the cause of it. Fix it or accept it.

Aida64 bench are okay but SisSandra is better for benchs, comparisons, etc. Yes Aida will show any performance increase, or decrease you might face while messing with timings. And the results also fluctuate. I can do 5 runs and 3 of them will give me 63.9 ns and 2 will be unrealistic, 60+- ns.

Well, all I can say about your temps then is they are great. Few seconds on prime95 smal fft and the 3900x goes to 75C.

Download the ZenTimings and post a print of it. It will show the voltages/timings/etc it's a useful software, or even RM does the job. But I avoid having RM installed so I ZenTimings and HWiNFO64 to monitor sensors.

You could work your way to lower the latency. 69.7 is high.
I cant wait for more people to get the motherboard I have maybe then i can safely push things by learning from others, as for the latency i assume tighter ram timings are required? Do you know where i need to focus that attention on the timings as well? As someone said previously ive paid to much for the quality of ram i got when i should of bought 3200mhz kit for the same price and better timings. But been able to reduce the original timings from 18-22-22-42 to 16-19-18-35 (trc reduced to 67 from 86 and tRFC from 630 to 470). It passed prime 95 large fft and memtest86 today. After the testing i tried disabling gdm mode i got a blue screen. Dont think ill squeeze much more out of this ram it works @3800mhz 16-19-19-36 but my flck wont match it even with higher voltages for soc and ram. If anything next i will only try reduce the tRFC some more as it reduces all my other timings as well. Im new to this ram timing stuff.

As for cinebench I finally hit 7224 range didnt occur to me that everyone benchmarks with as little software running wish i had done that before installing all my stuff.
 

Attachments

·
Registered
Joined
·
590 Posts
I cant wait for more people to get the motherboard I have maybe then i can safely push things by learning from others, as for the latency i assume tighter ram timings are required? Do you know where i need to focus that attention on the timings as well? As someone said previously ive paid to much for the quality of ram i got when i should of bought 3200mhz kit for the same price and better timings. But been able to reduce the original timings from 18-22-22-42 to 16-19-18-35 (trc reduced to 67 from 86 and tRFC from 630 to 470). It passed prime 95 large fft and memtest86 today. After the testing i tried disabling gdm mode i got a blue screen. Dont think ill squeeze much more out of this ram it works @3800mhz 16-19-19-36 but my flck wont match it even with higher voltages for soc and ram. If anything next i will only try reduce the tRFC some more as it reduces all my other timings as well. Im new to this ram timing stuff.

As for cinebench I finally hit 7224 range didnt occur to me that everyone benchmarks with as little software running wish i had done that before installing all my stuff.
Yes, tighter timings but that is for another thread, take a look at the [Official] AMD Ryzen DDR4 24/7 Memory Stability Thread and NEW!!! DRAM Calculator for Ryzen™ 1.7.3 (overclocking....

Well, I've read all your novel about your ram purchase in another topic as well. I know they've said it's not good but it is what it is, so, work you way to get lower timings. As you use 32GB my knowledge abou ram and timings is useless to you cause I use 2x8GB single rank and that changes a lot of stuff. Prime95 for RAM testing is not that good, you can try TM5 with 1usmus and Anta's profile, just add more cycles (you can see how in both links above), Ram Test (Karhu) and HCI, they are all good. Memtest86 it is not reliable. I usually do go for TM5 with 1usmus, if passes I go for Anta's, if passes I go for Karhu overnight. If no errors are found overnight while running Karhu that's it, from my point of view it's stable and I won't mess with the timings anymore.

If you cannot match fclk with mclk there's no reason for you to run the ram at 3800, try 3733 and fclk 1867 then. But don't run fclk mclk asynchronous.

By the ZenTimings print your VDDG is too high, you should be fine with just 0950mV/0.950V.\

As for Cinebench, well, in all of my three answer to you I've mentioned background apps holding you back. At least you nailed that down. So leave the CPU alone and go for you ram instead. I belive you can squeeze more performance from your ram tweaking timings. As I said before, I can't tell you which one, as either dual rank (2x16) or 4x8 will behave differently then 2x8GB which is what I habe experience with. Also, depends A LOT on the mobo, bios version, if is daisy chain or t topology. Just to illustrate, I've been rock solid with bios 6.20 since Fev/2020, and couple of weeks ago I've updated to 6.40, all settings are the same but I was getting erros in any ram test software, instantly. Turned out I needed to increase ClkDvrStr from 40 to 60. That was it, problem gone, a guy that helps around those topics above mentioned above nailed the issue. Dealing with ram it's a very tricky thingy. Just try do get your latency down.
 

·
Registered
Joined
·
236 Posts
Yes, tighter timings but that is for another thread, take a look at the [Official] AMD Ryzen DDR4 24/7 Memory Stability Thread and NEW!!! DRAM Calculator for Ryzen™ 1.7.3 (overclocking....

Well, I've read all your novel about your ram purchase in another topic as well. I know they've said it's not good but it is what it is, so, work you way to get lower timings. As you use 32GB my knowledge abou ram and timings is useless to you cause I use 2x8GB single rank and that changes a lot of stuff. Prime95 for RAM testing is not that good, you can try TM5 with 1usmus and Anta's profile, just add more cycles (you can see how in both links above), Ram Test (Karhu) and HCI, they are all good. Memtest86 it is not reliable. I usually do go for TM5 with 1usmus, if passes I go for Anta's, if passes I go for Karhu overnight. If no errors are found overnight while running Karhu that's it, from my point of view it's stable and I won't mess with the timings anymore.

If you cannot match fclk with mclk there's no reason for you to run the ram at 3800, try 3733 and fclk 1867 then. But don't run fclk mclk asynchronous.

By the ZenTimings print your VDDG is too high, you should be fine with just 0950mV/0.950V.\

As for Cinebench, well, in all of my three answer to you I've mentioned background apps holding you back. At least you nailed that down. So leave the CPU alone and go for you ram instead. I belive you can squeeze more performance from your ram tweaking timings. As I said before, I can't tell you which one, as either dual rank (2x16) or 4x8 will behave differently then 2x8GB which is what I habe experience with. Also, depends A LOT on the mobo, bios version, if is daisy chain or t topology. Just to illustrate, I've been rock solid with bios 6.20 since Fev/2020, and couple of weeks ago I've updated to 6.40, all settings are the same but I was getting erros in any ram test software, instantly. Turned out I needed to increase ClkDvrStr from 40 to 60. That was it, problem gone, a guy that helps around those topics above mentioned above nailed the issue. Dealing with ram it's a very tricky thingy. Just try do get your latency down.
Seeing the numbers people are getting on that ram thread is amazing, Im just gonna buy 16-16-16-36 @3600mhz and make sure its samsung bdie as that seems like the best way to go. Just ran the hci memtest it passed. Thats more then 24hrs of ram testing if i add prime95 large fft, adai64 and memtest86 to the mix. I cant get my second timing below 19 I can push 15-19-16-34 with tighter trf and tRFC timings.But since im gonna sell this kit now I might just stick with my current timings. Ill grab a new kit when i get a new 3080 or if by some miracle amd releases a gpu as good or better which i doubt.
 

·
Registered
Joined
·
29 Posts
I just got my 3900XT about a week ago and im curious on the overclocking on it myself. I'm trying to find what the safe voltage to use for daily use, but so much crazy information out there its hard to find anything definitive. In Cinebench clocks didn't go above 3.96ghz and the vcore 1.265, with a score of 7100 and the temp reaching 73 degrees. If all im doing is gaming and don't leave the pc on 24/7, what would be a good spot for voltage without fear of quick cpu degradation?
 

·
Registered
Joined
·
236 Posts
I just got my 3900XT about a week ago and im curious on the overclocking on it myself. I'm trying to find what the safe voltage to use for daily use, but so much crazy information out there its hard to find anything definitive. In Cinebench clocks didn't go above 3.96ghz and the vcore 1.265, with a score of 7100 and the temp reaching 73 degrees. If all im doing is gaming and don't leave the pc on 24/7, what would be a good spot for voltage without fear of quick cpu degradation?
😁 Be sure to quote me if you find a 100% answer, I managed this today (the image) and can push 4.5ghz in ccd0 if i lower ccd1/ccx0 to a limit of 4.3ghz and ccd1/ccx1 to 4.4ghz all @1.28v max, I went with 1.28v because that seems to be high safe margin most say is best. Everything works well cinebench 7800 score and gaming works however prime 95 small fft crashes aftet 5mins at 89c. Like Thomasck said its best to let the cpu manage itself.
 

Attachments

·
Registered
Joined
·
29 Posts
😁 Be sure to quote me if you find a 100% answer, I managed this today (the image) and can push 4.5ghz in ccd0 if i lower ccd1/ccx0 to a limit of 4.3ghz and ccd1/ccx1 to 4.4ghz all @1.28v max, I went with 1.28v because that seems to be high safe margin most say is best. Everything works well cinebench 7800 score and gaming works however prime 95 small fft crashes aftet 5mins at 89c. Like Thomasck said its best to let the cpu manage itself.
Is that what you use for daily usage, the 1.28v? And do you leave the pc on throughout the day?
 

·
Registered
Joined
·
236 Posts
Is that what you use for daily usage, the 1.28v? And do you leave the pc on throughout the day?
I used it for about 12hours then went back to normal settings just couldnt accept a failed prime95 small fft run in the back of my mind. I could go with the camp that sugest 1.35v is the safe max high but Im waiting to see people post long term effects on their own cpus first.
 

·
Registered
Joined
·
29 Posts
I used it for about 12hours then went back to normal settings just couldnt accept a failed prime95 small fft run in the back of my mind. I could go with the camp that sugest 1.35v is the safe max high but Im waiting to see people post long term effects on their own cpus first.
I am in the same boat haha. It's the unknown that keeps me from tinkering with the voltages so early on.
 

·
Registered
Joined
·
590 Posts
I used it for about 12hours then went back to normal settings just couldnt accept a failed prime95 small fft run in the back of my mind. I could go with the camp that sugest 1.35v is the safe max high but Im waiting to see people post long term effects on their own cpus first.
Don't fall for that. 1.35V is high and it's not safe. Yes there's a lot out there but 1.35 is definitely high. It's okay to say that every chip has a different max safe voltage, and to find that disable pbo (if not already disabled) and run prime95 small fft. The voltage you see in there it's your safe voltage based on the cpu's FIT.
 

·
Registered
Joined
·
236 Posts
Don't fall for that. 1.35V is high and it's not safe. Yes there's a lot out there but 1.35 is definitely high. It's okay to say that every chip has a different max safe voltage, and to find that disable pbo (if not already disabled) and run prime95 small fft. The voltage you see in there it's your safe voltage based on the cpu's FIT.
You see this? Gonna try it in a few days and post my results really busy right now
 

·
Registered
Joined
·
22 Posts
When is this tool out? Its been a while since this announcement but I haven't heard much about when its to be released (Linus said "next month" and the video was last month...)
 

·
Registered
Joined
·
590 Posts
You see this? Gonna try it in a few days and post my results really busy right now
Not fan of LLT unless Anthony is around the videos. But I've already told you about this to be realease tool in the previous answers I gave you :)

@alasdairvfr by the end of September bud.
 

·
Registered
Joined
·
236 Posts
Not fan of LLT unless Anthony is around the videos. But I've already told you about this to be realease tool in the previous answers I gave you :)
I cant find a download link guess we still gotta wait :confused:
 

·
Registered
Joined
·
590 Posts
@Ark-07 I've also said by the end of September :p Wait a little more young Padwan :), we are all waiting for that.
 

·
Registered
Joined
·
236 Posts
@Ark-07 I've also said by the end of September :p Wait a little more young Padwan :), we are all waiting for that.
I cant wait, hopefully amd takes something this and gives an update as well
 

·
Registered
Joined
·
29 Posts
Don't fall for that. 1.35V is high and it's not safe. Yes there's a lot out there but 1.35 is definitely high. It's okay to say that every chip has a different max safe voltage, and to find that disable pbo (if not already disabled) and run prime95 small fft. The voltage you see in there it's your safe voltage based on the cpu's FIT.
What about 1.30? So hard to find anything definitive thats straight from AMD themselves. I see more 1.28-1.30 for 24/7 usage, but I would love to see someone from AMD confirm that myself. With PBO on (just made boost to 200) Cinebench R20 my score was 7283 and my max temp was 80 (looks like an average score for a 3900XT it seems).

Edit: Ran Cinebench R20 Single Core and got 537 and have no idea if that is good or bad lol.
 

·
Registered
Joined
·
236 Posts
What about 1.30? So hard to find anything definitive thats straight from AMD themselves. I see more 1.28-1.30 for 24/7 usage, but I would love to see someone from AMD confirm that myself. With PBO on (just made boost to 200) Cinebench R20 my score was 7283 and my max temp was 80 (looks like an average score for a 3900XT it seems).

Edit: Ran Cinebench R20 Single Core and got 537 and have no idea if that is good or bad lol.
Better to go with safe and safer voltages I personally have limited myself too 1.20v if i decided to overclock with a higher llc
 

·
Registered
Joined
·
194 Posts
What about 1.30? So hard to find anything definitive thats straight from AMD themselves. I see more 1.28-1.30 for 24/7 usage, but I would love to see someone from AMD confirm that myself. With PBO on (just made boost to 200) Cinebench R20 my score was 7283 and my max temp was 80 (looks like an average score for a 3900XT it seems).

Edit: Ran Cinebench R20 Single Core and got 537 and have no idea if that is good or bad lol.
Single Core 537 is great. The 3900X scores between 501 and 517 for me, the 517 is with EDC bug. I get multi core between 7300 and 7600 depending on how the Cinebench program is set for Process Priority in Task Manager. When started it goes to below normal and scores 7200ish, change to normal and it scores 7400ish and when set to high priority it gets 7600ish.
 

·
Registered
Joined
·
590 Posts
What about 1.30? So hard to find anything definitive thats straight from AMD themselves. I see more 1.28-1.30 for 24/7 usage, but I would love to see someone from AMD confirm that myself. With PBO on (just made boost to 200) Cinebench R20 my score was 7283 and my max temp was 80 (looks like an average score for a 3900XT it seems).

Edit: Ran Cinebench R20 Single Core and got 537 and have no idea if that is good or bad lol.
Nobody will hear from AMD cause PROBABLY - I'm might be wrong make this statement -, as we've seen around every chip has a different FIT. So AMD can not say what is safe to everyone. Some chips is 1.25, some 1.28, mine seems to be 1.268. As I told Ark-07, turn PBO off and run Prime95 Small FFT and check what is the vcore, that is your "safe".

Your scores are fine and good, specially single core. I get around 522 single core and 7250-7330 multicore depending on how hot or cold is outside.

If you just purchased your 3900X/XT, try full auto. Do all your tests, gaming, benching, take note of them. Enjoy the system as it is, a then try CCX overclocking, try fixed all core overclock and make your own conclusions. I believe you will end up leaving on auto after all, and the only reason to not going with auto is if after setting CCX OC or all core OC you get better single core and multi core performance, which you have to have in mind is the extra performance (if you get there) is worthy having in mind the vcore and heat.

I would not mess with vcore a lot, I would not go above 1.275V. If I give a go, probably I will step back to auto later on cause degradation can happen on the log run, and once degraded there's no way back.
 
141 - 160 of 185 Posts
Top