Your score isn't higher than mine.
https://www.3dmark.com/compare/spy/51837341/spy/54134795#
But seriously though what is up with your scores? The benchmark must be worthless, yeah, or you're cheating somehow.
![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
-
Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.
AMD Ryzen 9 9950X3D Performance, Overclocking and Cooling Thread
- Thread starter Deimos
- Start date
https://www.3dmark.com/compare/spy/51837341/spy/54134795#
But seriously though what is up with your scores? The benchmark must be worthless, yeah, or you're cheating somehow.
This is my best result with the new ram installed.
https://www.3dmark.com/3dm/128988069?
I set my infinity fabric to 2067mhz UCKL=MEMCLK which gave me almost 600 extra points. I'm running +200 boost clock override and PBO -20 all cores, motherboard limits. I'm good with this, not going to bother tweaking further.
The kit I got is Hynix M die. I pretty much used the same settings as my previous kit, except for the main timings which are now 30-36-36 instead of 32-38-38.
Here are my full settings:
https://www.3dmark.com/3dm/128988069?
I set my infinity fabric to 2067mhz UCKL=MEMCLK which gave me almost 600 extra points. I'm running +200 boost clock override and PBO -20 all cores, motherboard limits. I'm good with this, not going to bother tweaking further.
The kit I got is Hynix M die. I pretty much used the same settings as my previous kit, except for the main timings which are now 30-36-36 instead of 32-38-38.
Here are my full settings:
Last edited:
- Joined
- Sep 17, 2019
- Messages
- 1,503
Awesome, this is exactly the type of thing I am looking for! People who actually know how to configure past "stock"
Thanks for not only the link, but how you have set things up as I am close to jumping on a new build and some tweaking ideas will help.So, contrary to popular belief, all core OC is the way to fly on AM4.Your score isn't higher than mine.
https://www.3dmark.com/compare/spy/51837341/spy/54134795#
But seriously though what is up with your scores? The benchmark must be worthless, yeah, or you're cheating somehow.
All core OC, no PBO, CPPC Off, SMT off. I went into this with graphs well over a year ago now, but basically, it gave me a major boost in not only synthetics, but a very measurable boost in gaming as well, even at 4K! If you are interested, I can send you the link, it's on these forums. Also, a great kit of Samsung B-die's pushed as fast as I could with the IMC and very tight timings really helped as well...Don't touch scalar, it affects the silicon reliability (unlike the other parameters). Unless you plan to only use it for 6 months to a year, I would not touch it. Scalar x10 basically takes it from a 10 year (typical consumer grade) expected lifespan to 1 year.
Removed scalar. Kept all other settings. Will test stability with AIDA and y cruncher when I get back home.
Skatrerbencher down played the risk of scalar in his recent video but quite a few people think degraded CPUs. I don’t want to take the risk. Thanks.
I mean, the description of the setting is to pretty much ignore silicon stress levels. My motherboard bios says it right there in the description and I always thought 'nah, I'm good thanks.'
I have changed my memory settings, is anyone having issues running tREFI really high? I used to run it at 50,000 and also tried the maximum setting but after watching buildzoid's video about tweaking 96GB kits, he said something about having hitching if tREFI is too high.
I've experienced pretty bad hitching when viewing multiple video streams, and any website that is heavy on graphics, for example, the unifi CKG2 network insights causes real bad video hitching. After I lowered tREFI to 25,000 it seems to be OK. I have 7 security cameras up on one screen , Emby on another, and the Unifi network console up on a third screen and there is no hitching what so ever. Damn I wish I knew tREFI was the culprit earlier, I have been putting up with this nonsense for months on my 7800X3D with 32GB ram.
I'll carry on using it like this for a while and report back. For now, here are my current timings:
Edit: BTW setting max tREFI isn't unstable and a full OCCT memtest runs fine, but at the max setting I get system-wide hitching, whereas 50,000 I get hitching only in very specific circumstances. 25,000 seems good so far.
I've experienced pretty bad hitching when viewing multiple video streams, and any website that is heavy on graphics, for example, the unifi CKG2 network insights causes real bad video hitching. After I lowered tREFI to 25,000 it seems to be OK. I have 7 security cameras up on one screen , Emby on another, and the Unifi network console up on a third screen and there is no hitching what so ever. Damn I wish I knew tREFI was the culprit earlier, I have been putting up with this nonsense for months on my 7800X3D with 32GB ram.
I'll carry on using it like this for a while and report back. For now, here are my current timings:
Edit: BTW setting max tREFI isn't unstable and a full OCCT memtest runs fine, but at the max setting I get system-wide hitching, whereas 50,000 I get hitching only in very specific circumstances. 25,000 seems good so far.
Last edited:
- Joined
- Sep 17, 2019
- Messages
- 1,503
So, being a PC junkie and nerd who loves to endlessly tweak and stay on top of tech, I Ordered my new PC parts for pickup tomorrow at Microcenter!
Hopefully I can get the ram im buying to work around 6000Mhz CAS 26 1:1 or so with the kit specified. They are out of 64GB 2 stick kits, so going for a 2 stick 96GB kit of the same type. I plan to tune all the timings manually anyway once I get things running.
Open to any suggestions as I pick this up tomorrow at some point.
- MSI X870E MPG Carbon AM5 board
- AMD 9950X3D CPU
- Corsair Domintor Titanium Black 96GB (2x48GB) 6600Mhz DDR5 CAS 32
- Samsung 9100 4TB nvme pcie 5.0 drive
- Reusing my MSI 4090 Suprim X until I can snag a 5090 someday
- Reusing my SB AE-7 Sound Card
- Seagate 16TB 7200rpm backup drive
- NZXT C1200 1200 Watt 80 Plus Gold ATX Fully Modular Power Supply - ATX 3.1
- Corsair 5000D Airflow Tempered Glass Mid-Tower ATX Computer Case - Black (my second one, love this case)
- Corsair iCUE LINK RGB TITAN 360 RX LCD 360mm AIO
- Corsair iCUE LINK QX120 RGB Magnetic Dome Bearing 120mm PWM Fans
Hopefully I can get the ram im buying to work around 6000Mhz CAS 26 1:1 or so with the kit specified. They are out of 64GB 2 stick kits, so going for a 2 stick 96GB kit of the same type. I plan to tune all the timings manually anyway once I get things running.
Open to any suggestions as I pick this up tomorrow at some point.
- MSI X870E MPG Carbon AM5 board
- AMD 9950X3D CPU
- Corsair Domintor Titanium Black 96GB (2x48GB) 6600Mhz DDR5 CAS 32
- Samsung 9100 4TB nvme pcie 5.0 drive
- Reusing my MSI 4090 Suprim X until I can snag a 5090 someday
- Reusing my SB AE-7 Sound Card
- Seagate 16TB 7200rpm backup drive
- NZXT C1200 1200 Watt 80 Plus Gold ATX Fully Modular Power Supply - ATX 3.1
- Corsair 5000D Airflow Tempered Glass Mid-Tower ATX Computer Case - Black (my second one, love this case)
- Corsair iCUE LINK RGB TITAN 360 RX LCD 360mm AIO
- Corsair iCUE LINK QX120 RGB Magnetic Dome Bearing 120mm PWM Fans
- Joined
- Sep 17, 2019
- Messages
- 1,503
Alrighty, so system built, was tweaking PBO / memory / timings all weekend. I ended up on the below, stable for gaming and P95 stable so far as well. They ran out of 2x32GB kits at MC, so I ended up with this (2x48) 96GB kit, I was surprised I was able to get it to 1:1 6000Mhz with tight timings with a DR kit of ram. However, I can't seem to get to lower 60ns range for latency, so if anyone has any tips/tricks, let me know. I can't push 6200Mhz or above with 1:1 ratio, unless there is some voltage I am missing. DDR5 is at 1.4V (along with the voltage that feeds this), IO Voltage is at 1.37V, SOC at 1.25V, rest is on auto. The RAM is actually rated for 6600Mhz, so the RAM is not really the limit here I feel. I need to have GDM enabled as well. Disabled "works", but as soon as I load a game, I get audio popping and whatnot, so clearly, not stable that way.
PBO limits are shown below as well...
I can push FCLK to 2133Mhz with no issues (2167 was unstable in testing), so to gain a little more I pushed BCLK to 101Mhz, which bumped FCLK, UCLK and MCLK up as well slightly, still below where I can no longer run 1:1. Guess I did not win the IMC 1:1 lottery, but with tight timings and the OC here, solid performance gain I believe. I feel I would just be chasing 1% or less at this point. Open to any suggestions!
Also of interesting note; disabling SMT like with AM4 provides a nice boost in 3Dmark, but more interesting is in games this time around, it makes little or negative difference. So far, SMT on for gaming (not benching) seems to have a slight boost for a change. I imagine this is because of the X3D scheduling through gamebar where having 8C/16T is more useful than it only having access to 8C/8T on the X3D chip.
3dmark score:
https://www.3dmark.com/spy/54837278
PBO limits are shown below as well...
I can push FCLK to 2133Mhz with no issues (2167 was unstable in testing), so to gain a little more I pushed BCLK to 101Mhz, which bumped FCLK, UCLK and MCLK up as well slightly, still below where I can no longer run 1:1. Guess I did not win the IMC 1:1 lottery, but with tight timings and the OC here, solid performance gain I believe. I feel I would just be chasing 1% or less at this point. Open to any suggestions!
Also of interesting note; disabling SMT like with AM4 provides a nice boost in 3Dmark, but more interesting is in games this time around, it makes little or negative difference. So far, SMT on for gaming (not benching) seems to have a slight boost for a change. I imagine this is because of the X3D scheduling through gamebar where having 8C/16T is more useful than it only having access to 8C/8T on the X3D chip.
3dmark score:
https://www.3dmark.com/spy/54837278
Last edited:
I'm finding the 9950X3D a bit better for gaming than the reviewers let on (not a jab at them), because the stock AMD drivers/game bar solution is imperfect. I set BIOS to "CPPC prefer vcache", uninstalled AMD drivers, and once in Windows I don't let the cores park. EXPO on (6000cl30), everything else stock.
Games like Total War Troy (it's the most recent TW game I own but I assume new ones scale the same?) or Bannelord can make use of the 2nd CCD, loading all cores and threads. The irony is when I watched one reviewer who used the default AMD setup instead, they got worse than 9800X3D performance in the same game.
For low threaded games however it is still better to manually assign the cores, for example Arma 3, BG3 and KC Deliverance (the first one) suffer a 10% penalty and uneven frametimes if I let them spill over to the 2nd CCD, but forcing the affinity via task manager (or any tool of your choosing) restores the single CCD buttery smooth experience (game bar and AMD drivers can do the same of course)
Windows will put some background tasks on the second CCD (like 1-2% load) while your game is focused so it doesn't hurt to leave it enabled instead of parking it like AMD drivers do, as long as you don't have thermal or power budget issues affecting the vcache CCD (at stock and while gaming, you will not).
I guess the AMD driver solution is really aggressive to try and avoid the issues with some games that don't like the cross CCD latency (it is kinda bad after all, 10% fps loss and bad frametimes), they would have to test every single game one by one. I'm sure Microsoft scheduler could be improved with dual CCDs in mind though, if they are to remain a thing.
Games like Total War Troy (it's the most recent TW game I own but I assume new ones scale the same?) or Bannelord can make use of the 2nd CCD, loading all cores and threads. The irony is when I watched one reviewer who used the default AMD setup instead, they got worse than 9800X3D performance in the same game.
For low threaded games however it is still better to manually assign the cores, for example Arma 3, BG3 and KC Deliverance (the first one) suffer a 10% penalty and uneven frametimes if I let them spill over to the 2nd CCD, but forcing the affinity via task manager (or any tool of your choosing) restores the single CCD buttery smooth experience (game bar and AMD drivers can do the same of course)
Windows will put some background tasks on the second CCD (like 1-2% load) while your game is focused so it doesn't hurt to leave it enabled instead of parking it like AMD drivers do, as long as you don't have thermal or power budget issues affecting the vcache CCD (at stock and while gaming, you will not).
Total War Troy in single CCD mode (=9800X3D + 3%) gives 117fps and 150fps on 2 different scenes. Dual CCD mode I get 164fps and 201fps. Repeatable results. (This is in line with City Skyline 2 as benchmarked by other users: https://www.reddit.com/r/CitiesSkyl..._x3d_cpu_insights_updated_7800x3d_vs_9800x3d/ )
For my testing in single CCD mode, the CPU is pegged at 100% from start to finish, with SMT on. In both cases the GPU is far far below 100%, down to 50ish average in the 8c run (4090), and up to 70-80 with dual CCD.
M&B 2 Bannerlord goes from 115 to 119fps on a quick torture test I made by enabling 2nd CCD, that's small but again that's still furthering the gap with a 9800X3D (not to mention the 7800X3D I come from). Also repeatable, that scene gives a flat line of frametimes.
Also worth noting that even with both CCDs loaded in those games, average CPU clocks are in the 5300-5400 range for all cores! That's higher than a 9800X3D. That is kinda nuts but pulling up to 160w is of course not efficient anymore.
I'm too lazy to benchmark multiplayer games but Rivals and BF2042 certainly load all threads heavily when I'm playing on the 360hz screen, whether they run smoother than single CCD it's hard to say but it isn't noticeably worse at least, and there are no drops that I can see or perceive, while in Arma 3 and BG3 I could definitely feel the hitches.
For my testing in single CCD mode, the CPU is pegged at 100% from start to finish, with SMT on. In both cases the GPU is far far below 100%, down to 50ish average in the 8c run (4090), and up to 70-80 with dual CCD.
M&B 2 Bannerlord goes from 115 to 119fps on a quick torture test I made by enabling 2nd CCD, that's small but again that's still furthering the gap with a 9800X3D (not to mention the 7800X3D I come from). Also repeatable, that scene gives a flat line of frametimes.
Also worth noting that even with both CCDs loaded in those games, average CPU clocks are in the 5300-5400 range for all cores! That's higher than a 9800X3D. That is kinda nuts but pulling up to 160w is of course not efficient anymore.
I'm too lazy to benchmark multiplayer games but Rivals and BF2042 certainly load all threads heavily when I'm playing on the 360hz screen, whether they run smoother than single CCD it's hard to say but it isn't noticeably worse at least, and there are no drops that I can see or perceive, while in Arma 3 and BG3 I could definitely feel the hitches.
I guess the AMD driver solution is really aggressive to try and avoid the issues with some games that don't like the cross CCD latency (it is kinda bad after all, 10% fps loss and bad frametimes), they would have to test every single game one by one. I'm sure Microsoft scheduler could be improved with dual CCDs in mind though, if they are to remain a thing.
Last edited:
Where would one find this setting?
AMD CBS > SMU on my Gigabyte board.
Note that this tells Windows to treat vcache cores as the best cores, anything other than games (or some vcache loving app) will thus be slower unless you manually clock them the same.
Personally I don't care, everything outside of games is more than fast enough, but that needs to be said.
Note that this tells Windows to treat vcache cores as the best cores, anything other than games (or some vcache loving app) will thus be slower unless you manually clock them the same.
Personally I don't care, everything outside of games is more than fast enough, but that needs to be said.
Last edited:
- Joined
- Sep 17, 2019
- Messages
- 1,503
So it turns out, I was just rusty at overclocking memory and tightening timings (plus learning the AM5 platform). I guess I have a great IMC afterall!
I'm now running 6400Mhz DDR5, 1:1 with FCLK at 2133Mhz as well. Improved latency AND bandwidth. I'm sure I can fine tune things more, but this feels pretty solid. Not sure I can get to the low 60ns range with this amount of memory, but pretty damn happy with 67ns on 96GB DR RAM at this speed so far!
Last edited:
1.1.2.3.5...
[H]ard|Gawd
- Joined
- Jul 13, 2006
- Messages
- 1,715
These are very respectable numbers imho. Especially for 96GB.
Can you remind me of the mobo and ram you got in case i want to duplicate your results?
- Joined
- Sep 17, 2019
- Messages
- 1,503
Sig has my specs now, but to make it EZ PZ....
- MSI X870E MPG Carbon Wifi 7 AM5 board
- Corsair Dominator Titanium Black 96GB (2x48GB) 6600Mhz DDR5
1.1.2.3.5...
[H]ard|Gawd
- Joined
- Jul 13, 2006
- Messages
- 1,715
Attachments
- Joined
- Sep 17, 2019
- Messages
- 1,503
From what I have been seeing online, I may have won the IMC lottery. These sticks are Hyinx, but they are also dual rank, so harder on the IMC. I also saw the last AGESA added up to 8ns to latency, so in reality, if I were to revert back, it's probably sub/lower 60ns. From what I read tho, the latest AGESA added performance and also X3D support for the 9xxx series, which I need.... lol. I can enable "latency killer", which reverts those changes in my bios, but at the cost of cpu performance. Seeing as I'm on X3D, 8ns isn't gonna mean squat compared to CPU speed.I don't see sigs.
Thanks!
I may try 6600 1:1 someday. Interestingly, I can get into Windows with it, but I think I would have to tweak timings more to get it stable. I tried it with only 1.3V SoC and 1.37VDDIO and it would insta-lock during P95... lol. But the fact it booted, means I might be able to do it.
EDIT: Tried latency killer... new AIDA64 attached (and updated timings). Cb23 / Cb24 scored SLIGHTLY higher with Latency Killer enabled, games and 3dmark bench almost exactly the same. The CPU does boost 50~100Mhz slower, but that is almost trivial, so maybe I'll try gaming this way for a bit just to see if it even matters.
Last edited:
Lucky sob =P. Yea...I got an ABSOLUTE DUD of a IMC on mine. I have to run 1.18v vsoc just to get 6000mhz stable. At 1.15 it wasn't stable....got errors in memory testing. Was too lazy to fine tune between those values. I can hit 6200mhz and 2067 fclk at 1.3 vsoc...Which is just terrible. I'm curious if it's my board or chip, most likely chip, running an x870e proart. I'm just worried about running 1.3v vsoc 24/7 considering my machine is always on and a workstation...Don't want to risk any down time if things go wrong. Luckily I don't have an ASRock board =P, so can avoid the current burned chip fiasco they are dealing with...haha.
I've seen two 9800X3Ds with burn marks on the Asrock subreddit (not to say there aren't a few more). Most are failures to boot due to memory compatibility issues - supposedly remedied with the 3.20 BIOS, but I still see some failures to boot popping up over there now and then. AMD and Asrock both released press statements to that effect.
Yea. I haven't read up on the continual posts regarding burnt up X3D cpu's, but one of the reasons why I don't want to push 1.3v vsoc 24/7. I bet most cases now are probably from people who haven't updated there bios...but that's just speculation on my behalf. Lucky for those that a bios update remedied their situation...but definitely a problem for sure. I honestly will never use ASRock myself, got my own personal bias. I gave them a shot twice in the past late 2000's and early 2010's, both were nothing but headaches...Probably much better now, but the experience has stuck and sworn me off using them for my personal builds.
- Joined
- Sep 17, 2019
- Messages
- 1,503
I think 1.3V SoC is upper limit, I think most that were dying were pushing 1.35V SoC. But yeah, for 6400, 1:1 UCLK=MEMCLK, 2133 FCLK I am using 1.275V SoC... I have no problem running this 24/7 and temps are well under control. Even if I wanted to push it, I paid for the 2 year MC warranty, so I could fry this puppy and just get a new one. However, I do feel like I got a golden sample. It's game and prime95 stable and of course can loop CB endlessly too without crashing so far.
Yea...Got a winner on your hands. I got mine from amazon on launch day...
BTW...is your CPU VDDIO accurate? that's pretty darn low. I tried .9 for shits and giggles after seeing yours, no boot. 1.1, no boot, 1.2...booted, running karhu right now to verify, then going to run some p95, y cruncher, etc. May as well lower whatever voltage I can considering my dud of a imc...lol. I'm going to be on water (custom loop) once I throw this from my test bench into my rig, so temps won't really be the biggest concern...But just worried about burning out the chip if I 6200/2067 @ 1.3v. 1.35v vsoc was indeed problematic, and AMD rectified that by lowering the maximum limit, but considering they did that once already, I have a feeling there in the "wait and see" what happens territory...speculation of course.
- Joined
- Sep 17, 2019
- Messages
- 1,503
My CPU VDDIO is 1.375V (which is below red limit of 1.4V on my motherboard).
Look on the plus side, if it is covered under warranty, maybe you will not get a dud next time? Based on what I see online, I might be a little more careful with mine as I think I lucked out IMC wise for sure, although I do not plan to push the memory any higher. I was going to attempt an all-core OC though of 5.6Ghz. On my old 5950X, all core OC was the way to fly vs PBO. Can't say I notice any issues with this CPU tho, the 9950X3D is a beast.
Ahhh... That makes a lot more sense, was a bit shocked seeing your CPU VDDIO on those zen timings screen shots.
Unfortunately I can not afford any down time. Hence why I'm on my 5950X machine right now for my workstation till I have my 9950X3D fully tuned. Plus, got a 1600w Seasonic Platinum psu coming on the 18th...So not in a rush to swap them out till then. If I did push 1.3v vsoc and something did happen, would be very problematic considering it would be my main machine, and my AM4 would be migrated over to a test rig.
- Joined
- Sep 17, 2019
- Messages
- 1,503
That's fair, in my case I'd just throw my 4090 back in my 5950X setup and keep rolling. But my local MC is 20 minutes away, so I'd be set that day honestly. My SoC is only at 1.275V, not 1.3V. It's just 1.3V is the accepted high safe limit (for now). I tried 1.25V, but for 6400 1:1, it was not p95 stable, it was gaming stable tho, but that's not acceptable to me, I want to ensure full stability when I OC.
Although booting into windows 1:1 with 6600 has me tempted to try and stabilize it, but I'm already just below what I feel are safe voltages for SoC and VDDIO.
- Joined
- Sep 17, 2019
- Messages
- 1,503
Anyone here mess with Scalar yet on this chip? I decided to give it a go today (after reading so many for and against it threads), landed on a scalar of around 5X. It definitely raises single core and multicore clocks from 1X.
At 5X, with CO of -30, FBoost of +150mhz on all of my cores (known stable for me after tons if testing), i was at about 85C under 10 minute Cinebench throttle testing. At 1X, it doesn't top 75C.
MT score increase in Cb23 by over 1000, single core was up as well. Saw this same general increase in 3dmark too. In terms of gaming, lows were slightly higher, but I really had to use CPU limiting settings to see it. Overall at 4K gaming, might help some 1% lows, that's about it. Holding higher boost clocks for longer is a nice touch.
Not sure I'd push past 5X (personally) for what is likely diminishing returns. Testing on 10X the CPU was topping 95C fairly quick under 100% load testing. I feel 5X for me at 85C under unrealistic loads is fairly safe, voltages were not that much higher either, it mostly seemed to really open up PPT, TDC and EDC limits while allowing clocks to run higher overall at those somewhat unlimited ranges it set.
That being said, it's a nice boost to the CPU if you play it right and carefully. I think next I might see if my non x3d CCD can do better than -30 CO.
At 5X, with CO of -30, FBoost of +150mhz on all of my cores (known stable for me after tons if testing), i was at about 85C under 10 minute Cinebench throttle testing. At 1X, it doesn't top 75C.
MT score increase in Cb23 by over 1000, single core was up as well. Saw this same general increase in 3dmark too. In terms of gaming, lows were slightly higher, but I really had to use CPU limiting settings to see it. Overall at 4K gaming, might help some 1% lows, that's about it. Holding higher boost clocks for longer is a nice touch.
Not sure I'd push past 5X (personally) for what is likely diminishing returns. Testing on 10X the CPU was topping 95C fairly quick under 100% load testing. I feel 5X for me at 85C under unrealistic loads is fairly safe, voltages were not that much higher either, it mostly seemed to really open up PPT, TDC and EDC limits while allowing clocks to run higher overall at those somewhat unlimited ranges it set.
That being said, it's a nice boost to the CPU if you play it right and carefully. I think next I might see if my non x3d CCD can do better than -30 CO.
LOCO LAPTOP
[H]F Junkie
- Joined
- May 4, 2006
- Messages
- 12,957
Anyone else notice high core voltage at between 20-30% cpu usage? AsRock x870e nova wifi, I have the voltage set to 1.2v in the bios.
I'm use to intel stuff that I don't recall having this issue.
I'm use to intel stuff that I don't recall having this issue.
At idle I'd see 1.31V with a stock 9800X3D. I've since manually set mine to 1.23V with a 5425MHz clock, and it stays there with my Asrock X670E Steel Legend. This is measured using HWiFO 64.
- Joined
- Sep 17, 2019
- Messages
- 1,503
Yeah, I would not worry about it, it is well within AMD specifications. Unless you are using Scalar, you are bound to the default FIT limits of the CPU. I hit 1.337V when CPU usage is low, under heavy load, it is way less. Unless you have manually OC'ed (which ignores AMD limits) or used Scalar, you are not overriding anything that AMD has set as a limit. Lower loads = higher clocks = more voltage required. Gaming I rarely see 60C, so I feel like it runs great, and that is with a 9950X3D as well.
PBO with a curve offset helps it run a lot cooler for sure.
LOCO LAPTOP
[H]F Junkie
- Joined
- May 4, 2006
- Messages
- 12,957
I was worried because of the reports of Asrock boards and 'burning' up 9800x3ds. At full load I am at or just under 1.2v which is what I would expect. I do have a +200 PBO and a -20 all core curve offset with x5 scalar but I do have limits set manually.
I peak at 78c on my setup during stress tests and ~60c gaming. Still playing around, the scores seem low and memory latency I been fighting with. I can't get it any lower and keep it stable. I don't think these numbers are great based on what I seen.
Last 3dmark with SMT enabled. https://www.3dmark.com/spy/55164495
I peak at 78c on my setup during stress tests and ~60c gaming. Still playing around, the scores seem low and memory latency I been fighting with. I can't get it any lower and keep it stable. I don't think these numbers are great based on what I seen.
Last 3dmark with SMT enabled. https://www.3dmark.com/spy/55164495
- Joined
- Sep 17, 2019
- Messages
- 1,503
I think the fried chips were related to SoC being set way above 1.3V from what I've seen.
Is your CDLO voltage correct? Mine is like 0.95V. Max I've ever seen mine on auto was like 1.0V.
What ICs do you have on your memory? Hynix A or M die? Or Samsung / Micron? Your secondaries and third levels seem somewhat loose. I feel like ur TFRC is insanely high and ur TREFI is way too low. Especially for only 6000.
Last edited:
LOCO LAPTOP
[H]F Junkie
- Joined
- May 4, 2006
- Messages
- 12,957
My CDLO is set to 1.15v, looking at other screenshots from others on reddit it does seem high. ill manually set it and see what happens. It was like this out of the box.
The IC appear to be SK Hynix M-die, What is showing in that screenshot is what was put in with EXPO/Auto, didn't not touch anything there. I was able to bring it down a bit with a few changes but lets be honest, I have no idea what I'm doing and reverted back.
I also just got this ram if you think it's garbage I can return it and get something better. I'm a Gskill guy but people mentioned they suck with AMD setups.
Last edited:
LOCO LAPTOP
[H]F Junkie
- Joined
- May 4, 2006
- Messages
- 12,957
Made some changes based on another user reddit post. I also changed the CLDO voltage to 1.0v
Ill have to stress test these changes but that help the latency a bunch.
Ill have to stress test these changes but that help the latency a bunch.
- Joined
- Sep 17, 2019
- Messages
- 1,503
Nice improvements! Great thing about Hynix A or M die is, you can legit copy & paste the 2nd and 3rd level timings and they will 99% of the time work at any speed. Only the primaries really change with speed settings and voltage. At 6000Mhz, I bet you can hit CL of 28 with 1.4V on the DIMMs, I did with mine before I decided to push to 6400 1:1 CL 30.
- Joined
- Dec 18, 2006
- Messages
- 497
Interesting to read all this info. Have no idea what most of it is lol. Last build was an i7 11 years ago.
This is my parts list....got everything except just waiting on the case to arrive this weekend: https://ca.pcpartpicker.com/user/abba77/saved/6Mxm3C
Got this RAM...it was on the QVL list and not crazy expensive.
I dont plan to overclock too crazy...run the machine 24/7...sometimes have to leave it unattended for a month while travelling...need access to it.
May need to bug some of you for help with some basic stuff to get it run well and little cooler.
This is my parts list....got everything except just waiting on the case to arrive this weekend: https://ca.pcpartpicker.com/user/abba77/saved/6Mxm3C
Got this RAM...it was on the QVL list and not crazy expensive.
I dont plan to overclock too crazy...run the machine 24/7...sometimes have to leave it unattended for a month while travelling...need access to it.
May need to bug some of you for help with some basic stuff to get it run well and little cooler.
LOCO LAPTOP
[H]F Junkie
- Joined
- May 4, 2006
- Messages
- 12,957
Unfortunately It's not stable with just those timing changes. Gonna have to play with it some more, may need to cool the ram, It hit 60c when hell breaks loose.
- Joined
- Sep 17, 2019
- Messages
- 1,503
What memory sticks do you have? If you can provide a link that will help. 60C is insane, even under 100% OCCT load mine don't even get above 52C, and I do not even have any fans on them, just case Airflow. I'm also using 1.39V on the Dimms.
LOCO LAPTOP
[H]F Junkie
- Joined
- May 4, 2006
- Messages
- 12,957
I have a 120mm fan directly on them now and it's about to reach 60c again, after 9 minutes instead of 7 minutes.
This is the ram I got from Central Computers: https://www.corsair.com/us/en/p/mem...s-cl30-amd-expo-memory-kit-cmh64gx5m2b6000z30
LOCO LAPTOP
[H]F Junkie
- Joined
- May 4, 2006
- Messages
- 12,957
I slapped my Intel Gskill set from my 14900k setup, F5-6400j3239g32gx2-rs5k (SK Hynix A-die) and it actually boots at 6200mhz (and 6400mhz but not stable) Ran 10 minutes of OCCT with no errors then raised the FCLK to 2133 and same thing. I know, that doesn't mean it's stable. But I'm already ahead with this set and finally in the 60's for latency.
Gonna stress test this overnight and see how it goes.
Gonna stress test this overnight and see how it goes.