Ryzen Owners Zen Garden

[ViperXTR]

oh, one shop here finally dropped 5800X3D price to 299usd

 

[Mussels]

Passed my 3700x systrem to partially disabled uncle, owed him a favour he never collected and figured his 3570k was due to be taken out to pasture and shot.
He had my old old 3570K w/ 16GB DDR3 2400

Now he's got a ryzen 3700x (45W eco mode)
32GB DDR4 3200 (@3133, board has issues with cold boots at 3200 for no known reason)
AX370 Gaming 5
Dark rock slim 120mm (30c idle, 45c load)
2x RX580 in Crossfire (65c/58c)

And of course, I took photos of it.

Pre-answering some comments:

Bottom GPU is using 2x molex to 8 pin adaptor because he lost the PSU cables from a new PSU somehow. Top GPU is raised up slightly to angle some airflow in there, it's over 15c colder than it was in his previous case despite the heat of a second GPU.


I absolutely love this case.
Antec P101 Silent

It's made with no random holes anywhere - 3 perfect 120mm intakes and 1x 120mm exhaust with the extra airflow being pushed out the vents near the GPU's, and the PSU if its facing up.
On the highest setting of the fan controller, you hear a very faint wind tunnel effect up close - but that passive airflow does amazing things for cooling, the CPU fan kept turning off even at load it was that cool - we had to force it to 100% speed to keep it on (which is still silent, be quiet fans are bad-ass)

Unlike most cases, the fronts not made for mixed fan sizes - so theres no screwholes and gaps for that positive air pressure to leak out of. No open top. Just guided airflow right over everything you want it to go over.

It screams "cut back Fractal R5 clone, with less air/noise leakage"

 

[Macro Device]

Trolls are welcome

Is this still the case?

 

[the54thvoid]

Is this still the case?

The next line was that I didn't mind a fight. Troll fights tend to be short lived and can end in bans.

 

[Ruru]

Finally managed to check the curve optimizer. Put -30 there, do I need to do something else?

A discord buddy linked a tutorial for me, but I'd rather ask that can any of you 5800X users tell me which values you use?

Also did some rearrangent with my custom loop, dunno did that help but I also slowed my pump down. In light use my rig is pretty damn quiet

 

[Nordic]

Finally managed to check the curve optimizer. Put -30 there, do I need to do something else?

A discord buddy linked a tutorial for me, but I'd rather ask that can any of you 5800X users tell me which values you use?

Also did some rearrangent with my custom loop, dunno did that help but I also slowed my pump down. In light use my rig is pretty damn quiet

Every chip has a different set of curve optimizer values that work with it. Stress test your cpu with something like OCCT

 

[Macro Device]

The next line was that I didn't mind a fight. Troll fights tend to be short lived and can end in bans.

Roger that. Will come back when I get a mobo for my 7700X.

 

[ViperXTR]

Finally managed to check the curve optimizer. Put -30 there, do I need to do something else?

A discord buddy linked a tutorial for me, but I'd rather ask that can any of you 5800X users tell me which values you use?

Also did some rearrangent with my custom loop, dunno did that help but I also slowed my pump down. In light use my rig is pretty damn quiet

Varies by chip to chip, the recent cinebench 2024 is very sensitive to undervolting and can be a good test.

 

[Ruru]

Every chip has a different set of curve optimizer values that work with it. Stress test your cpu with something like OCCT

Varies by chip to chip, the recent cinebench 2024 is very sensitive to undervolting and can be a good test.

So I'll leave it as is. Oh hell no, not going to spend hours and hours just for some fine-tuning. :/

 

[unwind-protect]

I have a 5950x in an Asus Prime x570 with 128 GB unbuffered ECC RAM (Micron MTA18ASF4G72AZ-3G2R).

Both the BIOS and in-OS facilities report that I run at 3200 MT/s. I thought that wasn't possible or at least not expected with 4 sticks.

Can somebody shed some light on this mystery? Am I just lucky?

 

[Belfaborac]

So I'll leave it as is. Oh hell no, not going to spend hours and hours just for some fine-tuning. :/

-30 across the board is unlikely to be entirely stable. I'd rather spend an hour or two (as opposed to hours and hours) stress testing now, instead of having random crashes down the line.

 

[Count von Schwalbe]

So I'll leave it as is. Oh hell no, not going to spend hours and hours just for some fine-tuning. :/

Just set -30 and loop overnight. No time wasted and you can tell if it is stable.

 

[Space Lynx]

So I'll leave it as is. Oh hell no, not going to spend hours and hours just for some fine-tuning. :/

just do -25 across the board like I did, its safer and literally no difference in gaming performance.

 

[Mussels]

Finally managed to check the curve optimizer. Put -30 there, do I need to do something else?

A discord buddy linked a tutorial for me, but I'd rather ask that can any of you 5800X users tell me which values you use?

Also did some rearrangent with my custom loop, dunno did that help but I also slowed my pump down. In light use my rig is pretty damn quiet

-30 is extremely unlikey to 'just work'

A lot of people who say it does blame their instability on other things because the crashes dont give you an obvious answer like IRQL errors for older systems and memory issues

Just set -30 and loop overnight. No time wasted and you can tell if it is stable.

You cant, because that only tests load. They usually screw up at lighter loads, like when you're using just a browser changing tabs or starting/stopping videos.

 

[Belfaborac]

Wot he said. ^

 

[Mussels]

My 5800x3D runs -30 and an offset undervolt, but it's weird.

My 5800x only does -3 on two cores, while others are around -10. I gave up testing it all and settled for -2 on all cores and left it - it could go days to weeks without issues, then something would change and the instability would return and smash me repeatedly.

 

[Ruru]

-30 across the board is unlikely to be entirely stable. I'd rather spend an hour or two (as opposed to hours and hours) stress testing now, instead of having random crashes down the line.

I guess I could put a stresstest running for a while.

 

[Mussels]

I guess I could put a stresstest running for a while.

Testing load isn't gunna show up any issues with a curve undervolt, it never does.

You need to leave the system at idle for a long time, instead - background tasks windows does will spike the cores randomly and probably trigger the issue fast enough.

Opening and closing a youtube tab tends to work, as does aida64's cache benchmark

 

[Ruru]

Testing load isn't gunna show up any issues with a curve undervolt, it never does.

You need to leave the system at idle for a long time, instead - background tasks windows does will spike the cores randomly and probably trigger the issue fast enough.

Opening and closing a youtube tab tends to work, as does aida64's cache benchmark

Screw curve, I'll go with classic undervolting. Got ~10C lower temps with undervolt vs curve with the same PPT/TDC/EDC values. Scored pretty similar vs the reference results of Cine2024.

Now I'll just make a little more ear-friendly fan profile as these Corsair ML120s I have on the rad sound like jet engine

And here's the AIDA64 cache results.

 

[pavle]

Passed my 3700x systrem to partially disabled uncle, owed him a favour he never collected and figured his 3570k was due to be taken out to pasture and shot.
He had my old old 3570K w/ 16GB DDR3 2400

Now he's got a ryzen 3700x (45W eco mode)
32GB DDR4 3200 (@3133, board has issues with cold boots at 3200 for no known reason)
AX370 Gaming 5
Dark rock slim 120mm (30c idle, 45c load)
2x RX580 in Crossfire (65c/58c)

And of course, I took photos of it...

What's with that upper graphics card being so bent up? Usually they sag down a bit, but isn't that overly bending it in the opposite direction?

 

[Ruru]

What's with that upper graphics card being so bent up? Usually they sag down a bit, but isn't that overly bending it in the opposite direction?

I'd place my bet for tight cable routing as Mussels' PCIe cable comes from up.

And now let's go and do some RAM OC on my 2nd rig. I have no idea how those Corsairs will overclock, but let's put testmem5 with extreme settings to the test.

 

[Nordic]

So I'll leave it as is. Oh hell no, not going to spend hours and hours just for some fine-tuning. :/

You don't have to be physically present during the testing. I recommended OCCT because it has useful per thread testing great for testing Curve Optimizer. Even a traditional undervolt should be tested.

If a few hours of stress testing to confirm stability is toooo much, you are probably better off at stock.

 

[AusWolf]

If a few hours of stress testing to confirm stability is toooo much, you are probably better off at stock.

This. Not to mention, stock is good enough for 99.9% of use cases. It's not 1998 anymore when you could gain +50% on a 300 MHz Pentium II.

 

[Ruru]

You don't have to be physically present during the testing. I recommended OCCT because it has useful per thread testing great for testing Curve Optimizer. Even a traditional undervolt should be tested.

If a few hours of stress testing to confirm stability is toooo much, you are probably better off at stock.

Damn, OCCT has evolved since I last tried it.... probably in the Core2 days though. Lowering the temperature (= fan noise) is the main reason why I want this to run with lower voltages.

Does it work that I put everything to auto in bios and then use Ryzen Master, THEN put those working values to bios?

Also a question for you guys, does running IF faster than 1:1 have any performance benefits?

 

[Klemc]

AGESA Combo AM5 PI 1.0.8.0
ROG STRIX B650E-E GAMING WIFI BIOS 1809

What's up, 1654 (1.07c) broke my PC (was about Samsung thingy anyway), 1.07b works well.