Ryzen Owners Zen Garden

[AusWolf]

I will become an AMD owner soon, i already have an APU 4650G Probut in a DeskMini X300, that works well at 45w without XMP. I don't even know if XMP enabled will make the PC asking more power, that i don't want.

I think about 7950X3D, for the bottlenecked 4070ti i own in the PC i use to play games and nothing else (from emulators to only 2D or 3D intensive games).

Welcome.

XMP asks for more power from the RAM and the CPU's memory controller, but only marginally. You have nothing to lose by enabling it.

If you only game, don't waste your money on a 7950X3D. A 7800X3D will give you the same results.

EDIT, i mistakenly read june bc i'm French and date formats... so i could easyly wait 7820.

I know that feeling. DD/MM/YYYY (going from smaller to bigger units) or YYYY/MM/DD (going from bigger to smaller units) both make perfect sense, but MM/DD/YYYY is the stupidest format ever.

 

[Klemc]

I know that feeling. DD/MM/YYYY or YYYY/MM/DD both make perfect sense, but MM/DD/YYYY is the stupidest format ever.

Yes, the best is Y/M/D, it's put in order (alphabetinumerically), now i'm ready to wait one month for the 7800X3D, allright

 

[Nordic]

I ran -10 CO on all cores for many months because it was stable enough. Eventually I did have a crash in a pivotal gaming moment so I decreased it to -5 on all cores.

Core cycler tells me on my preferred cores that there are errors with p95 at -5. It has not crashed yet. Y Cruncher doesn't tell me if there are errors but crashes when it fails. I have ran both p95 and y cruncher 1 hour on each core without a crash. It seems stable enough.... Right?

Core 00 >>> -5
Core 01 >>> -5
Core 02 >>> -30
Core 03 >>> -20
Core 04 >>> - 5
Core 05 >>> -30
Core 06 >>> -25
Core 07 >>> -25
Core 08 >>> -20
Core 09 >>> -20
Core 10 >>> -20
Core 11 >>> -15
Core 12 >>> -30
Core 13 >>> -30
Core 14 >>> -20
Core 15 >>> -30

 

[3x0]

I have ran both p95 and y cruncher 1 hour on each core without a crash. It seems stable enough.... Right?

CoreCycler author recommends 12h per core to be fully stable. You decide if it's worth the time

 

[trparky]

I had to back mine off to -5, I was having some weird stuttering in Windows. I tried -30 but damn, that was all kinds of weird. Kept stepping down until -5 was found to produce no stuttering.

 

[Nordic]

CoreCycler author recommends 12h per core to be fully stable. You decide if it's worth the time

I saw that in the read me. I was strongly considering running one core a night once I got through the one hour iterations.

 

[johnspack]

Darn threadripper... just keeps costing more. Got the wrong dam cooler.... baby noctua. This time got the right one, but you have to buy the 2nd 140mm fan extra.
So that cost me 170can...

 

[tabascosauz]

CoreCycler author recommends 12h per core to be fully stable. You decide if it's worth the time

I saw that in the read me. I was strongly considering running one core a night once I got through the one hour iterations.

12h per core is 100% useless if you don't also change other parts of the config to be more effective. Test a wider range of FFT sizes, P95 vs ycruncher, etc. The default FFT size is reasonably time effective but it's narrow.

My old testing config was like 1h iterations, 5-6 iterations per core. But it was too time consuming so in the end I went back to using default config as a quick test only

 

[Norton]

Darn threadripper... just keeps costing more. Got the wrong dam cooler.... baby noctua. This time got the right one, but you have to buy the 2nd 140mm fan extra.
So that cost me 170can...
View attachment 284324

Noctua NF-P14R Redux 1500 PWM fans usually go for cheaper- have a pair on the NH-U14S on my main rig and paid around 15 USD each for them. Performance is nearly the same as the NF-A15

 

[tpa-pr]

NTLite has a staggering amount of options. Not a lot of handholding but it's still a somewhat guided process, reminds me of walking through IKEA. You start with installer options, then setup, then services and processes etc. I'll look at it again in the morning. Yes, telemetry and whatnot is in there

I'd recommend a bit of care when using NTLite to make cut-down Windows installations. With 10 and 11 there's quite a bit of spaghetti code/components going on and you might end up removing something "unnecessary" which is actually vital. Same with any of those PowerShell scripts and executables you see on Github.

 

[tabascosauz]

I'd recommend a bit of care when using NTLite to make cut-down Windows installations. With 10 and 11 there's quite a bit of spaghetti code/components going on and you might end up removing something "unnecessary" which is actually vital. Same with any of those PowerShell scripts and executables you see on Github.

I mean, yes......I've been burned by a lot of those "debloat" scripts in the past which is why I only some bare minimum tweaks, mostly to do with privacy and BS bundled apps. Hyper-V items, some Explorer settings, reboot on BSOD. Things that I used to have to do after installation via command line, may as well get it done in NTLite.

 

[Veseleil]

I'd recommend a bit of care when using NTLite to make cut-down Windows installations. With 10 and 11 there's quite a bit of spaghetti code/components going on and you might end up removing something "unnecessary" which is actually vital. Same with any of those PowerShell scripts and executables you see on Github.

That's why it's much simpler to get one of the already tweaked ISOs of your liking from the reputable sources like the Ghost Spectre, K Lite, Sasnet, or from a various ones from an archive.org (look for comments).
With tools like NTLite you need to have a huge knowledge, gathered by either an experience from working for the MS on the Windows OS itself, or by an insane amount of trial and error testing. It is very time consuming, and just one wrong setting could render your iso unbootable. Been there, done that, Ghost Spectre saved the day.

 

[Radxge]

Recently bought a 7700 (could not resist $300 + free Jedi game), which is my first AMD cpu since Athlon 64 X2.

4K 120HZ gamer here with a 4090 and I have seen a significant performance uplift versus my 9700K in Ray Tracing games .
Frame rates are more stable and less subject to stuttering.

Also very happy with AsRock X670E Pro RS and system has been rock solid so far.

 

[freeagent]

Recently bought a 7700 (could not resist $300 + free Jedi game), which is my first AMD cpu since Athlon 64 X2.

4K 120HZ gamer here with a 4090 and I have seen a significant performance uplift versus my 9700K in Ray Tracing games .
Frame rates are more stable and less subject to stuttering.

Also very happy with AsRock X670E Pro RS and system has been rock solid so far.

Attaboy

I built this one in 2020 and was also my first since 939. It has been great! Have fun and enjoy

 

[Count von Schwalbe]

Recently bought a 7700 (could not resist $300 + free Jedi game), which is my first AMD cpu since Athlon 64 X2.

4K 120HZ gamer here with a 4090 and I have seen a significant performance uplift versus my 9700K in Ray Tracing games .
Frame rates are more stable and less subject to stuttering.

Also very happy with AsRock X670E Pro RS and system has been rock solid so far.

Welcome to the Garden!

Keep your eye out when the the X3D models drop, they are rumored to help the frame pacing.

 

[AusWolf]

Recently bought a 7700 (could not resist $300 + free Jedi game), which is my first AMD cpu since Athlon 64 X2.

4K 120HZ gamer here with a 4090 and I have seen a significant performance uplift versus my 9700K in Ray Tracing games .
Frame rates are more stable and less subject to stuttering.

Also very happy with AsRock X670E Pro RS and system has been rock solid so far.

Welcome and enjoy your new system!

The 7700 non-X is the best thing that's happened to AM5 so far, although I don't regret getting the X version which runs hotter and is more expensive. But I'm jealous of that free Jedi game!

 

[trparky]

What kind of improvements will be seen if I were to go from a DDR5-6000 CL36 to a DDR5-6000 CL32 kit?

 

[AusWolf]

What kind of improvements will be seen if I were to go from a DDR5-6000 CL36 to a DDR5-6000 CL32 kit?

Very minor, if any at all, in my opinion. Zen 4 doesn't seem to be as sensitive to memory speed and latency as previous generations.

 

[trparky]

Zen 4 doesn't seem to be as sensitive to memory speed and latency as previous generations.

But what about this?

He's basically saying that it's quite easy to cripple Zen 4 with cheap memory.

 

[freeagent]

He's basically saying that it's quite easy to cripple Zen 4 with cheap memory.

Its quite easy to cripple any machine with cheap memory..

 

[trparky]

Its quite easy to cripple any machine with cheap memory..

You do have a point.

 

[AusWolf]

But what about this?

He's basically saying that it's quite easy to cripple Zen 4 with cheap memory.

Sure, but he's comparing 5200 MHz CL 40 to 6000 MHz CL 30 which are basically the worst and best you can do with Zen 4. Your current kit is already much closer to Steve's best case.

 

[trparky]

Your current kit is already much closer to Steve's best case.

So, basically the kit that Microcenter threw in for free was a damn good kit.

 

[AusWolf]

So, basically the kit that Microcenter threw in for free was a damn good kit.

I would say so, yes.

Edit: The other thing is, Steve's testing with a 4090. Unless you have one as well, you won't see such big differences even if you drop down to 5200 MHz for some reason.

 

[The King]

So, basically the kit that Microcenter threw in for free was a damn good kit.

There is another video here.
Like ZEN 3 tuning RAM can increase performance on ZEN 4.