AMD Ryzen 5000 Builders Thread

Page 62 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

B-Riz

Golden Member
Feb 15, 2011
1,482
612
136
Last edited:

Asterox

Golden Member
May 15, 2012
1,026
1,775
136
Well it is AM4 socket, so potential builders they have no right to object. :mask:


 
  • Like
Reactions: lightmanek

Det0x

Golden Member
Sep 11, 2014
1,027
2,953
136
Seems like Santa came a few days early this year 🎅
unify x max.jpg
Had to order from Germany to find it in stock anywhere..

Will try to find the time to get it installed within a few days.. Anyone taking bets if it will let me run above 1900 fclk without reduced performance ?
(Crosshair VIII Hero currently limits me to 1900, and maximum 2033MT/s memspeed when decoupled from flck)
 

Ajay

Lifer
Jan 8, 2001
15,332
7,792
136
Seems like Santa came a few days early this year 🎅
View attachment 54656
Had to order from Germany to find it in stock anywhere..

Will try to find the time to get it installed within a few days.. Anyone taking bets if it will let me run above 1900 fclk without reduced performance ?
(Crosshair VIII Hero currently limits me to 1900, and maximum 2033MT/s memspeed when decoupled from flck)
Nice! Except, wait, is it brown?
 

Det0x

Golden Member
Sep 11, 2014
1,027
2,953
136
@Det0x

There's a lot to unpack in that picture. Gummies and cognac? I don't even know mang. And is that gin?

(enjoy the board, lol)
Just go with the flow :p
In the meantime, i'm trying something new!
1.jpg
It cant get worse, i hope atleast :p
Stock heatsink on Trident Z are brushed aluminum, so i thought what the hell, lets just overdo it on the outside to try to compensate for little spacing between the sticks on this 2dimm board :cool:
2.jpg3.jpg4.jpg5.jpg6.jpg
 
  • Wow
  • Love
Reactions: lightmanek and Ajay

DrMrLordX

Lifer
Apr 27, 2000
21,583
10,785
136
@Det0x

Pimp my RAM edition? Interesting nevertheless. Not sure that heat will really travel well with multiple thermal interfaces like that, but . . . should be interesting at least.
 

Det0x

Golden Member
Sep 11, 2014
1,027
2,953
136
Just go with the flow :p
In the meantime, i'm trying something new!
View attachment 54670
It cant get worse, i hope atleast :p
Stock heatsink on Trident Z are brushed aluminum, so i thought what the hell, lets just overdo it on the outside to try to compensate for little spacing between the sticks on this 2dimm board :cool:
View attachment 54671View attachment 54672View attachment 54673View attachment 54674View attachment 54675
Preliminary report :)

Extra heatsinks on memory sticks seems to be working fine, eventho the memory are so close together and the heatsinks are just on one side.. Only a ~7 degree delta between idle and load after 1 hour Karhu ramtest @ 1.56vdimm
1640043500604.png
Motherboard is working really exemplary, it boot every setting known to man, no fuss, no nothing... Haven't had a failed post yet (y) (10x better then the Asus board in this regard)
And as you can see, I'm also back to my normal "CL13 1900 baseline memory settings"

CPU is the same 5950x as before, everything above 1900 flck spits out ~100 WHEA each Aida64 mem testrun
But one thing is new with this motherboard.. with the Crosshair VIII Hero i got clearly reduced performance @ 1933 and everything above, even in superlight things like Aida.. With this board i seems to see somewhat scaling to atleast 4000:2000, maybe :p (haven't tried higher yet)

Hopefully this means there is hope for real performance scaling above 1900 when using the whea suppressor on this board.. :)

Much testing left to be done and lots of new settings to be learned, but i'm happy with what this board have achieved on first day.. This will be a very good platform for Zen 3d-$ in the future
 

Det0x

Golden Member
Sep 11, 2014
1,027
2,953
136
So i have done some further testing..

The good:
This motherboard + memory combo have no problems booting up (atleast) 4400MT/sec memory speeds on dual rank 2x16GB running pretty tight timings (flat CL15).
Really wish for a good sample of 6950x v-$ which let me run high fclk🙏

Motherboard boots pretty much every settings i try, and when it don't, the recovery-system works flawlessly :)

The bad:
I seem to get lower performance with this motherboard @ the same static clocks compared to my old asus motherboard.

If we take Cinebench r20 which don't care about memory for example:
Crosshair VIII Hero @ 4700/4600 static clocks = ~12.1k points in r20
Unify X MAX @ 4700/4600 static clocks = ~11.9k points in r20

Hopefully this comes down to the rather old AGESA on the only bios released for this motherboard atm, SMU 56.52.0 (1.2.0.3c)
Crossing my fingers this get fixed in new bios..

Some funkiness: (LLC lol scaling)
LLC 8 (weakest) = upto ~20% cpu vdroop
LLC 7 (second weakest) = ~3% cpu vdroop

My CPU is still the same 5950x, cant run above 1900fclk without reduced performance, and have pretty much given up trying at this point..

The interesting:
1640105527660.png VS 1640105536271.png

When testing how high memory-speed this motherboard + memory combo would run decoupled from flck with the minimum fiddling in the bios, (second day i have this motherboard) i also decided to perform some basic benchmark comparisons @ a static 4500/4400mhz cpu clock against my standard synced flck 1900 CL profile..

Since the IO die on my 5950x is so bad limiting me to lowish 1900:1900:3800 when running synced, it really depend on what the benchmark I'm running for it to be faster than running 1600:1100:4400 asynced..

1640105551157.png
VS
1640105561902.png

1600:1100:4400 asynced
Linpack = 659 average gflops
y-cruncher = 60.990 seconds

1900:1900:3800 synced
Linpack = 661 average gflops
y-cruncher = 63.711 seconds

Even more testing needed.. :p
 
Last edited:

Det0x

Golden Member
Sep 11, 2014
1,027
2,953
136
Pretty bored since no 5950x3d.. Playing around with async mode some more for a lack of other things to tweak.

This time its my single rank 2x8GB sticks that have to take the abuse at 1.63 vdimm :smilingimp:
1641538240013.png

And some Y-cruncher numbers @ CL17 4667MT/s at pure T1
Rank #1 on hwbot for 2.5B Zen3
1641538429698.png


Rank #1 on hwbot for 1B Zen3
1641538453151.png

Do note the 343watt PPT above :innocent:

Have 3 new 5950x's incoming next week, crossing my fingers for atleast one with the B2 stepping...
No 5950x3d ruined all my plans since i already had made a deal with my brother to sell him my current 5950x+crosshair viii hero+these 2x8gb mem sticks

Dual rank 2x16GB maxes out at 4400 MT/s (100% stable and bench-able)
Single rank 2x8GB maxed out at 4800 MT/s (need to lower to 4667 for benchmarks)

*edit*
Did also swap the fans on ramcooler :)
1641539526194.png1641539535545.png
 
Last edited:
  • Love
Reactions: lightmanek

deasd

Senior member
Dec 31, 2013
513
724
136
Since Zen3 thread is too old I decide to post this in builders thread, B2 stepping Zen3 already on shelve and you might buy it with some luck:

the owner said it SEEMS B2 has better OC headroom.

da400fb30f2442a73d39acbc8c43ad4bd1130281.jpgd31b304e251f95ca64e3c4af94177f3e670952a2.jpg
 
  • Like
Reactions: lightmanek

Justinbaileyman

Golden Member
Aug 17, 2013
1,980
249
106
So my New X570s Aorus Master arrived today from Newegg. I am all set for a total upgrade and plan to get a 5950x or 5950x3D to replace my 3950x if ever released. But for the meantime what should I go for next as far as an upgrade goes? Maybe a new GPU ?? If so what is the go to GPU these days ? I have both a 5700xt and a 2070 super which are probably way outdated by now so would like to get something a little more up to date?
 

lightmanek

Senior member
Feb 19, 2017
387
754
136
I've ordered 5950X to upgrade my 5900X since Zen3D launches later than I expected and with only 8 cores. I'm hoping to get B2 stepping, but regardless of that, this should be a small but nice upgrade to my rig (plus first NVME PCie4.0 Samsung 980Pro going as a system drive, replacing 970EVO).
 

lightmanek

Senior member
Feb 19, 2017
387
754
136
Does the 5950x have a weaker memory controller then the 3950x??
Edit: How do you tell what stepping you have??

Ryzen 5xxx has identical memory controller as they use the same IOD. In reality though, Ryzen 5000 series processors tend to clock memory and fabric slightly higher than older products.
Stepping can be decoded from the lid of processor if you know what to look for or by using tools like CPU-Z or HwInfo.
 

Justinbaileyman

Golden Member
Aug 17, 2013
1,980
249
106
Ryzen 5xxx has identical memory controller as they use the same IOD. In reality though, Ryzen 5000 series processors tend to clock memory and fabric slightly higher than older products.
Stepping can be decoded from the lid of processor if you know what to look for or by using tools like CPU-Z or HwInfo.
Okay , thanks for your reply..!!
 
  • Like
Reactions: lightmanek

scineram

Senior member
Nov 1, 2020
361
283
106
5700G, Fatal1ty B450 Gaming-ITX

Guys, when I enable Global C-states there is video output corruption after logging into Windows and at idle. But during a benchmark run the artifacts go away. Has anybody seen this before? What else should I change in the BIOS? Everythings is up to date.
Guys, Breakthrough!

I noticed that I had all modern BIOS features enabled, except IOMMU. When I turned that to Enabled or Auto the issue went away. Now there are no graphics artifacts even with Global C-states Control on. Not sure what the connection was between those two, but there it is.
 

Det0x

Golden Member
Sep 11, 2014
1,027
2,953
136
Have received the last two cpus and done preliminary silicon quality tests on these also, both were a older batch from week 30 2021.
First one was decent but nothing special, last one was a pretty good platinum/almost platinum sample i'm very happy with :)
Best sample can run 3x Cinebench r20 runs back to back @ 4825/4700mhz with only 1237mv under load, in normal ambient temp.. Together with high ST boosting on the correct cores this is a keeper. (core1)
Sadly i still cant make any headway in Linpack Xtreme in regards to scaling above 1900flck with these 4 dual CCD cpus ive tested, but i'm seeing close to perfect scaling in Aida and other light applications... Also don't get any WHEA in windows before i starts benchmarks when above 1900:3800.. (~around 100 errors per linpack run @ 2000:4000)

All tests below was run at stock bios settings --> PBO disabled and "auto" LLC which is pretty weak on this MSI UNIFY X MAX motherboard. (upto 10% vdroop under heavy load)
All stock diagnostic voltages in hydra.
Asus tool = "116 silicon quality" and "4381mhz max heavy load"

Hydra: CCD quality
CCD #1 = 4525mhz --> 4.05 rating = Golden CCD
CCD #1 = 4325mhz --> 3.87 rating = Bronze CCD

1T-2T:
CCD #1 = AVX1 = 4775mhz / FMA3 = 4650mhz
CCD #2 = AVX1 = 4725mhz / FMA3 = 4600mhz

3T-4T:
CCD #1 = AVX1 = 4750mhz / FMA3 = 4625mhz
CCD #2 = AVX1 = 4700mhz / FMA3 = 4575mhz

9T-12T:
CCD #1 = AVX1 = 4750mhz / FMA3 = 4625mhz
CCD #2 = AVX1 = 4575mhz / FMA3 = 4450mhz

13T-16T:
CCD #1 = AVX1 = 4675mhz / FMA3 = 4550mhz
CCD #2 = AVX1 = 4500mhz / FMA3 = 4375mhz

All threads:
CCD #1 = AVX1 = 4475mhz / AVX2 = 4400mhz / FMA3 = 4125mhz
CCD #2 = AVX1 = 4325mhz / AVX2 = 4250mhz / FMA3 = 3975mhz

Game profile:
CCD #1 = 4725mhz
CCD #2 = 4550mhz

Highest ST clockspeed for 4 highest rated cores: (1375mv set)
#1 = core1 @ 4925mhz (212 rating)
#2 = core3 @ 4900mhz (208 rating)
#3 = core5 @ 4875mhz (212 rating)
#4 = core7 @ 4875mhz (203 rating)

Highest clockspeed possible to complete Cinebench r20 3 times in a row @ 1375mv set (upto 10% vdroop)
CCD #1 = 4700mhz
CCD #2 = 4600mhz

Linpack Xtreme @ static 4500/4400mhz (slow "safe" memory-timings)
1900:3800 = 642.98 Gflops
1933:3866 = ~400 Glops (whea's and massive throttling)
2000:4000 = not even attempted

Asus tool = "117.5 silicon quality" and "4394mhz max heavy load"

Hydra: CCD quality
CCD #1 = 4575mhz --> 4.11 rating = Platinum CCD
CCD #1 = 4475mhz --> 4.02 rating = Golden CCD

1T-2T:
CCD #1 = AVX1 = 4750mhz / FMA3 = 4625mhz
CCD #2 = AVX1 = 4700mhz / FMA3 = 4575mhz

3T-4T:
CCD #1 = AVX1 = 4725mhz / FMA3 = 4600mhz
CCD #2 = AVX1 = 4675mhz / FMA3 = 4550mhz

9T-12T:
CCD #1 = AVX1 = 4775mhz / FMA3 = 4650mhz
CCD #2 = AVX1 = 4650mhz / FMA3 = 4525mhz

13T-16T:
CCD #1 = AVX1 = 4700mhz / FMA3 = 4575mhz
CCD #2 = AVX1 = 4575mhz / FMA3 = 4450mhz

All threads:
CCD #1 = AVX1 = 4525mhz / AVX2 = 4450mhz / FMA3 = 4175mhz
CCD #2 = AVX1 = 4450mhz / AVX2 = 4375mhz / FMA3 = 4100mhz

Game profile:
CCD #1 = 4750mhz
CCD #2 = 4600mhz

Highest ST clockspeed for 4 highest rated cores: (1375mv set)
#1 = core3 @ 4975mhz (208 rating)
#2 = core7 @ 4950mhz (212 rating)
#3 = core8 @ 4875mhz (212 rating)
#4 = core1 @ 4850mhz (203 rating)

Highest clockspeed possible to complete Cinebench r20 3 times in a row @ 1375mv set (upto 10% vdroop)
CCD #1 = 4800mhz
CCD #2 = 4650mhz

Linpack Xtreme @ static 4500/4400mhz (slow "safe" memory-timings)
1900:3800 = 643.93 Gflops
1933:3866 = ~616.5 Glops (some whea's)
2000:4000 = 199 Gflops (whea's and massive throttling)
  • Main drawback of CPU above: Best 212 cores on this #1 5950x are core 7 and 8. This is bad for a few reasons.. The two last cores on a full CCD always have around 6 degrees higher temps. Some benchmarks/application always ignore CPPC (such as CPU-Z) and forces workload to core1, this is very bad since core1 is the worst one in entire CCD on this cpu. (you always wish for your best 212 cores to be core1)
1642746714650.png
Asus tool = "116.5 silicon quality" and "4386mhz max heavy load"

Hydra: CCD quality
CCD #1 = 4575mhz --> 4.09 rating = Golden CCD
CCD #1 = 4475mhz --> 4 rating = Golden CCD

1T-2T:
CCD #1 = AVX1 = 4725mhz / FMA3 = 4600mhz
CCD #2 = AVX1 = 4725mhz / FMA3 = 4600mhz

3T-4T:
CCD #1 = AVX1 = 4700mhz / FMA3 = 4575mhz
CCD #2 = AVX1 = 4700mhz / FMA3 = 4575mhz

9T-12T:
CCD #1 = AVX1 = 4775mhz / FMA3 = 4650mhz
CCD #2 = AVX1 = 4650mhz / FMA3 = 4525mhz

13T-16T:
CCD #1 = AVX1 = 4700mhz / FMA3 = 4575mhz
CCD #2 = AVX1 = 4575mhz / FMA3 = 4450mhz

All threads:
CCD #1 = AVX1 = 4525mhz / AVX2 = 4450mhz / FMA3 = 4175mhz
CCD #2 = AVX1 = 4450mhz / AVX2 = 4375mhz / FMA3 = 4125mhz

Game profile:
CCD #1 = 4725mhz
CCD #2 = 4600mhz

Highest ST clockspeed for 4 highest rated cores: (1375mv set)
#1 = core3 @ 4925mhz (212 rating)
#2 = core5 @ 4900mhz (208 rating)
#3 = core2 @ 4875mhz (203 rating)
#4 = core1 @ 4875mhz (212 rating)

Highest clockspeed possible to complete Cinebench r20 3 times in a row @ 1375mv set = 1237mv under load
CCD #1 = 4800mhz
CCD #2 = 4650mhz

Linpack Xtreme @ static 4500/4400mhz (slow "safe" memory-timings)
1900:3800 = 643flops
1933:3866 = ~624 Glops
2000:4000 = 387 Gflops
"Perfect" scaling in aida64

CPU below will serves as my new 5950x
Asus tool = "116 silicon quality" and "4382mhz max heavy load"

Hydra: CCD quality
CCD #1 = 4625mhz --> 4.14 rating = Platinum CCD
CCD #1 = 4500mhz --> 4.03 rating = Golden CCD

1T-2T:
CCD #1 = AVX1 = 4825mhz / FMA3 = 4700mhz
CCD #2 = AVX1 = 4700mhz / FMA3 = 4575mhz

3T-4T:
CCD #1 = AVX1 = 4800mhz / FMA3 = 4675mhz
CCD #2 = AVX1 = 4675mhz / FMA3 = 4550mhz

9T-12T:
CCD #1 = AVX1 = 4825mhz / FMA3 = 4700mhz
CCD #2 = AVX1 = 4650mhz / FMA3 = 4525mhz

13T-16T:
CCD #1 = AVX1 = 4750mhz / FMA3 = 4625mhz
CCD #2 = AVX1 = 4575mhz / FMA3 = 4450mhz

All threads:
CCD #1 = AVX1 = 4575mhz / AVX2 = 4500mhz / FMA3 = 4200mhz
CCD #2 = AVX1 = 4450mhz / AVX2 = 4375mhz / FMA3 = 4150mhz

Game profile:
CCD #1 = 4775mhz
CCD #2 = 4625mhz

Highest ST clockspeed for 4 highest rated cores: (1375mv set)
#1 = core5 @ 4950mhz (208 rating)
#2 = core3 @ 4950mhz (203 rating)
#3 = core8 @ 4925mhz (212 rating)
#4 = core1 @ 4925mhz (212 rating)

Highest clockspeed possible to complete Cinebench r20 3 times in a row @ 1375mv set = 1237mv under load
CCD #1 = 4825mhz
CCD #2 = 4700mhz

Linpack Xtreme @ static 4500/4400mhz (slow "safe" memory-timings)
1900:3800 = 643.93 Gflops
1933:3866 = ~560 Glops
2000:4000 = 375 Gflops
"Perfect" scaling in aida64

Full Hydra diagnostic with elevated voltages on my new best 5950x
1642747035542.png
1642747050787.png
Have to admit, think this is among the best 5950x's ive ever seen 😁
 

Attachments

  • 1642746705775.png
    1642746705775.png
    63 KB · Views: 3
Last edited:

Det0x

Golden Member
Sep 11, 2014
1,027
2,953
136
When trying to extract the maximum performance = 100.25mhz baseclock lol
1642747212368.png
Memory settings stability tested:
1642747234593.png
200%RunmemtestPro = ~1.5 hour
25 cycles TestMEM5 1usmus cfg = ~2.5 hours
Y-cruncher, all tests = 6 hours
20000% karhu ram test = ~10.5 hours
 

Det0x

Golden Member
Sep 11, 2014
1,027
2,953
136
First bench session with new 5950x:

TechN AM4 block with liquid metal on custom watercooling = 60 degrees under load when running Cinebench r23 MT bench

Cinebench R23 MT Score = 33712 points
Cinebench R23 ST Score = 1726 points

Cinebench R20 MT Score = 13005 points
Cinebench R23 ST Score = 671 points

Cinebench R15 MT Score = 5638 points
Cinebench R15 ST Score = 287 points
1642747396518.png

Little CPU-Z run for fun
1642747534432.png
 
Last edited:

lightmanek

Senior member
Feb 19, 2017
387
754
136
First bench session with new 5950x:

TechN AM4 block with liquid metal on custom watercooling

Cinebench R23 MT Score = 33712 points
Cinebench R23 ST Score = 1726 points

Cinebench R20 MT Score = 13005 points
Cinebench R23 ST Score = 671 points

Cinebench R15 MT Score = 5638 points
Cinebench R15 ST Score = 287 points

****

What are your all core clocks for these runs?
My 5950x is a very efficient CPU at lower freq. but has hard time to run 4.7GHz all core as one CCD says NO.
Also my IF clocks cannot go giher than 1800MHz which is unheard of! My 5900x could boot at 2000 but was not stable, this one hasn't booted even once at 1900MHz and I've tried few times already.
When I get a chance I will post few scores, but I think I got an average sample this time ... well, at least the faster CCD has no problem boosting to 5.15GHz, haven't tried more yet :)
 

Det0x

Golden Member
Sep 11, 2014
1,027
2,953
136
What are your all core clocks for these runs?
My 5950x is a very efficient CPU at lower freq. but has hard time to run 4.7GHz all core as one CCD says NO.
Also my IF clocks cannot go giher than 1800MHz which is unheard of! My 5900x could boot at 2000 but was not stable, this one hasn't booted even once at 1900MHz and I've tried few times already.
When I get a chance I will post few scores, but I think I got an average sample this time ... well, at least the faster CCD has no problem boosting to 5.15GHz, haven't tried more yet :)
Benchmate info-snip with clocks :)
1642757245643.png