Ryzen Owners Zen Garden

[tabascosauz]

AsRock has an auto overclock feature that turns on PBO, 85 temp limit, and -30 CO

I turned that on and it was fully stable in every test, but I only scored 17,300 in r23 multi.

Guru3D review of 7800x3d has stock r23 multi getting a little over 18k. So, I don't know, it's not a big deal. I did see 4750 the entire time in r23, and boosts to 5050 sometimes in other tests. Temps are what shocked me the most, I never broke 75 celsius in any time, not even prime95. I did get to 77 celsius a few tines in r23 multi... I really like this AIO (I am also using the contact frame, picture below), kind of glad I didn't sell it now. The rig looks amazing in person.

If I just leave my 7800x3d at stock like you @AusWolf I only get 4.2-4.4 all core boosting in r23 multi.

@freeagent It did pass OCCT as well as Prime95 for 20 minutes each (which is enough for me to feel confident anyway).

Score isn't what I wanted, but I think I will just leave the Auto OC on with Expo, it was fully stable and ran cool, so meh.

@GerKNG since you have same board basically, do you have any recommendations for me to get over 18k in r23 score? what OC do you use?

I suspect AMD's response to the VSOC fiasco might also have dialed down the Vcore side of things. Seems that recent production 7800X3Ds on recent BIOSes just generally frequently are tame out of the box. Compared to launch day and early production 7800X3Ds. Mine was the same when stock.

If you aren't monitoring effective clocks through HWInfo or have snapshot polling turned on, then as far as I'm concerned those "clocks" readouts that both of you see are pretty worthless. If you can get to 18k then it should be fine. Double check that PBO is actually having a beneficial effect for you, before automatically turning it on.

I didn't touch ASRock's presets, they didn't seem particularly useful considering temps dont' reach 85. I'm thinking of messing around with some positive Vcore offset though, maybe it can recoup some of the lost performance.

 

[Space Lynx]

I suspect AMD's response to the VSOC fiasco might also have dialed down the Vcore side of things. Seems that recent production 7800X3Ds on recent BIOSes just generally frequently are tame out of the box. Compared to launch day and early production 7800X3Ds. Mine was the same when stock.

If you aren't monitoring effective clocks through HWInfo or have snapshot polling turned on, then as far as I'm concerned those "clocks" readouts that both of you see are pretty worthless. If you can get to 18k then it should be fine. Double check that PBO is actually having a beneficial effect for you, before automatically turning it on.

I didn't touch ASRock's presets, they didn't seem particularly useful considering temps dont' reach 85. I'm thinking of messing around with some positive Vcore offset though, maybe it can recoup some of the lost performance.

I do monitor effective clocks with HWINFO, I have both clock areas completely expanded cause I just think it's neat to look at lol

honestly it's all good, its only like 800 points less, not end of the world. I like how easy it was to do too. I just turned on EXPO and the asrock auto oc feature has several options to pick from, so just being easy i can set it and forget it is nice. In gaming it shouldn't really matter all that much anyway. So, unless @GerKNG has some advice for me, I am just going to leave it this way, he has same mobo and cpu at the moment, so he might know a trick I missed.

@AusWolf SoC voltage didn't need lowered for me, default was 1.230 on this board

 

[tabascosauz]

@Space Lynx oops my bad, no vcore offset available. Only LLC and I can't say if it makes a difference. ymmv

 

[GerKNG]

@Space Lynx
Your CPU should (if not thermal throttling) boost to 4600-4700 in R23 and score around 17000-17500 at stock. (latest bios?)

i played around with CO, LLC, FCLK and tweaked my RAM for months and in the end i gained basically nothing. Zen 4 X3D is just not worth tweaking. (90% of the performance gain was setting my FCLK to 2100 (stable limit).
i gave up on it and rather spend my time playing games instead of benching and stress testing for weeks because Curve Optimizer -30 runs linpack and P95 over night but -10 locks up while opening a folder.

CMOS Reset, EXPO, F10.

Curve Optimizer, FCLK, RAM maxed out:

 

[Space Lynx]

@Space Lynx
Your CPU should (if not thermal throttling) boost to 4600-4700 in R23 and score around 17000-17500 at stock. (latest bios?)

i played around with CO, LLC, FCLK and tweaked my RAM for months and in the end i gained basically nothing. Zen 4 X3D is just not worth tweaking. (90% of the performance gain was setting my FCLK to 2100 (stable limit).
i gave up on it and rather spend my time playing games instead of benching and stress testing for weeks because Curve Optimizer -30 runs linpack and P95 over night but -10 locks up while opening a folder.

CMOS Reset, EXPO, F10.
View attachment 347425

Curve Optimizer, FCLK, RAM maxed out:
View attachment 347427

thank you, I feel the same way now, EXPO and away I go! and hey it rhymes too!

edit: yes latest bios and similar score regardless of oc.

 

[AusWolf]

If I just leave my 7800x3d at stock like you @AusWolf I only get 4.2-4.4 all core boosting in r23 multi.

That's weird. Are you sure you're not hitting any limit? TDC, EDC, and such?

 

[Psychoholic]

If I just leave my 7800x3d at stock like you @AusWolf I only get 4.2-4.4 all core boosting in r23 multi.

That's weird. Are you sure you're not hitting any limit? TDC, EDC, and such?

Yeah, that is weird, my 7800X3d/B650 is around 4.85 all core CB23.
All stock, just enabled EXPO.

 

[Nordic]

You'll get those clocks with or without PBO as long as your cooling is good enough.

My 7800x3d only got 4600mhz max with good air cooling stock. I needed an aggressive CO profile to do better.

Then why do I get the same 5050 MHz single-core and 4800 MHz all-core at default as other people with CO at various settings?

You are just lucky.

I get between 17,900 and 18,100 depending on background programs, star alignment, etc.

I think glnn's 7800X3D was the highest I've seen so far, which was something like 19.5k at just over 5.0GHz effective all core. How exactly he got there I'm not sure but the score does correspond to the clock. Judging from the temps that might be direct die on a golden sample with close to -50 of CO, I can't remember.

I'm at 18.7k @ ~4.85GHz, it seems like most people top out at 4.9 or so all-core. ST seems like 1800-1850 is run of the mill for most people as they cap out at 5050, almost splitting the difference here at 1822.

My 7800x3d can do -50 on almost all cores. My core 8 can't go past -20 though. I can get 19400 in R23. I am now on custom water cooling which enables me to go this fast. I wish I had asynchronous clock engine so I could push further. I bet I could do 5400mhz.

 

[tabascosauz]

My 7800x3d only got 4600mhz max with good air cooling stock. I needed an aggressive CO profile to do better.

You are just lucky.


My 7800x3d can do -50 on almost all cores. My core 8 can't go past -20 though. I can get 19400 in R23. I am now on custom water cooling which enables me to go this fast. I wish I had asynchronous clock engine so I could push further. I bet I could do 5400mhz.

I don't have that much of a spread of a gap, but my worst core drops out somewhere between -40 and -45. Give me some time with it. Been testing with HYDRA because I'm lazy lol

Bone stock mine was similar, around 4.5 or so. Can currently make it past 4.9 and just shy of 19k but it seems a lil low

I only brought up that specific result because I don't think that was BCLK or eCLK at all, just pure CO and board perf. Have seen other people's runs with 5.3GHz+ ST but that doesn't interest me or reveal more about the 7800X3D at all.

 

[freeagent]

That is still pretty good for a octal core. I can squeeze off 23,xxx with PBO on my rustic 5900X. Not sure what stock is though.

 

[tabascosauz]

That is still pretty good for a octal core. I can squeeze off 23,xxx with PBO on my rustic 5900X. Not sure what stock is though.

Stock 3900X and 5900X are like 20-21k. Not there yet but we are getting close. Maybe some crappy 3900Xs are sub-20k

 

[Space Lynx]

My 7800x3d only got 4600mhz max with good air cooling stock. I needed an aggressive CO profile to do better.

You are just lucky.


My 7800x3d can do -50 on almost all cores. My core 8 can't go past -20 though. I can get 19400 in R23. I am now on custom water cooling which enables me to go this fast. I wish I had asynchronous clock engine so I could push further. I bet I could do 5400mhz.

this overclock passes OCCT runs? i had a negative CO pass everything except OCCT once.

 

[Nordic]

this overclock passes OCCT runs? i had a negative CO pass everything except OCCT once.

Yes. I have done extensive stability testing. I have tuned my memory sub timings since that screenshot. I don't know if R23 is very memory dependent so it could be higher.

 

[Space Lynx]

Yes. I have done extensive stability testing. I have tuned my memory sub timings since that screenshot. I don't know if R23 is very memory dependent so it could be higher.

man that is impressive, my mobo has that asynchronus clock engine, i see it in settings, i have no idea how to use it though

@Nordic I just went into bios to try and find it again, i can't find it now, i swear i saw something called that. but asrock boards have about a thousand sub menus in bios... so i lost it, lmao

 

[tabascosauz]

Yes. I have done extensive stability testing. I have tuned my memory sub timings since that screenshot. I don't know if R23 is very memory dependent so it could be higher.

R23 is one of the least memory intensive benchmarks. But it still won't accept obvious instability.

 

[Space Lynx]

Re: 7950X3D mysteriously running below stock speed on prime95 at 61C

The maximum boost frequency for Ryzen processors is the maximum frequency achievable by a single core on the processor running a bursty single-threaded workload. If the workload you are running use multiple cores, you willl see a lower clock frequency. For heavy workloads which draw high...

community.amd.com

looks like this guy was having similar issue as me, and after updating to latest bios also only got 4.4 all core.


I have to admit, getting 400mhz lower than everyone else is kind of annoying, 500mhz in my case. 4.3 all core in r23 a stock with expo.

takes -30 to reach 4700 still 100mhz less... sigh.

prime95 runs extremely cold for me on this chip too, same as him, i find it very confusing. cause im used to prime95 choking me to death on temps lol

 

[tabascosauz]

Re: 7950X3D mysteriously running below stock speed on prime95 at 61C

The maximum boost frequency for Ryzen processors is the maximum frequency achievable by a single core on the processor running a bursty single-threaded workload. If the workload you are running use multiple cores, you willl see a lower clock frequency. For heavy workloads which draw high...

community.amd.com

looks like this guy was having similar issue as me, and after updating to latest bios also only got 4.4 all core.


I have to admit, getting 400mhz lower than everyone else is kind of annoying, 500mhz in my case. 4.3 all core in r23 a stock with expo.

takes -30 to reach 4700 still 100mhz less... sigh.

prime95 runs extremely cold for me on this chip too, same as him, i find it very confusing. cause im used to prime95 choking me to death on temps lol

Perhaps AGESA has made changes again. I'm not on the latest BIOS, there's no need to be - only 2.02 with AGESA 1100.

 

[Space Lynx]

Perhaps AGESA has made changes again. I'm not on the latest BIOS, there's no need to be - only 2.02 with AGESA 1100.

dude look at this... i changed temp limit to 75 celsius and then did the CO 30 thingy. now my temps are insanely good, and i got better all core at 4.7 than 85 celsius limit which was 4.6 all core and 82 celsius temps/ish

man amd confuses the hell out of me, lmao but hey i will take it. im leaving it here. now i am beating stock peoples scores at less temps, eh w.e the hell its doing im cool with it

fyi the red temps don't even matter apparently, its just hitting its artificial wall, and it doesn't lower clocks either when it does. lmao its so confusing

I'm half tempted to the 65 celsius limit and see if i get same scores and better boosts, lol it makes no sense, but it improves performance at colder temps... complete opposite of w.e OC'ing used to be

 

[Cifu]

My Ryzen 7700X got a delid and sorrowfully I do not experience any improvement, instead I got worse temps even after several attempt to clean and re-install Artic Cooling MX-6 paste (I even try Noctua NT-H1 paste). The delid show an soldering defect with the IOD but not on the CCD...

Seems to be the Delid won't help - at least in this case.

 

[tabascosauz]

My Ryzen 7700X got a delid and sorrowfully I do not experience any improvement, instead I got worse temps even after several attempt to clean and re-install Artic Cooling MX-6 paste (I even try Noctua NT-H1 paste). The delid show an soldering defect with the IOD but not on the CCD...

Seems to be the Delid won't help - at least in this case.

Perhaps it's time to tried a direct die solution and better thermal interface like Kryosheet/LM. Stock IHS is a big part of the thermal problem on AM5.

Or you could try the TG High Performance heatspreader, since you've already delidded, @dgianstefani has one

 

[Cifu]

The integrated heat spreader glued back after the delid, only changed to liquid metal.
I won't really fond on direct die solutions, it's simply cost too much and give too few. If I will want better experience i change the CPU an X3D version later but until now I do not experience any CPU limited case.

Originally does not want too much, just wanna try it how it's fare after the delid.
At least for me it's won't too good.

 

[Nordic]

R23 is one of the least memory intensive benchmarks. But it still won't accept obvious instability.

Good to know. I don't think I could get much higher in R23 with my cpu. At least not without asynchronous clock.

I am planning to buy a 9800x3d day 1 and I want to delid it. Star Citizen isn't as cpu hungry as it used to be but it is still cpu hungry.

I am also hoping the 9800x3d will allow me to overclock my memory faster than 6200mhz. Getting even 6200mhz was a struggle. Star Citizen can get 20% higher fps between 6000mhz xmp and 6400mhz tuned.

 

[tabascosauz]

Good to know. I don't think I could get much higher in R23 with my cpu. At least not without asynchronous clock.

I am planning to buy a 9800x3d day 1 and I want to delid it. Star Citizen isn't as cpu hungry as it used to be but it is still cpu hungry.

I am also hoping the 9800x3d will allow me to overclock my memory faster than 6200mhz. Getting even 6200mhz was a struggle. Star Citizen can get 20% higher fps between 6000mhz xmp and 6400mhz tuned.

Yeah but what about 6000 properly tuned? Half decent timings should account for bigger gains than a 200MHz bump to UCLK alone.

 

[Nordic]

Yeah but what about 6000 properly tuned? Half decent timings should account for bigger gains than a 200MHz bump to UCLK alone.

I am not sure. Others have done extensive performance testing and were able to demonstrate as much as a 20% performance increase in game.

In my own synthetic tests the difference between 6000 tuned and 6200 tuned was 3%.

 

[tabascosauz]

I am not sure. Others have done extensive performance testing and were able to demonstrate as much as a 20% performance increase in game.

In my own synthetic tests the difference between 6000 tuned and 6200 tuned was 3%.

It's not that I don't believe you, but I just really have a hard time believing that either of these situations (tuned 6000 vs 6200, tuned 6000 vs 6400) will result in more than 3% of real performance. And on X3D, no less. Maybe if there is a specific well-controlled in-game benchmark that you can run.

Not even tuned vs XMP, usually results in massive double digit real world gains.

As to Star Citizen, not a space simmer but only thing I've ever heard people say about performance is that it is optimized like garbage. Which might open up to a lot of variables and thus not a very reliable test of performance.

The integrated heat spreader glued back after the delid, only changed to liquid metal.
I won't really fond on direct die solutions, it's simply cost too much and give too few. If I will want better experience i change the CPU an X3D version later but until now I do not experience any CPU limited case.

Originally does not want too much, just wanna try it how it's fare after the delid.
At least for me it's won't too good.

How were your temps originally?

I agree on the direct die cost, but you've already delidded. It would honestly be a waste to do all that just to go back to the original crappy IHS. Those aren't exactly made to very exacting physical standards. I only know of the TG IHS, idk if companies like rockitcool make anything for AM5.