Ryzen Owners Zen Garden

[Mussels]

Oh yeah, i ran eco mode for summer here as well - i LOVE that a simple toggle can change our power consumption so wildly

 

[tabascosauz]

Finally made it over 9000 stable on Geekbench 3 mem score, doesn't mean much outside of OC but hey bandwidth is good



DR E-die @ 3800CL16 + 4650G: 8142
DR CJR @ 3733CL16 + 3700X: 8323
DR CJR @ 3600CL16 + 5900X: 8543
SR A0 B-die @ 4000CL16 + 4650G: 8618
SR A0 B-die @ 4200CL17 + 4650G: 8867
SR A2 B-die @ 4200CL16 + 4650G: 9022

With a view to the coming heatwaves, I've been tinkering with my CPU. 11w output in Cinebench R23 and still feels adequate for typical home use. Curiously, lowering V_soc didn't seem to affect its power draw.

Now I'm ready for those 30c ambients

What VSOC are you running? You might be able to shave off about 1-2W every 50mV lower VSOC, but Matisse/Vermeer I/O die has a pretty high base power draw. It's a pretty big chip.

 

[QuietBob]

What VSOC are you running? You might be able to shave off about 1-2W every 50mV lower

I'm at 1.1v, which I believe was set by the XMP profile. There was virtually zero difference in SOC power between 1.1 and 0.95v as observed in R23. I set both SOC and VDDG to the same value. Perhaps a memory benchmark would show some gains?

 

[tabascosauz]

Ryzen 5000 seems to keep getting better. Now I'm on negative 30 all cores with the curve optimizer and haven't had any errors for seven days. I tried the same settings a few months ago and it made the system hard reset after a few days of use.

I'm pretty curious about this, -30 on all 6 cores should be an outstanding result if reddit is to be believed - what did you test the UV with? Your comment reminded me to finally get around to using Curve Optimizer. The settings today seemed okay through two runs of corecycler (~2h30m) and some other heavy loads I threw at it.

• Stock 142/95/140 limits, +0MHz override, 1x scalar, -10/10/12/12/12/12/15/15/15/15/15/15

Wanted to cut down temps on Core 0 more (temp-wise a bit of a problem child on this chip), but it wouldn't take even one step lower than -10 without failing corecycler. Any temp improvements are welcome on this cooler that performs somewhere between U12S and U12A.

Got snapshot polling enabled here, so core clock = effective clock, two birds w/ one stone. Funny how Windows always prefers Core 7 for background tasks 99.9% of the time? It's the worst core and it shows. Maybe that's not a bad thing, the chip and scheduler are much happier together now than they were in 2019.

 

[INSTG8R]

I'm pretty curious about this, -30 on all 6 cores should be an outstanding result if reddit is to be believed - what did you test the UV with? Your comment reminded me to finally get around to using Curve Optimizer. The settings today seemed okay through two runs of corecycler (~2h30m) and some other heavy loads I threw at it.

• Stock 142/95/140 limits, +0MHz override, 1x scalar, -10/10/12/12/12/12/15/15/15/15/15/15

Wanted to cut down temps on Core 0 more (temp-wise a bit of a problem child on this chip), but it wouldn't take even one step lower than -10 without failing corecycler. Any temp improvements are welcome on this cooler that performs somewhere between U12S and U12A.

Got snapshot polling enabled here, so core clock = effective clock, two birds w/ one stone. Funny how Windows always prefers Core 7 for background tasks 99.9% of the time? It's the worst core and it shows. Maybe that's not a bad thing, the chip and scheduler are much happier together now than they were in 2019.

View attachment 200470 View attachment 200471 View attachment 200473

I agree! I’m running-15 all core that gets me 1.40 max voltage so using that to do the math…I don’t like that it’s arbitrary numbers vs my usual concrete negative offsets of the past I’d like to try -20 but even that seems a little too extreme despite it probably ending up around 1.35 which was my previous offset target on my previous 3700X

 

[8BitZ80]

I'm pretty curious about this, -30 on all 6 cores should be an outstanding result if reddit is to be believed - what did you test the UV with? Your comment reminded me to finally get around to using Curve Optimizer. The settings today seemed okay through two runs of corecycler (~2h30m) and some other heavy loads I threw at it.

• Stock 142/95/140 limits, +0MHz override, 1x scalar, -10/10/12/12/12/12/15/15/15/15/15/15

Wanted to cut down temps on Core 0 more (temp-wise a bit of a problem child on this chip), but it wouldn't take even one step lower than -10 without failing corecycler. Any temp improvements are welcome on this cooler that performs somewhere between U12S and U12A.

Got snapshot polling enabled here, so core clock = effective clock, two birds w/ one stone. Funny how Windows always prefers Core 7 for background tasks 99.9% of the time? It's the worst core and it shows. Maybe that's not a bad thing, the chip and scheduler are much happier together now than they were in 2019.

View attachment 200470 View attachment 200471 View attachment 200473

It's been perfectly stable for almost three weeks now, but some of the cores don't like Prime95. I ran CoreCycler for three hours and cores 1 and 5 failed. I'll try dialling them back a bit to see where their limits are.

UPDATE: I had to tweak cores 0, 1 and 5 down to negative 25, the others are still on negative 30. Now it runs Prime95 for hours with no issues.

 

[toilet pepper]

From what I read in reddit and my experience as well, p95 might seem stable when dealing with Curve Optimizer but in reality it really isn't.

I've had mine pass p95 for hours but simply crash on random low workloads. I wish there was a simpler way to tell which of the cores are bad to make tuning these easier.

 

[freeagent]

I've had mine pass p95 for hours but simply crash on random low workloads.

This is the main reason why I stopped using that program years ago.

 

[8BitZ80]

From what I read in reddit and my experience as well, p95 might seem stable when dealing with Curve Optimizer but in reality it really isn't.

I've had mine pass p95 for hours but simply crash on random low workloads. I wish there was a simpler way to tell which of the cores are bad to make tuning these easier.

I noticed the same thing in the earlier months. I would run Prime95 with no issues and then crash when I'm watching Youtube or something. The strange thing is that those crashes are gone, but the cores have also changed in their worker failure behavior. E.g. Core 0 used to be perfectly fine at negative 30 and Core 5 wouldn't tolerate anything above 20. My best guess is that using a better thermal paste or BIOS version has changed my results somehow.

 

[tabascosauz]

Kinda crazy how much curve optimizer can accomplish even with PBO completely disabled. The MT score is intentionally 142W-limited, and still improved. The previously higher ST score in the R23 thread is 1635 on a 360mm custom loop, didn't think I could beat that on a 140mm downdraft cooler lol. Last I heard there are some crazy people out there managing 1700+ on more exotic cooling.........

From what I read in reddit and my experience as well, p95 might seem stable when dealing with Curve Optimizer but in reality it really isn't.

I noticed the same thing in the earlier months. I would run Prime95 with no issues and then crash when I'm watching Youtube or something. The strange thing is that those crashes are gone, but the cores have also changed in their worker failure behavior. E.g. Core 0 used to be perfectly fine at negative 30 and Core 5 wouldn't tolerate anything above 20. My best guess is that using a better thermal paste or BIOS version has changed my results somehow.

Y'all running vanilla P95 or using it indirectly through a per-core testing script? Vanilla AVX P95 is pretty bad at just about everything.

On the other hand per-core scripts running SSE P95 seem to be the best fit for curve optimizer; the SSE test basically maxes out clocks and the boost algorithm generally applies pretty hefty offsets for AVX so it should be unlikely that settings are consistently stable in the script but not elsewhere.

The thermal paste probably made no difference but it's definitely possible for Vcore behaviour to change with a new BIOS.

 

[8BitZ80]

Kinda crazy how much curve optimizer can accomplish even with PBO completely disabled. The MT score is intentionally 142W-limited, and still improved. The previously higher ST score in the R23 thread is 1635 on a 360mm custom loop, didn't think I could beat that on a 140mm downdraft cooler lol. Last I heard there are some crazy people out there managing 1700+ on more exotic cooling.........

View attachment 200704 View attachment 200701




Y'all running vanilla P95 or using it indirectly through a per-core testing script? Vanilla AVX P95 is pretty bad at just about everything.

On the other hand per-core scripts running SSE P95 seem to be the best fit for curve optimizer; the SSE test basically maxes out clocks and the boost algorithm generally applies pretty hefty offsets for AVX so it should be unlikely that settings are consistently stable in the script but not elsewhere.

The thermal paste probably made no difference but it's definitely possible for Vcore behaviour to change with a new BIOS.

Yeah I was running CoreCycler before and after, plus some regular Prime95. After the adjustments I was getting up to Iteration 7 until I ended the program with no errors.

 

[toilet pepper]

From what I can tell, core cycler is using p95 as well. I just test using CB20 and usually my rig restarts when some settings are too aggressive. I could tune some more but at this point its more trouble than what its worth. My 3080 is around 20 inches from the AIO I'm using inside the NR200p.

 

[tabascosauz]

From what I can tell, core cycler is using p95 as well. I just test using CB20 and usually my rig restarts when some settings are too aggressive. I could tune some more but at this point its more trouble than what its worth. My 3080 is around 20 inches from the AIO I'm using inside the NR200p.

Unless you're manually configuring P95 to run a non-AVX instance on one specific core at a time, which is what corecycler does, it doesn't quite "use" P95.

The point is that P95 is a decent stress test.........just not in its default form, where it just runs useless AVX all-core, which proves nothing because the boost algorithm recognizes it's AVX P95 and slashes clocks. To test per-core stability, you need to max out the boost clock to the limit on that specific core, hence running no-AVX P95 and running only on 1 core at a time. Running AVX and using more cores are both things that reduce clocks, which is pointless when verifying individual core stability.

Cinebench is terrible for stability testing. It's great for some quick performance numbers or showing off scores, but it will not do anything for you if daily stability is what you're after. And the ST test will only ever run on your preferred 2 cores.

Yeah I was running CoreCycler before and after, plus some regular Prime95. After the adjustments I was getting up to Iteration 7 until I ended the program with no errors.

Made me think I was doing something wrong to barely finish Iteration 2 after 3 hours, then I realized you have half the number of cores, that's probably why

 

[8BitZ80]

Made me think I was doing something wrong to barely finish Iteration 2 after 3 hours, then I realized you have half the number of cores, that's probably why

Half the number of cores AND a 200MHz lower maximum boost clock lol

I still wonder why AMD went with higher clocks on the high core count SKUs. A reduced core count always helps with high clocking ability.

 

[GoldenX]

Stock on the Ryzen APUs is zero right now here, and I was begged to give mine so an audio/video editor student could start.
So I'm getting a 5600X

 

[freeagent]

So.. I am looking to upgrade my 5600x..

I was going to put it in my kids machine instead of the 3600XT..

I was considering the 5800X because I like the price, when you compare it to the 5900X.. which is in my cart right now on Amazon.ca because its about 110 cheaper than Memory Express.. but its still 800 bucks. I have never spent that much on a CPU.. So holy shit. But I want everything. I want the boost, I want the memory performance.. I don't want the price though! I might sell off some older hardware to help pay it off. And I probably wont be moving to AM5 or whatever it will be called until later in its life cycle.

I could use the horsepower for some video stuff too, but I don't do much right now, and its not professional in the slightest.

 

[toilet pepper]

So.. I am looking to upgrade my 5600x..

I was going to put it in my kids machine instead of the 3600XT..

I was considering the 5800X because I like the price, when you compare it to the 5900X.. which is in my cart right now on Amazon.ca because its about 110 cheaper than Memory Express.. but its still 800 bucks. I have never spent that much on a CPU.. So holy shit. But I want everything. I want the boost, I want the memory performance.. I don't want the price though! I might sell off some older hardware to help pay it off. And I probably wont be moving to AM5 or whatever it will be called until later in its life cycle.

I could use the horsepower for some video stuff too, but I don't do much right now, and its not professional in the slightest.

The 5800x is $800CAD!??! And I thought prices in my country were always insane.

 

[freeagent]

The 5800x is $800CAD!??! And I thought prices in my country were always insane.

No the 5900X is 790ish and the 5800X is roughly 600.

I'm probably going to go with the 5900, I will buy something when I get home..

 

[toilet pepper]

That's roughly the same price here. You already have the 5600x might as well double your core count and get the 5900x. I have the 5800x and it is a pain to cool.

 

[tabascosauz]

So.. I am looking to upgrade my 5600x..

I was going to put it in my kids machine instead of the 3600XT..

I was considering the 5800X because I like the price, when you compare it to the 5900X.. which is in my cart right now on Amazon.ca because its about 110 cheaper than Memory Express.. but its still 800 bucks. I have never spent that much on a CPU.. So holy shit. But I want everything. I want the boost, I want the memory performance.. I don't want the price though! I might sell off some older hardware to help pay it off. And I probably wont be moving to AM5 or whatever it will be called until later in its life cycle.

I could use the horsepower for some video stuff too, but I don't do much right now, and its not professional in the slightest.

Are we looking at the sane product pages? 5800X is $569 and 5900X is $769 (same price I paid 4 months ago) at ME.

Both with crazy amounts of availability. I'm surprised that AMD slipped the 5800X price cuts under the radar and no one noticed, it's still a bit more expensive than the 11700K but the boards aren't nearly as expensive. That said the 11700 is still like $100 cheaper. At launch the 5800X was like $639, I took one look, laughed, and put the 5900X in my cart.

 

[freeagent]

I was going to get the 5900X from Amazon it was the cheapest, and would have had 1 day shipping. I just got home and went to buy it and it went up a hundred bucks. I went back to memory express and put it back in my cart wanted to pick it up in store, but they aren't doing that here because the virus is bad. I would have been 885.48 delivered from M.E. On a whim I checked back into Newegg and they had it in stock, not the third party seller.. so I bought one from their newest batch for 808 delivered.

She doesn't know yet tee hee..

I'm gonna be in trouble if I get caught lol.. Might just flip my 5600X and 3600XT to save some bacon..

 

[tabascosauz]

I was going to get the 5900X from Amazon it was the cheapest, and would have had 1 day shipping. I just got home and went to buy it and it went up a hundred bucks. I went back to memory express and put it back in my cart wanted to pick it up in store, but they aren't doing that here because the virus is bad. I would have been 885.48 delivered from M.E. On a whim I checked back into Newegg and they had it in stock, not the third party seller.. so I bought one from their newest batch for 808 delivered.

She doesn't know yet tee hee..

I'm gonna be in trouble if I get caught lol.. Might just flip my 5600X and 3600XT to save some bacon..

1. boys gotta be boys sometimes hon, just have to accept it
2. i think i deserve a little present to myself for surviving the virus
3. i didn't change anything, it's still a Ryzen? you sure you didn't just buy a purse and forget about it

I hope you get a gold chip but I'd caution against expecting all-core OC on air or 2000+ IF, not a 2CCD's strong suit. I really think you should keep the 5600X around, never let go of a unicorn like that

 

[freeagent]

You are absolutely right man! Ok yes.. I did just build a computer last November.. but outside of that I have not bought myself anything since we started our family like 9 years ago.. I footed the bill for everything, I am still paying Now she is working so it is helpful

I hope I get a good one too.. I had a good feeling when I bought the 5600X, I didn't get it when I was going to buy a 5800X, but I got that feeling today when I saw they were back in stock at Newegg, rather than 3rd party hax. 2K is probably unrealistic.. which is... this is hard to say.. ok.. because I only have 3200C14

Yeah I don't think I will get rid of the 5600X, she's my other baby. Such a wicked CPU. Only reason I considered the 5800X is because guys were hitting 2K with it.. whatever. Its just a number..

I would still be ok with 1900.. wold be groovy if it would do that no muss no fuss.. but probably not going to happen like that. My 5600X doesnt like pushing 4 sticks as much as my XT does, which is weird.

 

[Space Lynx]

You are absolutely right man! Ok yes.. I did just build a computer last November.. but outside of that I have not bought myself anything since we started our family like 9 years ago.. I footed the bill for everything, I am still paying Now she is working so it is helpful

I hope I get a good one too.. I had a good feeling when I bought the 5600X, I didn't get it when I was going to buy a 5800X, but I got that feeling today when I saw they were back in stock at Newegg, rather than 3rd party hax. 2K is probably unrealistic.. which is... this is hard to say.. ok.. because I only have 3200C14

Yeah I don't think I will get rid of the 5600X, she's my other baby. Such a wicked CPU. Only reason I considered the 5800X is because guys were hitting 2K with it.. whatever. Its just a number..

I would still be ok with 1900.. wold be groovy if it would do that no muss no fuss.. but probably not going to happen like that. My 5600X doesnt like pushing 4 sticks as much as my XT does, which is weird.

yep, its just a number. same as me lately. on my 5600x i changed the paste from mx-4 to mx-5 cause i was just bored. regrets. lost 2 celsius across the board, everything else the same including ambient.

very frustrating... should have just left it alone... LOL

 

[freeagent]

I saw that man, that’s a bummer. My cooler came with TF7, it’s pretty good! Might be cheap too? If you have a few bucks you should check it out!

I do have to say I am a little nervous when it comes to cooling..

I talk a lot of shit when it comes to AIO’s. And I put Thermalright on this magical platform because I have been smoking a ton of weed over the last 6 months. I sure hope their coolers are as good as I keep saying they are.. because summer is pretty much here now.. find out in a few days I reckon..