AMD Ryzen 3000 Builders Thread

[VirtualLarry]

I've basically resigned myself to accepting a 4.20Ghz OC on this new 3600.

I tried for 44.0 multi @ 1.35V vcore (highest vcore that I was willing to try), that was a no-go, wouldn't boot windows, but it would POST.

Tried 43.50 multi @ 1.35V, no-go.

Tried 43.00 multi @ 1.35V, no-go.

So, back to 42.00 multi, now @ 1.25V rather than 1.28125V.

Thinking about this for a bit - these CPUs are binned with a Fmax for each core, of 4.2Ghz (max single-core stock boost). Maybe 0.25Ghz over that, due to PBO or XFR or whatever.

So, being able to finally do an all-core OC of 4.20Ghz with these, is essentially like MCE mode on Intel. Something that I couldn't effectively do on my initial-release CPU. (Although, I probably should have tried it again, with the newest AGESA already on my board. Maybe that was the issue, and I didn't need to change chips?)

But I think that this newer silicon is better, I'm getting 63C max now with 1.25V. Package power is 67W. That's pretty darn good, for 4.20Ghz all-core. WAAAAY better than Zen1. Of course, 7nm TSMC versus 14nm GF. What a difference!

 

[DrMrLordX]

@VirtualLarry

Yeah I meant "false". You won't get accurate readings when using a static OC. You have to run defaults. Try bone stock with CBR20.

 

[Hans Gruber]

I've basically resigned myself to accepting a 4.20Ghz OC on this new 3600.

I tried for 44.0 multi @ 1.35V vcore (highest vcore that I was willing to try), that was a no-go, wouldn't boot windows, but it would POST.

Tried 43.50 multi @ 1.35V, no-go.

Tried 43.00 multi @ 1.35V, no-go.

So, back to 42.00 multi, now @ 1.25V rather than 1.28125V.

Thinking about this for a bit - these CPUs are binned with a Fmax for each core, of 4.2Ghz (max single-core stock boost). Maybe 0.25Ghz over that, due to PBO or XFR or whatever.

So, being able to finally do an all-core OC of 4.20Ghz with these, is essentially like MCE mode on Intel. Something that I couldn't effectively do on my initial-release CPU. (Although, I probably should have tried it again, with the newest AGESA already on my board. Maybe that was the issue, and I didn't need to change chips?)

But I think that this newer silicon is better, I'm getting 63C max now with 1.25V. Package power is 67W. That's pretty darn good, for 4.20Ghz all-core. WAAAAY better than Zen1. Of course, 7nm TSMC versus 14nm GF. What a difference!

Are you running 4200ghz all the time or is that the turbo speed?

 

[VirtualLarry]

Are you running 4200ghz all the time or is that the turbo speed?

All the time. I tried PrimeGrid. Had to go up to 1.30V for 4.20Ghz. Otherwise, I got reboots.

Edit: Now up to 1.32500V, woke up to a rebooted PC.

I hope that I didn't damage something, in my attempts at 4.40Ghz @ 1.35V.

Edit:: Still not "PrimeGrid stable" at 4.20Ghz @ 1.32500V. Going to have to go higher? Or is this freq. a no-go for full-on AVX2 applications?

 

[Hans Gruber]

All the time. I tried PrimeGrid. Had to go up to 1.30V for 4.20Ghz. Otherwise, I got reboots.

4200mhz is the wall for the 3600. When the XT comes out, it's a new ballgame. Your voltage is very good on your chip.

 

[TheGiant]

I've basically resigned myself to accepting a 4.20Ghz OC on this new 3600.

I tried for 44.0 multi @ 1.35V vcore (highest vcore that I was willing to try), that was a no-go, wouldn't boot windows, but it would POST.

Tried 43.50 multi @ 1.35V, no-go.

Tried 43.00 multi @ 1.35V, no-go.

So, back to 42.00 multi, now @ 1.25V rather than 1.28125V.

Thinking about this for a bit - these CPUs are binned with a Fmax for each core, of 4.2Ghz (max single-core stock boost). Maybe 0.25Ghz over that, due to PBO or XFR or whatever.

So, being able to finally do an all-core OC of 4.20Ghz with these, is essentially like MCE mode on Intel. Something that I couldn't effectively do on my initial-release CPU. (Although, I probably should have tried it again, with the newest AGESA already on my board. Maybe that was the issue, and I didn't need to change chips?)

But I think that this newer silicon is better, I'm getting 63C max now with 1.25V. Package power is 67W. That's pretty darn good, for 4.20Ghz all-core. WAAAAY better than Zen1. Of course, 7nm TSMC versus 14nm GF. What a difference!

do you have a power meter? can you please run the primegrid and if you can handbrake x265(slow) and post the results of package and whole system power?
what is your board?
thank you

 

[lightmanek]

I have a quick update after installing cheap Chinese AIO 240mm before my custom loop bits arrive.
This enabled me to create virtually silent PC as there are no fans spinning till my CPU hits 70C, so Cinebench 1T can run for 10 minutes before they kick in (small ommision is my still air cooled Radeon VII which spins silently when idle).

And now my PC can do this:


4350MHz @1.35V with 3600MHz Mem and 1800IF.
Temperature on HWInfo is after running single core test, when running MT it goes up to 82C.
I'm very happy whit this £165 CPU so far!
PS. Tried 1.35V and 1.38V at 4GHz, but R20 is hanging around half way through when temp reaches 88C. Custom loop might get me there, time will tell ...

 

[Elfear]

Traded my 3900X for a newer 3600 and I'm impressed with the results. It's a week 10 batch. Cooling is custom H2O. 4.5Ghz@1.35V (with vdroop, I was at 1.32V with MT and 1.34V with single). Temps hit 65C in the multicore test. Ram/IF was 1800/3600 and 14-15-14-28.

With the above settings it's stable for at least an hour in RealBench.

 

[DrMrLordX]

The 3600XT will likely be a heck of a chip.

Speaking of which, has anyone even gotten a review sample yet? We're less than two weeks out from availability.

 

[VirtualLarry]

The 3600XT will likely be a heck of a chip.

Speaking of which, has anyone even gotten a review sample yet? We're less than two weeks out from availability.

Looking forward to Hardware Unboxed review. Hopefully they attempt an all-core OC with it.

 

[CHADBOGA]

The 3600XT will likely be a heck of a chip.

Not at $250US

 

[VirtualLarry]

Not at $250US

Is there going to be a 3700XT? Wouldn't it be something, if the 3600XT (6C/12T, higher single-core boost clocks) and the 3700X (8C/16T) ended up at the SAME street prices? They're headed that way already.

 

[CHADBOGA]

Is there going to be a 3700XT? Wouldn't it be something, if the 3600XT (6C/12T, higher single-core boost clocks) and the 3700X (8C/16T) ended up at the SAME street prices? They're headed that way already.

I don't think there will be a 3700XT, because of the 3800XT.

Whilst it is not a big deal, I always found it odd how AMD had the 3600 & 3600X at 6 cores, but at 8 cores, rather than have the 3700 and 3700X, they went with 3700X and 3800X.

 

[DrMrLordX]

Not at $250US

Why not? The 3600X launched at the same price.

 

[VirtualLarry]

Why not? The 3600X launched at the same price.

And did that sell as well as the $200 3600, or the $300-310 3700X? No?

 

[DrMrLordX]

And did that sell as well as the $200 3600, or the $300-310 3700X? No?

No idea. But AMD isn't refreshing those CPUs for whatever reason. Only the 3600X, 3800X, and 3900X. The big problem with the original 3600X is that it really wasn't that much better than the 3600. The 3600XT finally makes good on that promise.

Yeah, it's a refresh CPU, which in and of itself is actually a potentially bad sign (as I've mentioned before). But AMD is finally making the 3600X's price point sensible. Ish.

 

[EXCellR8]

I feel like a 3700X refresh would just be in line with 3900X performance... just less cores.

 

[mopardude87]

The 3800x honestly looks very temping as a possible dedicated gaming box chip. I was drooling over a 10900k but ugh to many issues with supply it looks like, that and well its a damn furnace. My room heats up with the rig in my sig, but imagine if i went for that chip LOL. Prob gonna set this rig up in the living room as a htpc/folder.

Most likely won't even bother with a 3800x though, i could wait for the 4000 series stuff and see what Intel offers to compete with it as well and jump on the best option for gaming at that time. I got plans to swap the motherboard, and well i got a 16gb 3600mhz kit in my closet so i need little in the way of rocking out with another system. I got a 2500k rig with a 660 that would just get the 3900x, then i would donate the 2500k/mobo/ram combo to someone in need. Honestly had that setup WAY to long, i wouldn't mind seeing someone else enjoy it for a change. Had it since 2011.

 

[lightmanek]

I have another update, especially for those wondering what's the difference between 240mm cheap AIO and not so cheap custom water loop.



Some of these scores are my personal records now!
Now I can run absolutely stable at 4.4GHz with the same 1.35V which was only good for 4.3GHz on AIO and remember, on box cooler I was only able to achieve 4.2GHz fully stable.
Temperatures captured in HwInfo are during R20 run and you can see how slow my pump and fans are spinning, this is on purpose as I wanted to create silent overclock, when fans were at more reasonable 1300RPM and pump at 2500RPM (4400RPM max.) then my max. temps were 60C during the same run. This with 25C room temperature.

I have EK 480XS rad + 4x120mm CoolerMaster ARGB fans, Corsair XC7 CPU block and Corsair XD5 pump. On top of that 1x240mm (System 1) intake fan at the front of my case and 1x120mm exhaust fan above CPU VRM (System 2).

I love how well memory is working, I have 4 8GB DIMMs running at 3800MHz which is way way better than what my Threadripper 1920X could do, but I still miss QuadChannel

Later I will try 4.5GHz to see where are the limits of this sample.

EDIT - just check SC score compared to my i5 2500K at almost the same freq. I think we can see a small progress (don't even compare it to my score from FX-8320 @4500MHz!).




This is now at 1.45V and still not CB R20 stable, so I will stop here as clearly I'm past optimum volt/clock curve.
Very happy with this little CPU, now bring on XT siblings so I can upgrade (to moar coars).

 

[CHADBOGA]

Why not? The 3600X launched at the same price.

The 3600X's price was nonsensical, as literally every review site stated, not that one needs to have a review site say something so obvious.

With the 4000 series Ryzen's supposedly launching this year, AMD should have priced these XT chips at prices much closer to what current chips sell at.

 

[DrMrLordX]

With the 4000 series Ryzen's supposedly launching this year, AMD should have priced these XT chips at prices much closer to what current chips sell at.

Now you're on to the real reason why AMD is probably selling the XT chips. I expect price increases coming in October.

Personally, I'm still glad that AMD finally made the 3600X price-point sensible(ish) by producing a CPU that's demonstrably faster than the 3600. But it does not bode well for the future.

 

[Makaveli]

For those on Asus X570 Prime Pro boards.

Asus released on june 22

PRIME X570-PRO BIOS 2203
Update AMD AM4 AGESA PI 1.0.0.1

Which from a post I read is suppose to have Zen 3 support.

Also someone on Hardocp forum that has already installed says 5-7 seconds faster boot times.

Haven't updated to this bios yet will wait a couple days.

 

[JimKiler]

For those on Asus X570 Prime Pro boards.

Asus released on june 22

PRIME X570-PRO BIOS 2203
Update AMD AM4 AGESA PI 1.0.0.1

Which from a post I read is suppose to have Zen 3 support.

Also someone on Hardocp forum that has already installed is says 5-7 seconds faster boot times.

Haven't updated to this bios yet will wait a couple days.

My boot went from 17.0 seconds to 12.8 per task manager. It was noticeably faster!

 

[Makaveli]

My boot went from 17.0 seconds to 12.8 per task manager. It was noticeably faster!

I just installed the Bios and it is a much quicker boot, this one was slow to install so you can tell they made alot of changes. Liking this one so far!

 

[mopardude87]

Anyone using a motherboard like the one in my sig, on a up to date bios with no issues or improved performance or anything? I am using the october 2019 release, from what i understand up till recently if not still its the best one to use? I am having issues getting 3600mhz speeds out of my current kit, had to drop to 3200mhz. Also been wanting to enable a 105w TDDP and oddly this bios just randomly resets that setting on reboots with a no post prior.

I could live with the 3200mhz memory maybe, i get it given the kit i am using. I may be asking to much there, but the TDDP issue is beyond me. I love the temps and clocks, but not gonna deal with that issue.