• 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

D

Deimos

[H]ard|Gawd
2FA
Joined
Aug 10, 2004
Messages
1,276
I'm interested to see how people are running these and if anyone has tips, issues, fixes etc.

I picked mine up on Friday and installed it Saturday to replace my 7800X3D. When I built my system a little over a year ago I wanted a 16 core system but cut this corner to make the price a little more agreeable to my wife and I'm glad I did. The 7800X3D was great and I'm passing it on to my Son to enjoy.

Specs (in the sig)
Gigabyte X670 Aorus Elite AX
Gigabyte 4090
32GB 6000MT/s ram

Issues

I had some issues with Winblows 11 but mostly got them ironed out. My 'Balanced' profile was set to limit CPU speeds to 50%, I just did a reset of the profile to make sure I wouldn't run in to any weirdness and CPU scheduling seems to working as intended with the exception being GTA V Enhanced. Game Bar is broken (will not remember things as games).

OpenRGB with hardware monitoring no longer reads the CPU temps so I can no longer show CPU temps with RGB lighting. I have a feeling that the project is on its last legs so I'll have to find an alternative.

Overclocking and performance
I haven't done much other than a negative 10 offset. All core clocks are 4.9-5Ghz on my custom loop. I might try a higher offset later when I'm not so busy with work.

My system would often lag with multiple video streams running at once on the 78003XD (I often leave my security cameras open on a side screen) which I don't get any more. Overall I'm fairly pleased with the performance (why the hell wouldn't I be?)
 
Last edited:
Unfortunately I didn't get a very good 9950X3D IMC wise. Been tweaking it here and there on my new testbench platform before I throw the parts into my main rig. Testing at 6000mhz mclk on 2x48gb DR dimms, vsoc 1.22 Vsoc stable, testing 1.18 right now and so far so good. Unfortunately not stable at 1.15 vsoc. I can not push the memory clock high either...I can get 3100 (6200mhz) mclock stable with a 2066 fclk stable at 1.3 vsoc. At 1.28 is pushed errors an hour into p95. Not sure if 1.29 is stable, but pushing that much voltage for a 24/7 workstation system lingers on my mind and makes me uncomfortable. Unfortunately looks like I'll have to stick with 6000mhz as I'm not sure if 1.3v vsoc would be safe for an always on machine.

Pics below....Apologies for camera phone shots, system is on my test bench, no internet...Thumbnails, so can click to enlarge.

@6000 (tweaked subtimings, stable (P95, OCCT, Asus Realbench), testing at 1.18 vsoc now and looking good) Forgot to take a picture of aida64.
20250317_085411~3.jpg

@6200 mclk 2066 fclk (stable, P95, OCCT, Realbench) Still need to tweak main timings, sub timings are same as above.

20250316_164513~2.jpg
 
Last edited:
KickAssCop said:
Skatterbench is proper OG. However, most of his clocks never work for me :(.
Click to expand...
I used strategy 3 but less aggressive with a -20 all core offset and no boost clock override. I ran his recommended suite for stability and it seems OK. I'm seeing about 5.1Ghz on the v-cache cores during AVX stress testing and I'm happy with that.
 
Under heavy all core loads I'm bumping up against the temperature limit, is that normal? HWMonitor reported the CPU as using almost 260W. I'm tempted to lap the IHS, I got really good results lapping the 7800X3D (temps, didn't help OC at all though).
 
Got a 9950x3D. Following this video and the overclock.net thread. I set it with PBO + 200. Scalar x 10. CO -20 so far like you. The max power I saw was 226 W. Temps Mid 80s with Arctic Freezer 280. No thermal throttling.
 
Ikasu said:
Unfortunately I didn't get a very good 9950X3D IMC wise. Been tweaking it here and there on my new testbench platform before I throw the parts into my main rig. Testing at 6000mhz mclk on 2x48gb DR dimms, vsoc 1.22 Vsoc stable, testing 1.18 right now and so far so good. Unfortunately not stable at 1.15 vsoc. I can not push the memory clock high either...I can get 3100 (6200mhz) mclock stable with a 2066 fclk stable at 1.3 vsoc. At 1.28 is pushed errors an hour into p95. Not sure if 1.29 is stable, but pushing that much voltage for a 24/7 workstation system lingers on my mind and makes me uncomfortable. Unfortunately looks like I'll have to stick with 6000mhz as I'm not sure if 1.3v vsoc would be safe for an always on machine.

Pics below....Apologies for camera phone shots, system is on my test bench, no internet...Thumbnails, so can click to enlarge.

@6000 (tweaked subtimings, stable (P95, OCCT, Asus Realbench), testing at 1.18 vsoc now and looking good) Forgot to take a picture of aida64.
View attachment 717479

@6200 mclk 2066 fclk (stable, P95, OCCT, Realbench) Still need to tweak main timings, sub timings are same as above.

View attachment 717480
Click to expand...
What brand and model # of memory you using?
 
Ski said:
What brand and model # of memory you using?
Click to expand...
T-Create from teamgroup. I bought them back in november (pre-tarif panic buy) and had them along with my x870E just sitting collecting dust till these cpu's came out. Was expecting a January release, that didn't pan out =P.

They are 6800 mhz hynix modules @ 36-36-36-84 M-dies. Just tightened the timings and running them at 6000mhz for now. Actually bought two, but ended up returning one after messing with a 9800X3D on a test bench with the board and ram. Yea, buildzoid was right, the four memory slots on AM5 boards is a design flaw, not a feature...haha. Couldn't get any respectable timings or frequency, so backed down to one kit and returned the other to Amazon.

Tinkering around with overclocking now. Got a -20 offset per ccd on right now testing. Going to switch over to curve shaper in a little bit once I play around a bit more with Curve offset. Have to say, not having experienced overclocking on the zen platform...This is a breeze and has been a pleasant experience so far. AMD's overclocking toolset is truly damn good on zen 5.
 
Last edited:
  • Like
Reactions: Ski
like this
blurp said:
Got a 9950x3D. Following this video and the overclock.net thread. I set it with PBO + 200. Scalar x 10. CO -20 so far like you. The max power I saw was 226 W. Temps Mid 80s with Arctic Freezer 280. No thermal throttling.
Click to expand...
10x scalar shouldn't be used for daily settings. And in my experience, x2 is enough to (when combined with a decent negative CO offset and good cooling) hold the all core boost at or close to the 1 and 2 thread boost speed. And even that really isn't necessary for daily use. It makes your cinibench numbers go up. But, it will have marginal or no effect in gaming and even most productivity workloads.
 
Ikasu said:
T-Create from teamgroup. I bought them back in november (pre-tarif panic buy) and had them along with my x870E just sitting collecting dust till these cpu's came out. Was expecting a January release, that didn't pan out =P.

They are 6800 mhz hynix modules @ 36-36-36-84 M-dies. Just tightened the timings and running them at 6000mhz for now. Actually bought two, but ended up returning one after messing with a 9800X3D on a test bench with the board and ram. Yea, buildzoid was right, the four memory slots on AM5 boards is a design flaw, not a feature...haha. Couldn't get any respectable timings or frequency, so backed down to one kit and returned the other to Amazon.

Tinkering around with overclocking now. Got a -20 offset per ccd on right now testing. Going to switch over to curve shaper in a little bit once I play around a bit more with Curve offset. Have to say, not having experienced overclocking on the zen platform...This is a breeze and has been a pleasant experience so far. AMD's overclocking toolset is truly damn good on zen 5.
Click to expand...
Keep updating your progress cuz I got a similar spec'd machine I'm about to build and it's been years since I've overclocked a machine so your information will be invaluable for me. Thanks dude.
 
chameleoneel said:
10x scalar shouldn't be used for daily settings. And in my experience, x2 is enough to (when combined with a decent negative CO offset and good cooling) hold the all core boost at or close to the 1 and 2 thread boost speed. And even that really isn't necessary for daily use. It makes your cinibench numbers go up. But, it will have marginal or no effect in gaming and even most productivity workloads.
Click to expand...
Those settings come from SkatterBencher who suggested it had minimal voltage increase. I'll test with lower Scalar next week when I get back home and report back.
 
Ski said:
Keep updating your progress cuz I got a similar spec'd machine I'm about to build and it's been years since I've overclocked a machine so your information will be invaluable for me. Thanks dude.
Click to expand...
Will do. Honestly I'm shocked at how easy it's been so far on oc'ing this platform. Has come a long way from what I had to do in the past...If you haven't overclocked in a while (like me), you'll also be pleasantly surprised. Def check out that skatterbencher #85 that was linked, or the youtube video variant of it. It's so incredibly helpful to understand the OC platform for am5...Pretty much giving you the cookie cutter guides for oc'ing. Also help she's using an Asus board as well for me =P.
 
  • Like
Reactions: Ski
like this
chameleoneel said:
10x scalar shouldn't be used for daily settings. And in my experience, x2 is enough to (when combined with a decent negative CO offset and good cooling) hold the all core boost at or close to the 1 and 2 thread boost speed. And even that really isn't necessary for daily use. It makes your cinibench numbers go up. But, it will have marginal or no effect in gaming and even most productivity workloads.
Click to expand...
What's your recommendation then that's considered safe but still within acceptable performance for heavy workstation use.
 
Ikasu said:
Will do. Honestly I'm shocked at how easy it's been so far on oc'ing this platform. Has come a long way from what I had to do in the past...If you haven't overclocked in a while (like me), you'll also be pleasantly surprised. Def check out that skatterbencher #85 that was linked, or the youtube video variant of it. It's so incredibly helpful to understand the OC platform for am5...Pretty much giving you the cookie cutter guides for oc'ing. Also help she's using an Asus board as well for me =P.
Click to expand...
Will do thanks for the link brother.
 
Yea. I'm going to stay away from the scalar option. This is going to be a 24/7 system, so if I can avoid any type of chip degradation the better. For my uses, Curve shaper and f max will be enough for me.
 
Ski said:
What's your recommendation then that's considered safe but still within acceptable performance for heavy workstation use.
Click to expand...
If you have decent or good cooling: Quick settings would be to activate PBO, set the limits to "manual" to make sure you are on defaults and not whatever the mobo maker thinks is appropriate.
+200 of frequency
-20 for Curve optimizer. and 2x on scalar.

Much higher than -25 CO and I would then start doing per-core CO settings. and find your weakest cores, set them with a lower offset number. I think per-core may be particularly useful for Zen5, due to the higher boosting, etc. Zen 5 seems to pull relatively more power with PBO active, than Zen 4 does.

--------------------------------

I haven't personally used a 9800X3D. I have used a 7700X and a 7800X3D.

7800X3D is incredibly efficient. So is Zen 4, in general.
For 7800X3D, Those settings above and really average average cooling, is enough for it to make all core boost match single core boost.
For non-X3D, Zen4 parts responded really well to Eco mode. Only losing a small amount of performance. You can then activate PBO and Curve Optmizer to gain back the very small performance loss. and still be using less power and heat, than stock.

Another easy way to help your CO setting to result in lower temps and power use, is to not do +200 on the clockspeed.

------------------------

Zen 5 X3D boost higher and use more power to do so. And the architecture in general, seems a bit more power hungry at higher utilizations. Eco mode seems to drop relatively more performance. But, I would still be doing pretty much the same PBO settings.


This optimum tech video is a great little functional workshop on ways to leverage PBO

View: https://www.youtube.com/watch?v=FaOYYHNGlLskk
 
  • Like
Reactions: Ski
like this
chameleoneel said:
If you have decent or good cooling: Quick settings would be to activate PBO, set the limits to "manual" to make sure you are on defaults and not whatever the mobo maker thinks is appropriate.
+200 of frequency
-20 for Curve optimizer. and 2x on scalar.

Much higher than -25 CO and I would then start doing per-core CO settings. and find your weakest cores, set them with a lower offset number. I think per-core may be particularly useful for Zen5, due to the higher boosting, etc. Zen 5 seems to pull relatively more power with PBO active, than Zen 4 does.

--------------------------------

I haven't personally used a 9800X3D. I have used a 7700X and a 7800X3D.

7800X3D is incredibly efficient. So is Zen 4, in general.
For 7800X3D, Those settings above and really average average cooling, is enough for it to make all core boost match single core boost.
For non-X3D, Zen4 parts responded really well to Eco mode. Only losing a small amount of performance. You can then activate PBO and Curve Optmizer to gain back the very small performance loss. and still be using less power and heat, than stock.

Another easy way to help your CO setting to result in lower temps and power use, is to not do +200 on the clockspeed.

------------------------

Zen 5 X3D boost higher and use more power to do so. And the architecture in general, seems a bit more power hungry at higher utilizations. Eco mode seems to drop relatively more performance. But, I would still be doing pretty much the same PBO settings.


This optimum tech video is a great little functional workshop on ways to leverage PBO

View: https://www.youtube.com/watch?v=FaOYYHNGlLskk
Click to expand...

Just for context this is my workstation I'm currently building and it's primarily a 3d modeling/rendering workstation (3dx Max & V-Ray) gaming will be an afterthought but once ina blue moon type of deal:
• 9950x 3D
• ASUS x870e Crosshair Hero
• ASUS TUF 3090 (plan to upgrade down the pipeline when I sell my kidney)
• EK Nucleus CR360 Dark AIO
• G.Skill Trident 96GB RAM DDR5 6400 32-28-28-102

I definintely intend to overclock because the programs I'm utizilizing will benefit for with the higher speeds and time is money for me these days. And unlike my previous build memory was such a huge problem because 4 sticks of RAM as you well know don't play nice when it comes to overclock so I'm trying to keep my specs within acceptable parameters. Plus it's hard to justice paying $800+ for RAM these days. Appreciate your follow-up by the way.
 
Ski said:
Just for context this is my workstation I'm currently building and it's primarily a 3d modeling/rendering workstation (3dx Max & V-Ray) gaming will be an afterthought but once ina blue moon type of deal:
• 9950x 3D
• ASUS x870e Crosshair Hero
• ASUS TUF 3090 (plan to upgrade down the pipeline when I sell my kidney)
• EK Nucleus CR360 Dark AIO
• G.Skill Trident 96GB RAM DDR5 6400 32-28-28-102

I definintely intend to overclock because the programs I'm utizilizing will benefit for with the higher speeds and time is money for me these days. And unlike my previous build memory was such a huge problem because 4 sticks of RAM as you well know don't play nice when it comes to overclock so I'm trying to keep my specs within acceptable parameters. Plus it's hard to justice paying $800+ for RAM these days. Appreciate your follow-up by the way.
Click to expand...
We literally have very similar uses. Although I'm using Maya/Arnold, as well as zbrush/painter for character creation among other software in the toolbox.

Same cpu, similar board, I have a 3090 atm (strix waterblocked), but have a 5090 to toss in. Using a Mora for my cooling though, and 96gb of T-Create ram. I'm not sure about your specific memory...But 100% make sure it's hynix, do not mess around with anything else. It's a huge factor for AMD if you want to tighten your timings and decrease memory latency. Also don't bank on running 6400mhz though. Pretty much luck of the draw. I unfortunately drew poorly with my 9950x3d in the lotto =*(. Buildzoid had it right, it's better to just nab a 6000mhz memory set and push it further, rather than buying higher and have to tighten the timings on your own. Although that is what I ended up doing...lol. If you get a hynix kit with very tight timings, would make your life easier.

Being a fellow 3d user, I'd def suggest curve shaper and just tightening your timings. Considering the constant render barrage of usage and multicore workloads, I'd leave the Scalar option on the bench. But that's just me. Def on the right track with 2x48gb. When I tested 4x48gb, the speed penalty was massive. I couldn't even get 5200mhz stable on a 9800X3D on the test bench. Memory latency was high as well. As much as I want to have 192gb and not have to worry about being ram amount limited, the latency and speed benefit for 2x48 is going to be more beneficial for the actual workflow and creation process.
 
  • Like
Reactions: Ski
like this
Ikasu said:
We literally have very similar uses. Although I'm using Maya/Arnold, as well as zbrush/painter for character creation among other software in the toolbox.

Same cpu, similar board, I have a 3090 atm (strix waterblocked), but have a 5090 to toss in. Using a Mora for my cooling though, and 96gb of T-Create ram. I'm not sure about your specific memory...But 100% make sure it's hynix, do not mess around with anything else. It's a huge factor for AMD if you want to tighten your timings and decrease memory latency. Also don't bank on running 6400mhz though. Pretty much luck of the draw. I unfortunately drew poorly with my 9950x3d in the lotto =*(. Buildzoid had it right, it's better to just nab a 6000mhz memory set and push it further, rather than buying higher and have to tighten the timings on your own. Although that is what I ended up doing...lol. If you get a hynix kit with very tight timings, would make your life easier.

Being a fellow 3d user, I'd def suggest curve shaper and just tightening your timings. Considering the constant render barrage of usage and multicore workloads, I'd leave the Scalar option on the bench. But that's just me. Def on the right track with 2x48gb. When I tested 4x48gb, the speed penalty was massive. I couldn't even get 5200mhz stable on a 9800X3D on the test bench. Memory latency was high as well. As much as I want to have 192gb and not have to worry about being ram amount limited, the latency and speed benefit for 2x48 is going to be more beneficial for the actual workflow and creation process.
Click to expand...

Yea they're hynix ((F5-6400J3239F48GX2-TZ5RK).) did my hoemwork on that one. I was skeptical of getting the Trident Royal series which has a CL28 at 6000mhz, especially the additional $100 cost. However, I might change my tune depending on how this memory performs. Thanks again btw :).
 
Ski said:
Yea they're hynix ((F5-6400J3239F48GX2-TZ5RK).) did my hoemwork on that one. I was skeptical of getting the Trident Royal series which has a CL28 at 6000mhz, especially the additional $100 cost. However, I might change my tune depending on how this memory performs. Thanks again btw :).
Click to expand...
I use to be all in the aesthetics of my build in prior rigs. But now I don't really care as much. I mean heck, I have a giant mora rad sitting next to my pc, and have quick disconnect fittings on the outside and inside with lots of tubing about...Want to be able to remove gpu or cpu without draining, and upgrade when necessary. Which is indeed beautiful for ease of maintenance and upgrading.

Not sure if you got the tridentz rgb's due to aesthetics and rgb, just keep in mind rgb will needlessly heat up your memory as a result. Would def recommend installing the gskill RGB software and turning them off. Especially If you plan to OC them and tighten your timings as much as possible, they will add heat needlessly to your modules which can limit your headroom, or even cause errors due to high memory temps as dual rank dimms run hot. If you put a fan over it though, should alleviate that problem a bit....but takes away the point of having an RGB set if you are going for an aesthetic build. I personally picked up a 90 degree fan mount from aliexpress that mounts to another fan, so plan to put either a 120mm or 140mm fan over my memory modules to provide active cooling once I get this new hardware into my main rig and off the test bench. Want to ensure I have no issues whatsoever temp wise, although a bit overkill considering I'm water blocking my gpu...If it was air cooled and dumping that heat during gpu rendering through the passthrough gpu's are using now-a-days, would of cooked my modules leading to a higher likelihood of errors....Me no likey.
 
Ikasu said:
I use to be all in the aesthetics of my build in prior rigs. But now I don't really care as much. I mean heck, I have a giant mora rad sitting next to my pc, and have quick disconnect fittings on the outside and inside with lots of tubing about...Want to be able to remove gpu or cpu without draining, and upgrade when necessary. Which is indeed beautiful for ease of maintenance and upgrading.

Not sure if you got the tridentz rgb's due to aesthetics and rgb, just keep in mind rgb will needlessly heat up your memory as a result. Would def recommend installing the gskill RGB software and turning them off. Especially If you plan to OC them and tighten your timings as much as possible, they will add heat needlessly to your modules which can limit your headroom, or even cause errors due to high memory temps as dual rank dimms run hot. If you put a fan over it though, should alleviate that problem a bit....but takes away the point of having an RGB set if you are going for an aesthetic build. I personally picked up a 90 degree fan mount from aliexpress that mounts to another fan, so plan to put either a 120mm or 140mm fan over my memory modules to provide active cooling once I get this new hardware into my main rig and off the test bench. Want to ensure I have no issues whatsoever temp wise, although a bit overkill considering I'm water blocking my gpu...If it was air cooled and dumping that heat during gpu rendering through the passthrough gpu's are using now-a-days, would of cooked my modules leading to a higher likelihood of errors....Me no likey.
Click to expand...
I absolutely loathe RGB, but between the reviews, compatibility from the QVR, and performance, those were my only options... plus it was on sale so I had to try it out. Again I have 30 days to play with them and see what happens. I'll add that i got a Fractal North XL so I might rig the side fans to hit the memory.
 
Ski said:
Just for context this is my workstation I'm currently building and it's primarily a 3d modeling/rendering workstation (3dx Max & V-Ray) gaming will be an afterthought but once ina blue moon type of deal:
• 9950x 3D
• ASUS x870e Crosshair Hero
• ASUS TUF 3090 (plan to upgrade down the pipeline when I sell my kidney)
• EK Nucleus CR360 Dark AIO
• G.Skill Trident 96GB RAM DDR5 6400 32-28-28-102

I definintely intend to overclock because the programs I'm utizilizing will benefit for with the higher speeds and time is money for me these days. And unlike my previous build memory was such a huge problem because 4 sticks of RAM as you well know don't play nice when it comes to overclock so I'm trying to keep my specs within acceptable parameters. Plus it's hard to justice paying $800+ for RAM these days. Appreciate your follow-up by the way.
Click to expand...
I don't think Buildzoid has played with 48GB Dimms. But he has done 32GB dimms. Some of his approach to 32GB dimms may apply, since they are both dual rank:


View: https://www.youtube.com/watch?v=wnDEbqpvZvY

Also, here's a recent video where he goes over various timings for AM5


View: https://www.youtube.com/watch?v=sDxkous2Ua4
 
Where I'm at, atm....

memory.jpg


So far, stable. Unfortunately as stated previously. Can't push the mclk/uclk very high on my 9950x3d. Errors at 1.28vsoc for 3100 mclk (6200mhz), and 2067 fclk. Stable though at 1.3...but really don't want to risk a 24/7 with a 1.3 vsoc. So think I'm just going to stay with 3000/2000. Definitely did NOT win the lotto on the IMC...pretty atrocious honestly. at 1.15 vsoc I got p95 errors an hour in...and 1.18 though stable. I would normally tighten as much as I can, but doubt it's worth the time to tighten the 1.18v down further, considering it's well within a safe range already.

Going to start decreasing memory voltages to see how low I can go and maintain these timings, rather then bumping up voltage and tightening further. Seems good enough for me..

In regards to CPU. Got a curve shaper going and a fmax boost of 200 mhz stable from tests so far. Min freq. 0, low freq. - 30, medium freq. -30, high. freq -25, max freq. -10. So far so good, passing all the tests I've thrown at it. Been pretty impressive so far. Nice reduction in wattage/temps, and boost in performance. Ran asus realbench with its image based benchmark, and cpu sits consistently around 5.9ghz for the single threaded usage from the looks of it. Although when running a single thread of p95, or ycruncher, boosts to 5.9, but then drops to 5.7....not sure if that's normal for this or not...First time OC'ing on an AMD platform with Zen.
 
Ikasu said:
Where I'm at, atm....

So far, stable. Unfortunately as stated previously. Can't push the mclk/uclk very high on my 9950x3d. Errors at 1.28vsoc for 3100 mclk (6200mhz), and 2067 fclk. Stable though at 1.3...but really don't want to risk a 24/7 with a 1.3 vsoc. So think I'm just going to stay with 3000/2000. Definitely did NOT win the lotto on the IMC...pretty atrocious honestly.
Click to expand...
Eh. you're running 96GB of RAM. I would have guessed 6400 would be out of the question. 1.3 is safe for daily. If you can get 6200m stable, I would say that's a win. But I also don't think 6200 is that beneficial over 6000 on Ryzen. Especially with X3D.
X3D aren't as RAM sensitive.
 
chameleoneel said:
Eh. you're running 96GB of RAM. I would have guessed 6400 would be out of the question. 1.3 is safe for daily. If you can get 6200m stable, I would say that's a win. But I also don't think 6200 is that beneficial over 6000 on Ryzen. Especially with X3D.
X3D aren't as RAM sensitive.
Click to expand...

Considering Buildzoid is running his in that video at 6400/2133 at 1.22 vsoc...It's def possible if you get a good chip. Doesn't seem the difference in single/dual rank matters as much for this factor. I just got bad luck on the draw from the looks of it. System's main purpose is a 3d workstation, so any performance I can milk out of it is definitely a plus, but chip degradation weighs on my mind always. Also AMD did reduce and put a limit on vsoc voltage during the 7000 series as there were burnout issues. Although 7000X3D didn't have the advantage of the shifted v-cache design that 9000 has had...So it may indeed be safe for 1.3...but still weighs on my mind.
 
Last edited:
Ikasu said:
Considering Buildzoid is running his in that video at 6400/2133 at 1.22 vsoc...It's def possible if you get a good chip. Doesn't seem the difference in single/dual rank matters as much for this factor. I just got bad luck on the draw from the looks of it. System's main purpose is a 3d workstation, so any performance I can milk out of it is definitely a plus, but chip degradation weighs on my mind always. Also AMD did reduce and put a limit on vsoc voltage during the 7000 series as there were burnout issues. Although 7000X3D didn't have the advantage of the shifted v-cache design that 9000 has had...So it may indeed be safe for 1.3...but still weighs on my mind.
Click to expand...
Yeah BZ has a golden sample.

6400 1:1 isn't that common, even when using 16GB single rank sticks. Doing it with 48GB dual rank sticks at 1.22 SOC voltage is unusual.

1.3v is AMD's safe upper limit. If you can get 6400 stable at 1.3 and it noticeably speeds up your workflow, I would run it without worry.
 
chameleoneel said:
Yeah BZ has a golden sample.

6400 1:1 isn't that common, even when using 16GB single rank sticks. Doing it with 48GB dual rank sticks at 1.22 SOC voltage is unusual.

1.3v is AMD's safe upper limit. If you can get 6400 stable at 1.3 and it noticeably speeds up your workflow, I would run it without worry.
Click to expand...
I wish I could get 6400 at that. Unfortunately I get 6200/2066 at 1.3v =*(.
 
Wait, you can run dual rank at 6000MT/s on these CPUs? My 7800X3D could not swing it. I guess it's time to buy more ram then...
 
Deimos said:
Wait, you can run dual rank at 6000MT/s on these CPUs? My 7800X3D could not swing it. I guess it's time to buy more ram then...
Click to expand...
As long as you're running two sticks, not a problem.
 
Deimos said:
Wait, you can run dual rank at 6000MT/s on these CPUs? My 7800X3D could not swing it. I guess it's time to buy more ram then...
Click to expand...

Ikasu said:
As long as you're running two sticks, not a problem.
Click to expand...
Yeah for clarity to all:

At this point in time, all 48GB sticks should be dual rank.

The majority of 32GB sticks should be dual rank. Although I believe I saw some news sometime in the past year, that single rank 32GB sticks could be introduced soon. So, that could become a YMMV.

All 24GB or 16GB or 8GB, are single rank.
 
Deimos said:
Alright, I'm going to go for a 6000MT/s kit with tighter timings. Just need to check around for the best pricing...
Click to expand...
It's still dependent on your chips IMC, and motherboard. But 6000 is pretty much a guarantee now-a-days.

If you do plan to pick up some new memory, make sure it's Hynix A or M die based.
 
Ikasu said:
It's still dependent on your chips IMC, and motherboard. But 6000 is pretty much a guarantee now-a-days.

If you do plan to pick up some new memory, make sure it's Hynix A or M die based.
Click to expand...
Yeah I wish the memory manufactures would make a little more obvious, I'll pop into my local store tomorrow and check the serial number. Unfortunately they only have 1 in stock at my closest branch.
 
Has anyone here with a 9950X3D and a 4090 ran Timespy, port royal or anything of the likes for comparison?

I've been debating finally moving on from my 5950X, but comparing online, I'm still scoring HIGHER with my 4090 and 5950X than people with a 4090 and a 9950X3D. I figured being a [H] forum, maybe some people here have tweaked beyond "stock" PBO boosting and have some numbers to compare with.

For reference... this is what I am snagging with my 5950X setup:
http://www.3dmark.com/spy/51837341
 
III_Slyflyer_III said:
Has anyone here with a 9950X3D and a 4090 ran Timespy, port royal or anything of the likes for comparison?

I've been debating finally moving on from my 5950X, but comparing online, I'm still scoring HIGHER with my 4090 and 5950X than people with a 4090 and a 9950X3D. I figured being a [H] forum, maybe some people here have tweaked beyond "stock" PBO boosting and have some numbers to compare with.

For reference... this is what I am snagging with my 5950X setup:
http://www.3dmark.com/spy/51837341
Click to expand...
As someone who has a 5950x...I find these claims literally impossible. You have to have looked at wrong information or you're smoking something =P. Just scroll down on this page, as they are direct comparisons using the same gpu...They even date the data...3/25 being march this year. So apples to apples. Also real world is more important than 3dmark.

https://gamersnexus.net/cpus/amd-ryzen-9-9950x3d-cpu-review-benchmarks-vs-9800x3d-285k-9950x-more
 
Ikasu said:
As someone who has a 5950x...I find these claims literally impossible. You have to have looked at wrong information or you're smoking something =P. Just scroll down on this page, as they are direct comparisons using the same gpu...They even date the data...3/25 being march this year. So apples to apples. Also real world is more important than 3dmark.

https://gamersnexus.net/cpus/amd-ryzen-9-9950x3d-cpu-review-benchmarks-vs-9800x3d-285k-9950x-more
Click to expand...
I agree, real world is always better, but most comparisons are also using potato settings. Who TF with this kind of hardware is gaming on potato settings? I know WHY they do it, but I really want some good "Max Settings" / 4K tests as well. I have found a few to compare, but I am specifically looking for 3dmark scores as well from people who may be pushing past default PBO settings (enthusiasts).

Also, click my link and compare with 9950X3D tests using a 4090, I'm legit right up there with a 5950X. Trying to figure out how close max tweaking and OC'ing has helped overall.

Here is a search for just 9950X3D and 4090: https://www.3dmark.com/search?_ga=2...ck=&maxGpuMemClock=&minCpuClock=&maxCpuClock=
 
You must log in or register to reply here.
Back
Top