AMD Ryzen 5000 Builders Thread

[B-Riz]

	Cores Threads	Base Freq	Turbo Freq	L3 Cache	TDP	MSRP
Ryzen 9 5950X	16c/32t	3400	4900	64 MB	105 W	$799
Ryzen 9 5900X	12c/24t	3700	4800	64 MB	105 W	$549
Ryzen 7 5800X	8c/16t	3800	4700	32 MB	105 W	$449
Ryzen 5 5600X	6c/12t	3700	4600	32 MB	65 W	$299

5600X comes with cooler, the rest do not.

above from https://www.anandtech.com/show/1614...en-3-on-nov-5th-19-ipc-claims-best-gaming-cpu

 

[DrMrLordX]

@Space Tyrant

XMP can screw with a lot of voltages (like vSoC) without telling you it's going to do so. If you are not hand-tuning your memOC, now would be a good time to do so.

Otherwise, yeah, try messing with different LLC settings and voltage offsets to see how it affects PBO and non-PBO performance. It could take a little while.

 

[Kenmitch]

MicroCenter (at least the one near me) has the F4-3600C16D-32GTZNC kits.

Is that the same thing? My inclination after reviewing is to give G.skill a re-try and ensure I get B-die, but the added C to the part name gives me fear.

Nope. Look at the timings. The ones I linked are 16-16-16-36 those ones are 16-19-19-39 and probably not b-die.

 

[Space Tyrant]

@Space Tyrant

XMP can screw with a lot of voltages (like vSoC) without telling you it's going to do so. If you are not hand-tuning your memOC, now would be a good time to do so.

Otherwise, yeah, try messing with different LLC settings and voltage offsets to see how it affects PBO and non-PBO performance. It could take a little while.

I do fine-tune my memory timings, but I usually start from the XMP profile. I didn't realize that XMP might affect other voltages though. I have new memory coming in a few days and will attack the mem timings with an eye on voltages. After I get the memory and fabric settled, I'll be looking into voltage offset possibilities. The board seems to have defaulted to an LLC setting of '2' -- and may have a few other unexpected (to me, at least) default settings.

 

[reqq]

yeah didnt know that either. gonna turn XMP off..need full controll..

 

[Noid]

5600 at NewEgg NOW

Yesterday one of my nearby MicroC had " One " 5900x in stock
I could not believe my eyes.
Only available in store.
I drove 45 mins to be told it was gone an hour before.

I thought it was worth a try.

As for XMP profiles = I never use them.

[ edit ] just got back from another hour trip to the other MC that had 2 5900x's on the shelf. --- yea ... too late again.

 

[MrTeal]

MicroCenter (at least the one near me) has the F4-3600C16D-32GTZNC kits.

Is that the same thing? My inclination after reviewing is to give G.skill a re-try and ensure I get B-die, but the added C to the part name gives me fear.

3600 16-19-19-39 is probably not going to be B-die.

 

[amrnuke]

Nope. Look at the timings. The ones I linked are 16-16-16-36 those ones are 16-19-19-39 and probably not b-die.

Thank you -- makes sense.

At MicroCenter, have a couple of B-die options:
G.skill Flare X 2x8GB 3200CL14 (B-die)
G.skill Ripjaws V 2x8GB 3600CL16 (B-die)
Same price, $109

Newegg has:
Trident Z RGB kits:
4000CL18 - 2x8GB $149 (this is only RAM on QVL list for my mobo!)
3600CL16 - 2x8GB $134
Trident Z Neo:
3600CL16 - 2x8GB $144
3600CL14 - 2x8GB $189

Other option is Crucial Ballistix line, though it's harder to elucidate what is truly Micron E-die as there is no simple tool like the B-die finder, and none of the RAM over 3200 is on my motherboard's QVL.

My inclination is to grab the 4000 kit. It's on QVL, would be fairly likely to run at 3600 or 3800 with tight-ish timings (I have seen multiple reports of 3600CL14, 3800CL16), offers an option in the future if FCLK 2000 becomes a possibility -- even if it requires a motherboard upgrade to X570 or B550), and isn't as overpriced as the Neo line.

 

[amrnuke]

5600 at NewEgg NOW

Yesterday one of my nearby MicroC had " One " 5900x in stock
I could not believe my eyes.
Only available in store.
I drove 45 mins to be told it was gone an hour before.

I thought it was worth a try.

As for XMP profiles = I never use them.

[ edit ] just got back from another hour trip to the other MC that had 2 5900x's on the shelf. --- yea ... too late again.

FYI, at least at my Dallas location, they tend to have the stock online updated around midnight or at the latest 5-6AM when they get new CPUs in stock, so you know whether to get there at opening or not. No one is camping out any more and they take an hour or two to sell out. What I did to get a 5600X was just check on it when I woke up as usual at 6AM and if it's in stock, go to work, get stuff done, and head over to be there around opening time. Waited a couple of minutes (there were people in line to buy 3080, 6900XT, etc) but they had solid stock (double digits) still available even though I didn't get there til 45 minutes after opening.

 

[Elfear]

I must be doing something wrong with PBO. I've been messing around with it for 3 days and the highest clocks I see are with -10 offset for all cores, +50 Mhz, voltage all on Auto, PBO limits set to (280/235/245). My two best cores are both on the same CCD (3 & 4) and both boost to 5050-5060 effective. CB23 scores are:
ST - 1643
MT - 29839

So I decided to try turning off PBO but kept the wattage limits raised to 280/235/245. CB23 scores are:

ST - 1636
MT - 29716

When I try and lower the voltage offset to the other 14 cores, I get BSODs.

Here is what I tried so far:

All core -10 / +50, PBO limit at 280/235/245	Best core (boost max/effective max) - 5090/5059	2nd best core - 5090/5014	MT test max (boost/effective) = 4666/4488; avg (boost/eff) = 4470/4350
-5 / +50	5065/4935	5040/4930
-5 / +50, vcore = +.05	5040/4999	5040/4950
-10 / +50, vcore = +.05	5065/5033	5065/4982
0 / +100	5016/4977	4991/4923
0 / +125	4990/4966	4990/4921
All stock, PBO limits at stock	5040/5033	5040/5034	MT test max (boost/effective) = 4242/4005; avg (boost/eff) = 4005/3820
All stock, PBO limits at 280/235/245	5040/5010	5040/4986	MT test max (boost/effective) = 4616/4585; avg (boost/eff) = 4460/4350

Am I doing something wrong or is my chip just a dud overclocker? Seems like just raising the PBO power limits gets you 98% of what messing around with the curve optimizer does.

 

[EXCellR8]

So, I did a dumb or two and forgot that the 5600X was not natively supported on the 550 board I got so had to take a couple of detours. Of course I realized this AFTER the system was fully assembled but that, by far, wasn't even the worst part. After then reassembling the PC with a first generation Zen chip I was like oh right, same issue duh. Ripped that back out along with the 3700X from my main PC and then put that in the 550 in order to flash the BIOS; then the 5600X went back in. Ugh...

...I was going to omit the part where I mistakenly threw the 1500X back in after flashing the 550 board but I digress (that did 100% happen)

I'm up and running now though; since I was planning to re-seat the 3700X anyway because temps seemed a touch too high--but that one certainly beat me up real nice.

/tired

 

[Tup3x]

I'm tuning for maximum power efficiency and single threaded performance so for now I'll just added +50 MHz for max boos and -8 for all core curve optimiser (probably could go even lower but for now I'll just use this). Boosts to 5 GHz max, all core load is obviously lower but that's fine when there's 12 cores available... Power usage stays in check.

 

[Space Tyrant]

So, I did a dumb or two and forgot that the 5600X was not natively supported on the 550 board I got so had to take a couple of detours. Of course I realized this AFTER the system was fully assembled but that, by far, wasn't even the worst part. After then reassembling the PC with a first generation Zen chip I was like oh right, same issue duh. Ripped that back out along with the 3700X from my main PC and then put that in the 550 in order to flash the BIOS; then the 5600X went back in. Ugh...

...I was going to omit the part where I mistakenly threw the 1500X back in after flashing the 550 board but I digress (that did 100% happen)

I'm up and running now though; since I was planning to re-seat the 3700X anyway because temps seemed a touch too high--but that one certainly beat me up real nice.

/tired

LOL! I was a little nervous myself about that, but my manual claimed that it supported the "4000" series so I went with it -- and it worked. I still wound up stealing parts from two different machines though, so I am up one 5800x and down one 3600x and one 2700x.

/lucky

 

[DrMrLordX]

I do fine-tune my memory timings, but I usually start from the XMP profile. I didn't realize that XMP might affect other voltages though. I have new memory coming in a few days and will attack the mem timings with an eye on voltages. After I get the memory and fabric settled, I'll be looking into voltage offset possibilities. The board seems to have defaulted to an LLC setting of '2' -- and may have a few other unexpected (to me, at least) default settings.

You can copy the XMP timings/subtimings if you like (as a baseline), but actually activating the XMP profile may be a bad idea. Some folks report no problems, while others have difficulties (such as those you noticed).

 

[ArisVer]

Some news for the 5000 family.

Two Zen 3 CPUs — Ryzen 9 5900 and 7 5800 — with 65 W TDP headed for OEMs

Not confirmed by AMD yet, the Ryzen 9 5900 and Ryzen 7 5800 processors are apparently headed for OEMs with 12 and 8 cores and twice as many threads, as well as 6/64 MB and 4/32 MB of L2/L3 cache. Both will feature a TDP value of just 65 W, but their names and clock speeds are not final yet.

www.notebookcheck.net

 

[KentState]

Obligatory picture of build. Waiting on EK to ship the GPU water block but while I wait, the 5950X and VRM have a lot of radiator (EK 360 XE and 480PE) to cool them. Will have to read through all the overlocking discussions again and see what I can get out of it.

Since I just swapped the CPU from a 3900XT to this, should I reinstall the chipset driver?

 

[Justinus]

Obligatory picture of build. Waiting on EK to ship the GPU water block but while I wait, the 5950X and VRM have a lot of radiator (EK 360 XE and 480PE) to cool them. Will have to read through all the overlocking discussions again and see what I can get out of it.

View attachment 36483

If you play with all core overclocking I'm interested to see where you get.

I've got 3x hwlabs LX 360mm rads and an Optimus block on my 5950x right now pending GPU, and best overclock I'm satisfied with at this point is 4.675GHz on both CCDs at 1.251v. 4.7 requires 1.3v and even with the Optimus block and massive radiator space it peaks just over 80C which is too hot for my taste.

 

[KentState]

If you play with all core overclocking I'm interested to see where you get.

I've got 3x hwlabs LX 360mm rads and an Optimus block on my 5950x right now pending GPU, and best overclock I'm satisfied with at this point is 4.675GHz on both CCDs at 1.251v. 4.7 requires 1.3v and even with the Optimus block and massive radiator space it peaks just over 80C which is too hot for my taste.

Maybe I'm confused, but I'm showing 4.65 all core at 1.42V and around 49C in Ryzen Master. This is with Prime95 using the In-place Large FFT test. The Blended test doesn't seem to produce as much heat. I would think that you would be cooler than me at a lower voltage.

 

[Justinus]

Maybe I'm confused, but I'm showing 4.65 all core at 1.42V and around 49C in Ryzen Master. This is with Prime95 using the In-place Large FFT test. The Blended test doesn't seem to produce as much heat. I would think that you would be cooler than me at a lower voltage.

What are you scoring in synthetics? Like CB R23? I noticed when I briefly experimented with PBO I was not scoring what I should have at the given frequency, nor was it running as hot as it should have.

I haven't ruled out something being wrong with my loop but I've switched coolants and disassembled my CPU block to verify its clean, unclogged, and flowing as expected.


I checked Prime95 with in place large FFTs and it doesn't hold a candle to the heat generation or power consumption of running cinebench R23. It's only hitting ~185W and 55C whereas my 8 hour cinebench r23 loop hits 260W and 74C.

 

[gk1951]

Heavens KentState that is a beautiful rig!

I just received today a 5900x to replace my 3900x in my custom water cooled rig (MSI X570 Unify, Optimus AMD Foundation cpu wb, 32gigs (2x16) Gskill DDR4-3600 ram, Rad VII under an EK wb.)

I have it housed in a Fractal Define 7XL case. Loop cooled by a Corsair 480 7XL rad and a Magicool 360 rad. I'm using a EK D5-140 pump/res combo.

Preliminary benches show the 5900x to be fast. Your 5950x should be a BRUTE!

Now to find a 6800XT to replace my Rad VII with. UGH!!!!!!!!!

 

[amrnuke]

Out of curiosity, I have seen a lot of people on the interwebs mention their 5800X, 5900X, 5950X running at 90-95C at times.

My 5600X will not exceed 82C, even on PBO. This is curious to me... I'm hitting 4.575 GHz all-core in CB23 (4.65 GHz peak boost), and very periodically that will drop to 4.5 if temps hit 82C, then once temps get down to 80-81C, the chip goes back up to 4.575 and 4.6, dropping at times to 4.575 again, and so on. I opened my case and blew a little Honeywell 12" fan on high into it, and it holds 4.6-4.625 all-core, again throttling down at 82C. So it seems temps have something to do with it, but why are my temps capped at 82C?

Is the limiter here wattage (115W per HWInfo), amps, V, or something else?

Don't get me wrong, I'm tickled that it holds all-core speeds equal to the advertised max boost clock, but it seems like it has more to give...

 

[DrMrLordX]

@amrnuke

Just speculation, but temps do affect voltage/clockspeed curves for Matisse pretty severely, so Vermeer might behave the same way. I am reminded of an early 3900X review by der8auer:

It's not entirely applicable since he's dealing with static OC, but you can see that there is an inverse correlation between achieveable clockspeed and temperature. The boost algo in Matisse can adjust itself just based on temperature (I saw some early 3900X reviews pick up clocks just by dunking a waterloop's rad in icewater while running defaults). According to der8auer, -20C = +50-100MHz for Matisse. For Vermeer? No idea. He hasn't released an analogous video for a 5900X.

 

[amrnuke]

@amrnuke

Just speculation, but temps do affect voltage/clockspeed curves for Matisse pretty severely, so Vermeer might behave the same way. I am reminded of an early 3900X review by der8auer:

It's not entirely applicable since he's dealing with static OC, but you can see that there is an inverse correlation between achieveable clockspeed and temperature. The boost algo in Matisse can adjust itself just based on temperature (I saw some early 3900X reviews pick up clocks just by dunking a waterloop's rad in icewater while running defaults). According to der8auer, -20C = +50-100MHz for Matisse. For Vermeer? No idea. He hasn't released an analogous video for a 5900X.

Interesting... I'm about to find out I guess! I bought an Arctic Freezer II 240mm AIO for $99 a couple of days ago, so we will see how it does.

(I "needed" another cooler as I turn my 3600+CoolerMaster Hyper 212 into a headless dedicated Rosetta@Home/WCG and Plex/home file server... since it'll be running 24/7 DC loads I don't want to use the base Stealth cooler. I also figured, what the heck, and bought some G.skill Trident Z 4000CL18 B-die RAM. Eventually I'll find a reason to buy a new motherboard and case and power supply!)

 

[Midwayman]

Out of curiosity, I have seen a lot of people on the interwebs mention their 5800X, 5900X, 5950X running at 90-95C at times.

That seems really high unless they are measuring during a stress test. I don't think mine has made it out of the 70s.

 

[Space Tyrant]

That seems really high unless they are measuring during a stress test. I don't think mine has made it out of the 70s.

I probably have the worst cooler on this forum on my 5800x and the highest (stock) temp I've seen is 90C while running y-cruncher (1B digits of pi mode). The HSF is an original (old style) Wraith Spire with a 120mm fan+adapter replacing the stock 92mm fan. One other very stressful benchmark I ran also topped out at 90C.

I'll replace the spire with my dark rock 4 shortly and I expect to see lower temps.

Outside those math-intensive benchmarks, nothing else I run produces particularly high temps.

 

[Justinus]

I think I figured out why my temps are so high attempting static overclocks on my 5950x - turns out optimus forgot to install the internal jet plate oring, so presumably the water is escaping over the top of the fin channels instead of being forced through them.

They're sending me some orings so I can fix it. No idea how much temps are affected but my intuition is maybe 5-10C.