• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

What is your max infinity fabric on ryzen 5000? September 2021 edition

What is your max stable ram speed/infinity fabric?

  • 3666/1833 or lower

    Votes: 29 17.1%
  • 3733/1866

    Votes: 32 18.8%
  • 3800/1900

    Votes: 61 35.9%
  • 3866/1933

    Votes: 7 4.1%
  • 3933/1966

    Votes: 4 2.4%
  • 4000/2000

    Votes: 25 14.7%
  • 4066/2033

    Votes: 2 1.2%
  • 4133/2066

    Votes: 3 1.8%
  • 4200/2100 or above

    Votes: 7 4.1%

  • Total voters
    170
Joined
Sep 27, 2021
Messages
157 (0.15/day)
Location
Portugal
Processor Ryzen 7 5800x
Motherboard Msi b450 tomahawk max II
Cooling Artic liquid freezer II 240mm
Memory 2x8Gb B-die @3800 cl14/ 53ns on aida
Video Card(s) Evga 3060 ti ftw3
Storage M.2 Samsung 970 Evo Plus 250gb + 1tb ssd
Case Corsair 275r airflow
Audio Device(s) Hyperx
Power Supply Seasonic core gc 650w 80 plus gold
Mouse Zowie ec2-a
Keyboard Trust
Software Windows 10
Benchmark Scores Cinebench r20: 631/6232 Cinebench r23: 1628/15901 CPU-Z: 672/6960 SotTR:221FPS
It's what I got. I boot at 4000 and 3933 but got whea errors on occt memory test. At least I got 3866 cl16 stable later I post screenshots of Aida an timinings to got some advices.
 

Dux

Joined
May 17, 2016
Messages
511 (0.17/day)
Still if you tune your curve optimizer max it will generally beat allcore OC.

Here is my scores:
4.4@1.05V 60-65W load
4.5@1.09V 65-70W load
4.6@1.15V 75-80W load
4.7@1.23V 90-100W load
4.8@1.32V 115-125W+ load

Curve optimizer +200 pbo and -30x4/-29x2
4.85@1.34V single core
4.57-4.63@1.16-1.17V@76W

In idle temps with PBO+CO is 40-45C and voltage 0.83-0.85V.

The most comparable 4.6GHz is 45-50W idle and 1.15V.

As you see CO+PBO is superior to allcore OC since you get better single core at lower temp/consumption and very low temp/consumption in idle.
hey, can someone please explain this Curve optimizer to me? I had 5900X but sold it today and got 5600X because I had no more need for high core count. I have no idea how to do this PBO thing and CO. I have X570 mbo. Any tutorial somewhere?

EDIT:
OK so to answer my own question. I did find a yt video. And i found this thing in bIOS. description in BIOS says that setting the value to negative and then number below that will lower the voltage of CPU which it uses for certain frequency and the other way around. However after testing i see it is completely different for me. Setting the value to negative and numberto 20, my CPU became very hot when testing with cinebench R23 and package power consumption vent to 115W and voltage over 1.3V all core. . When i set the value to positive and number to 20. The power consumption, max voltage and temperatures went down.

So it turns out it is completely opposite of description in BIOS.
 
Last edited by a moderator:
Joined
Nov 5, 2021
Messages
40 (0.04/day)
pass 2.png

did a little more tweaking @Taraquin , got it down to the 59ns range on aida and in the 15800 range on r23. ill start tweaking those other timings soon
 
Joined
Aug 9, 2019
Messages
1,572 (0.87/day)
Processor 7800X3D 2x16GB CO
Motherboard Asrock B650m HDV
Cooling Peerless Assassin SE
Memory 2x16GB DR A-die@6000c30 tuned
Video Card(s) Asus 4070 dual OC 2610@915mv
Storage WD blue 1TB nvme
Display(s) Lenovo G24-10 144Hz
Case Corsair D4000 Airflow
Power Supply EVGA GQ 650W
Software Windows 10 home 64
Benchmark Scores Superposition 8k 5267 Aida64 58.5ns
hey, can someone please explain this Curve optimizer to me? I had 5900X but sold it today and got 5600X because I had no more need for high core count. I have no idea how to do this PBO thing and CO. I have X570 mbo. Any tutorial somewhere?

EDIT:
OK so to answer my own question. I did find a yt video. And i found this thing in bIOS. description in BIOS says that setting the value to negative and then number below that will lower the voltage of CPU which it uses for certain frequency and the other way around. However after testing i see it is completely different for me. Setting the value to negative and numberto 20, my CPU became very hot when testing with cinebench R23 and package power consumption vent to 115W and voltage over 1.3V all core. . When i set the value to positive and number to 20. The power consumption, max voltage and temperatures went down.

So it turns out it is completely opposite of description in BIOS.
Don't touch scalar. I would stay away from + pbo. Remember that curve optimizer does not improve temps or consumption. It only does force the CPU to use less voltage at a specific frequency. Example: My 5600X stock uses 76W in CB R23. Without curve optimizer clockspeed is 4.275-4.325GHz avg at 1.23-1.25V. With negative -30 allcore it still uses 76W, but clockspeed is now 4.55-4.6GHz at 1.2-1.21V. Performance is 6% better. If you want to improve temps set a PPT limit. On my setup I 65W limit with CO and 76W limit stock makes clocks allcore run at 4.3GHz avg allcore. This means that - 30 CO allcore lowers consumption by 15% if clocks are kept equal. Temps about 5C lower.
 
  • Like
Reactions: Dux

Dux

Joined
May 17, 2016
Messages
511 (0.17/day)
Don't touch scalar. I would stay away from + pbo. Remember that curve optimizer does not improve temps or consumption. It only does force the CPU to use less voltage at a specific frequency. Example: My 5600X stock uses 76W in CB R23. Without curve optimizer clockspeed is 4.275-4.325GHz avg at 1.23-1.25V. With negative -30 allcore it still uses 76W, but clockspeed is now 4.55-4.6GHz at 1.2-1.21V. Performance is 6% better. If you want to improve temps set a PPT limit. On my setup I 65W limit with CO and 76W limit stock makes clocks allcore run at 4.3GHz avg allcore. This means that - 30 CO allcore lowers consumption by 15% if clocks are kept equal. Temps about 5C lower.
Hey, that's fantastic. I just set it to negative 25 and PPT to 75. I use aida64 to monitor voltages, consumption, temperatures and clocks while running cinebench R23 on the second monitor. CPU did respect that 75W PPT limit and clocks went to 4450MHz all cores when running Cinebench. CPU used 1.140V for that frequency.

Also I'm curious about Cinebench R23 scores. I read that people get around 11500 points for multi thread test. But i get around 10500 when everything in BIOS is on default. Almost 11000 when i tune this curve optimizer.
 
Joined
Aug 9, 2019
Messages
1,572 (0.87/day)
Processor 7800X3D 2x16GB CO
Motherboard Asrock B650m HDV
Cooling Peerless Assassin SE
Memory 2x16GB DR A-die@6000c30 tuned
Video Card(s) Asus 4070 dual OC 2610@915mv
Storage WD blue 1TB nvme
Display(s) Lenovo G24-10 144Hz
Case Corsair D4000 Airflow
Power Supply EVGA GQ 650W
Software Windows 10 home 64
Benchmark Scores Superposition 8k 5267 Aida64 58.5ns
Hey, that's fantastic. I just set it to negative 25 and PPT to 75. I use aida64 to monitor voltages, consumption, temperatures and clocks while running cinebench R23 on the second monitor. CPU did respect that 75W PPT limit and clocks went to 4450MHz all cores when running Cinebench. CPU used 1.140V for that frequency.

Also I'm curious about Cinebench R23 scores. I read that people get around 11500 points for multi thread test. But i get around 10500 when everything in BIOS is on default. Almost 11000 when i tune this curve optimizer.
Try -30 on all if you can or use corecycler to find which cores can't do -30.
 
  • Like
Reactions: Dux

Dux

Joined
May 17, 2016
Messages
511 (0.17/day)
Try -30 on all if you can or use corecycler to find which cores can't do -30.
OK. I did try -30. It did nothing for the maximum frequency in Cinebench R23. It is still maxing out at 4450MHz. This is because CPU is respecting that 75W PPT limit i set, i guess. And voltage is also 1.140V for that frequency. What did change however is the minimum voltage when CPU is idle and working at 3.7GHz. It is now down to 0.950V. And this lowered idle consumption and temperatures. So this really is fantastic.
Thank you again. :toast:
 
Joined
Aug 9, 2019
Messages
1,572 (0.87/day)
Processor 7800X3D 2x16GB CO
Motherboard Asrock B650m HDV
Cooling Peerless Assassin SE
Memory 2x16GB DR A-die@6000c30 tuned
Video Card(s) Asus 4070 dual OC 2610@915mv
Storage WD blue 1TB nvme
Display(s) Lenovo G24-10 144Hz
Case Corsair D4000 Airflow
Power Supply EVGA GQ 650W
Software Windows 10 home 64
Benchmark Scores Superposition 8k 5267 Aida64 58.5ns
OK. I did try -30. It did nothing for the maximum frequency in Cinebench R23. It is still maxing out at 4450MHz. This is because CPU is respecting that 75W PPT limit i set, i guess. And voltage is also 1.140V for that frequency. What did change however is the minimum voltage when CPU is idle and working at 3.7GHz. It is now down to 0.950V. And this lowered idle consumption and temperatures. So this really is fantastic.
Thank you again. :toast:
Seems your motherboard handles your CPU very different than mine. Mine idles at 0.83-0.85V at 3.7GHz with all cores at -30. May change if you change bios etc, AMD chipset drivers might affect it too.
 
Joined
Sep 27, 2021
Messages
157 (0.15/day)
Location
Portugal
Processor Ryzen 7 5800x
Motherboard Msi b450 tomahawk max II
Cooling Artic liquid freezer II 240mm
Memory 2x8Gb B-die @3800 cl14/ 53ns on aida
Video Card(s) Evga 3060 ti ftw3
Storage M.2 Samsung 970 Evo Plus 250gb + 1tb ssd
Case Corsair 275r airflow
Audio Device(s) Hyperx
Power Supply Seasonic core gc 650w 80 plus gold
Mouse Zowie ec2-a
Keyboard Trust
Software Windows 10
Benchmark Scores Cinebench r20: 631/6232 Cinebench r23: 1628/15901 CPU-Z: 672/6960 SotTR:221FPS
Im here again for more tips from everyone for now I'm stable at 3866mhz. Everything with all voltages on auto except ram voltage at 1.45v. If I go to more frequency I got whea errors and lower CL doesn't boot, tried this 2 things with the same voltage and timings except CL lowered to 14.
Any recommendations for voltages or timings?
Attached my timmings and aida64 benchmark.
 

Attachments

  • ZenTimings_Screenshot.png
    ZenTimings_Screenshot.png
    32.4 KB · Views: 144
  • cachemem1.png
    cachemem1.png
    92.9 KB · Views: 132
Joined
Nov 5, 2021
Messages
40 (0.04/day)
current.png

Bit of an update. It had 3 errors with trrds/l 4/6 and tfaw of 16 so i backed it off to 6/8 and 24 respectively and it passed 20 runs with no errors. I don't know if its worth trying to go any lower or not, what do you think @Taraquin ?
I've only changed the primaries, TRC, TRFC. TRRDS/L, TFAW and AddrCmdDrvStr. Everything else is on auto still
 
Joined
Aug 9, 2019
Messages
1,572 (0.87/day)
Processor 7800X3D 2x16GB CO
Motherboard Asrock B650m HDV
Cooling Peerless Assassin SE
Memory 2x16GB DR A-die@6000c30 tuned
Video Card(s) Asus 4070 dual OC 2610@915mv
Storage WD blue 1TB nvme
Display(s) Lenovo G24-10 144Hz
Case Corsair D4000 Airflow
Power Supply EVGA GQ 650W
Software Windows 10 home 64
Benchmark Scores Superposition 8k 5267 Aida64 58.5ns
View attachment 225465
Bit of an update. It had 3 errors with trrds/l 4/6 and tfaw of 16 so i backed it off to 6/8 and 24 respectively and it passed 20 runs with no errors. I don't know if its worth trying to go any lower or not, what do you think @Taraquin ?
I've only changed the primaries, TRC, TRFC. TRRDS/L, TFAW and AddrCmdDrvStr. Everything else is on auto still
Try trrd 5/7 and faw 20, if error go back to 6/8, 24, wtr 4/12, trdrd/wrwrscl 4, remember that rc should be tras + trp, so either tras 36 or rc 55. You can try 35 ras and 54 rc, might work, rfc 480 or 464 might work, wr should be 16, rtp 8, wrrd 3.
 
Joined
Nov 5, 2021
Messages
40 (0.04/day)
Try trrd 5/7 and faw 20, if error go back to 6/8, 24, wtr 4/12, trdrd/wrwrscl 4, remember that rc should be tras + trp, so either tras 36 or rc 55. You can try 35 ras and 54 rc, might work, rfc 480 or 464 might work, wr should be 16, rtp 8, wrrd 3.
alright cheers ill give that a crack later tonight. I just get confused about rules. The only one i vaguely remembered was that tfaw is meant to be 4x trrds
 
Joined
Aug 9, 2019
Messages
1,572 (0.87/day)
Processor 7800X3D 2x16GB CO
Motherboard Asrock B650m HDV
Cooling Peerless Assassin SE
Memory 2x16GB DR A-die@6000c30 tuned
Video Card(s) Asus 4070 dual OC 2610@915mv
Storage WD blue 1TB nvme
Display(s) Lenovo G24-10 144Hz
Case Corsair D4000 Airflow
Power Supply EVGA GQ 650W
Software Windows 10 home 64
Benchmark Scores Superposition 8k 5267 Aida64 58.5ns
alright cheers ill give that a crack later tonight. I just get confused about rules. The only one i vaguely remembered was that tfaw is meant to be 4x trrds
I just remembered that Hynix often don't like the lowest RRDS/FAW. 5/20 might work though.

Rules are:
RAS=On B-die CL+RCDRD, can work higher or lower. On CJR/DJR it can work, but sometimes RC gets too low so you must use CL+RCDRD+1 or 2. On Micron rev E and Samsung D you must add even more as RP can run really low on these, but RC can't.
RAS+RP=RC
RFC=lowest stable number divideable by 16 if on 2x16sticks or by 8 if on 2x8 or 4x8
FAW=RRDS x 4, RRDL=RRDS or +1 or 2
WR=CL and RTP=WR/2, WR should be even so if CL=15 set WR to 14 and RTP to 7
WTRS=3 or 4, WTRL=RTP or lowest stable above RTP
WRWR/RDRD-SCL: 2, 3 or 4
CWL=CL
RDWR=RCDRD/2 or +1 or 2 if you have dual rank
WRDD=1 if 2x8GB or 4x8, 3 if 2x16 or 4x16.

1T and geardown mode rounds up CL, CWL, WR and RTP to even numbers so I recommend 2T instead.

If you don't follow the rules everything can work out fine, but you get higher chance of errors, timeouts etc. I run WR/RTP at 10/5 on my setup as performance is slightly better vs 16/8 which I should according to rules.

I learned this from anta777, he has made some of the configs for TM5 and seems to know much.
 
Last edited:
Joined
Nov 5, 2021
Messages
40 (0.04/day)
19.11.21.png

i changed only the wtrs/l values to 4 and 12 and got 2 errors, is this supposed to be changed in conjuction with something else?
 
Joined
Aug 9, 2019
Messages
1,572 (0.87/day)
Processor 7800X3D 2x16GB CO
Motherboard Asrock B650m HDV
Cooling Peerless Assassin SE
Memory 2x16GB DR A-die@6000c30 tuned
Video Card(s) Asus 4070 dual OC 2610@915mv
Storage WD blue 1TB nvme
Display(s) Lenovo G24-10 144Hz
Case Corsair D4000 Airflow
Power Supply EVGA GQ 650W
Software Windows 10 home 64
Benchmark Scores Superposition 8k 5267 Aida64 58.5ns
View attachment 225645
i changed only the wtrs/l values to 4 and 12 and got 2 errors, is this supposed to be changed in conjuction with something else?
No, I think they can be changed by themselves. 4 and 12 is very relaxed so weird that you get error. Revert back to what worked and try the next timings. Wtr don't impact performance that much.
 
Joined
Jun 1, 2021
Messages
87 (0.08/day)
System Name NR200 SuperSport Speed Machine yet quiet
Processor Ryzen 7 5800x
Motherboard Gigabyte B550I AORUS PRO AX
Cooling Noctua NH-D15 w/Thermal Grizzly Kryonaut
Memory Corsair Vengeance LPX 32 GB (2 x 16 GB) DDR4 3600Mhz
Video Card(s) Sapphire AMD Radeon RX 6800 Nitro plus 16GB With SAM and OCed
Storage Samsung 970 EVO PLUS 1TB NVMe M.2 Solid State Drive
Display(s) Gigabyte G27Q GAMING MONITOR 1440p 144hz IPS HDR "1ms"
Case Cooler Master NR200P w/Mods w/Full Noctua fans
Audio Device(s) Beyerdynamic DT 770 PRO 80 Ohm/Ifi Zen Dac v2
Power Supply Corsair SF 600 W 80+ Platinum Certified Fully Modular SFX
Mouse Glorious Model D Matte White w/Mousepad: Cooler Master MP511 (Size XL)
Keyboard Cooler Master CK550 Gateron Red Switch w/Tai-Hao Doubleshot PBT Backlit - Cool Gray/Navy
Software Microsoft Windows 10 Pro Full - USB 32/64-bit
Benchmark Scores https://www.3dmark.com/spy/27840416
View attachment 225645
i changed only the wtrs/l values to 4 and 12 and got 2 errors, is this supposed to be changed in conjuction with something else?
Put your VSOC to 1.15-1.17V and your VDDG CCD to 1000mv or 1010mv

1637265293757.png

1900 MLCK/FCLK/UCLK for me with no WHEA errors
 

DLD

New Member
Joined
Nov 21, 2021
Messages
23 (0.02/day)
Can somebody help me out with my RAM OC?

My kit is G.Skill Trident Z F4-3600C14D-32GTZNA. So B-die with with XMP 14-14-14-34 timings at 3600mhz.

I feel this kit could be doing a lot better than it is currently. My previous kit was a cheaper G.Skill Trident Z at 3600mhz 16-19-19-39. I was able to run it at 3800mhz no problems.

My new better bin kit I can't seem to get it to be stable at 3800mhz. I have it running at 1.5v but have not messed with any other voltages yet.

Here are my current timings. Any help from the experts would be great!
 

Attachments

  • RAM_Latency.PNG
    RAM_Latency.PNG
    172.7 KB · Views: 107

tabascosauz

Moderator
Supporter
Staff member
Joined
Jun 24, 2015
Messages
7,885 (2.38/day)
Location
Western Canada
System Name ab┃ob
Processor 7800X3D┃5800X3D
Motherboard B650E PG-ITX┃X570 Impact
Cooling NH-U12A + T30┃AXP120-x67
Memory 64GB 6400CL32┃32GB 3600CL14
Video Card(s) RTX 4070 Ti Eagle┃RTX A2000
Storage 8TB of SSDs┃1TB SN550
Case Caselabs S3┃Lazer3D HT5
Can somebody help me out with my RAM OC?

My kit is G.Skill Trident Z F4-3600C14D-32GTZNA. So B-die with with XMP 14-14-14-34 timings at 3600mhz.

I feel this kit could be doing a lot better than it is currently. My previous kit was a cheaper G.Skill Trident Z at 3600mhz 16-19-19-39. I was able to run it at 3800mhz no problems.

My new better bin kit I can't seem to get it to be stable at 3800mhz. I have it running at 1.5v but have not messed with any other voltages yet.

Here are my current timings. Any help from the experts would be great!

3600CL14 is a relatively high bin for dual rank. Why are you trying to drop tRCD and tRP so low when you haven't touched any of the other timings? If you haven't gotten a feel yet for how much volt-happy your kit is, you're just torpedoing your testing by trying to force it down to 14-13-13.

I run 1.51V for 3800 14-14-14, and ~1.42V for 3600 14-14-14. Kinda depends on the board though, I was up to 1.53V for 14-15-15 on another board.

3800.png


Try something like this to start, let's see how much volts you need to be stable:

3600cl14 final timings.png
 

DLD

New Member
Joined
Nov 21, 2021
Messages
23 (0.02/day)
Ok I set everything to what you suggested. I'll run Karhu for a while and see if I get any errors.

I would ultimately like to get this kit running at 3800mhz with lowest timings possible.

3600CL14 is a relatively high bin for dual rank. Why are you trying to drop tRCD and tRP so low when you haven't touched any of the other timings? If you haven't gotten a feel yet for how much volt-happy your kit is, you're just torpedoing your testing by trying to force it down to 14-13-13.

I run 1.51V for 3800 14-14-14, and ~1.42V for 3600 14-14-14. Kinda depends on the board though, I was up to 1.53V for 14-15-15 on another board.

View attachment 225997

Try something like this to start, let's see how much volts you need to be stable:

View attachment 225998
Ok after 17 min of Karhu at your settings I got an error. The settings I posted got 0 errors after 1 hour. I'm guessing its the tRFC?
 

tabascosauz

Moderator
Supporter
Staff member
Joined
Jun 24, 2015
Messages
7,885 (2.38/day)
Location
Western Canada
System Name ab┃ob
Processor 7800X3D┃5800X3D
Motherboard B650E PG-ITX┃X570 Impact
Cooling NH-U12A + T30┃AXP120-x67
Memory 64GB 6400CL32┃32GB 3600CL14
Video Card(s) RTX 4070 Ti Eagle┃RTX A2000
Storage 8TB of SSDs┃1TB SN550
Case Caselabs S3┃Lazer3D HT5
Ok I set everything to what you suggested. I'll run Karhu for a while and see if I get any errors.

I would ultimately like to get this kit running at 3800mhz with lowest timings possible.

Ok after 17 min of Karhu at your settings I got an error. The settings I posted got 0 errors after 1 hour. I'm guessing its the tRFC?

140ns tRFC should be easily doable for a kit of this caliber. It's Bdie, everything is relative to how much VDIMM you're running - how much voltage did you set for those settings?

3800 15-15-15 or 16-16-16 should be a walk in the park, I'd say 1.45-1.48V and 1.35-1.4V respectively. If you want 14-14-14, expect to pay for it in VDIMM, up to 1.5-1.55V or so.

What's your RAM cooling setup like and how are your RAM temps in Karhu? If you don't have good passive or active airflow directly over the DIMMs, you can forget about 3800CL14 and probably 3800CL15 immediately. At those timings I sent, if you hit 50C on the DIMMs you're done, errors will start rolling in. If not even earlier at a lower temperature.
 
Last edited:

DLD

New Member
Joined
Nov 21, 2021
Messages
23 (0.02/day)
140ns tRFC should be easily doable for a kit of this caliber. It's Bdie, everything is relative to how much VDIMM you're running - how much voltage did you set for those settings?

3800 15-15-15 or 16-16-16 should be a walk in the park, I'd say 1.45-1.48V and 1.35-1.4V respectively. If you want 14-14-14, expect to pay for it in VDIMM, up to 1.5-1.55V or so.

What's your RAM cooling setup like and how are your RAM temps in Karhu? If you don't have good passive or active airflow directly over the DIMMs, you can forget about 3800CL14 and probably 3800CL15 immediately. At those timings I sent, if you hit 50C on the DIMMs you're done, errors will start rolling in. If not even earlier at a lower temperature.
You are right. Me being a noob at this, still had it at 1.5v. I put it back to 1.45 (which is XMP) and it ran 55 min and over 2,000% coverage without an error.

It still reached 51c on 1 stick and 46 on another. My room was getting hot though so after opening the door the hottest stick stayed at 49c.

Here is a pic of my PC. Not the greatest quality. 3 AL120 Uni Fans as intake in the front. Two 140mm exhaust on the top for radiator and one 140mm exhaust in the back.

Is there any way to get airflow directly over the DIMMs? I know Corsair made something for their sticks.
 

Attachments

  • IMG_20211121_162808987.jpg
    IMG_20211121_162808987.jpg
    1.4 MB · Views: 56

tabascosauz

Moderator
Supporter
Staff member
Joined
Jun 24, 2015
Messages
7,885 (2.38/day)
Location
Western Canada
System Name ab┃ob
Processor 7800X3D┃5800X3D
Motherboard B650E PG-ITX┃X570 Impact
Cooling NH-U12A + T30┃AXP120-x67
Memory 64GB 6400CL32┃32GB 3600CL14
Video Card(s) RTX 4070 Ti Eagle┃RTX A2000
Storage 8TB of SSDs┃1TB SN550
Case Caselabs S3┃Lazer3D HT5
You are right. Me being a noob at this, still had it at 1.5v. I put it back to 1.45 (which is XMP) and it ran 55 min and over 2,000% coverage without an error.

It still reached 51c on 1 stick and 46 on another. My room was getting hot though so after opening the door the hottest stick stayed at 49c.

Here is a pic of my PC. Not the greatest quality. 3 AL120 Uni Fans as intake in the front. Two 140mm exhaust on the top for radiator and one 140mm exhaust in the back.

Is there any way to get airflow directly over the DIMMs? I know Corsair made something for their sticks.

Unless you run 2-3000rpm fans on the front, airflow is nowhere near enough if you rely on case fans.

Just stand a small 92mm fan or something on top of the GPU backplate and point it at the dimms. It's easy and nothing permanent that you can't easily unplug.

Corsair's RAM cooler is sucky - I returned it because it was barely better than running a 120mm exhaust above the RAM. I run a push pull C14S setup that provides plenty of downwards airflow, the Corsair cooler couldn't hold a candle even at near max speed. But if your existing benchmark is basically zero airflow, it'll help.
 

DLD

New Member
Joined
Nov 21, 2021
Messages
23 (0.02/day)
Unless you run 2-3000rpm fans on the front, airflow is nowhere near enough if you rely on case fans.

Just stand a small 92mm fan or something on top of the GPU backplate and point it at the dimms. It's easy and nothing permanent that you can't easily unplug.

Corsair's RAM cooler is sucky - I returned it because it was barely better than running a 120mm exhaust above the RAM. I run a push pull C14S setup that provides plenty of downwards airflow, the Corsair cooler couldn't hold a candle even at near max speed. But if your existing benchmark is basically zero airflow, it'll help.
Ok so if I put a fan over it and keep them at or below 50c what would you recommend I try next? Should I shoot for a 3800mhz clock or just keep tightening the timings at 3600?
 
Joined
Aug 9, 2019
Messages
1,572 (0.87/day)
Processor 7800X3D 2x16GB CO
Motherboard Asrock B650m HDV
Cooling Peerless Assassin SE
Memory 2x16GB DR A-die@6000c30 tuned
Video Card(s) Asus 4070 dual OC 2610@915mv
Storage WD blue 1TB nvme
Display(s) Lenovo G24-10 144Hz
Case Corsair D4000 Airflow
Power Supply EVGA GQ 650W
Software Windows 10 home 64
Benchmark Scores Superposition 8k 5267 Aida64 58.5ns
Ok so if I put a fan over it and keep them at or below 50c what would you recommend I try next? Should I shoot for a 3800mhz clock or just keep tightening the timings at 3600?
I would go for 3800 if stable and no whea 19 in eventviewer. 3800cl15 is about as hard as 3600cl14, but 3800 is generally a bit faster.

As for temps, there is binning involved. Some B-dies can handle a bit over 50C while tweaked, while others need below 45C.

I can run up to 1.5V without direct fan over ram, but I have a airflow case whith intake partially blowing at ram, GPU-blows a bit of air above ram and CPU draws air over ram. Over 1.5V I get overheat after a few minutes in karhu/TM5.

Remember you must run 2T and gear down mode off for cl15 to work.

If you try 3800 keep most timings as tabasco recommends but run:
15 15 15 30 45
272 tRFC
14/7 or 12/6 tWR/tRTP
4/8 or 4/10 wtr if doable.

PS: You won't like this, but disabling led will lower ram temp a bit since led draws a bit of power, thus creating heat, nit by much, but it can be the difference of overheating at 1.48V or 1.5V ;)

3600CL14 is a relatively high bin for dual rank. Why are you trying to drop tRCD and tRP so low when you haven't touched any of the other timings? If you haven't gotten a feel yet for how much volt-happy your kit is, you're just torpedoing your testing by trying to force it down to 14-13-13.

I run 1.51V for 3800 14-14-14, and ~1.42V for 3600 14-14-14. Kinda depends on the board though, I was up to 1.53V for 14-15-15 on another board.

View attachment 225997

Try something like this to start, let's see how much volts you need to be stable:

View attachment 225998
I mostly agree, but tRTP should be half tWR so 12/6, tRFC is better off at 256 as it is best timed if diviable by 16 for dual rank or 8 for SR, these suggestions according to anta777 :) Except for that rest great advice :)

As for you own settings 272 tRFC might be better at 2T allows you to run 5 tRTP, changing tWR to 12 also might improve things :) Might not improve anything ;)

OK so this made no difference it was still unstable at 4/12/24
Change voltages back if it made no difference. Tried lowering tRFC, tWR/tRTP and scls yet?
 
Last edited:
Top