Overclock.net banner
2,481 - 2,500 of 10,009 Posts
But I can't get above 24k with pbo +200 5425 at all. Nowhere near. It's almost always around 23800. Even static 5450 doesn't scale at all in R23. I always have low scores in R23. Could it be because I'm using windows store version or something dumb?
I noticed a stronger score when only using PBO and curve optimizer, that eCLK drops the cores further down coming in from a distance
 
I noticed a stronger score when only using PBO and curve optimizer, that eCLK drops the cores further down coming in from a distance
Yeah that is what I tried to explain. Eclk drops it even further unless you drop PBO fMax down a bit. So PBO +200 + eCLK should be slower than PBO +0 + more eCLK. That's just how it chooses a multiplier against the v/f curve. It makes sense. eCLK makes it run a higher frequency than the CPU thinks it is so it wants to drop boost bins down to for example 51.5 multi so it does that but actual frequency stays higher with eCLK so you can trick it into boosting higher.

@donnieyeen the BIOS is freshly flashed about 3 days ago, only things touched are RAM + FCLK all manual + SVM, TSME, SMEE, data scramble, Thunderbolt, SATA Controller, iGPU, ASUS armoury crate installer disabled. Secure boot is set to Windows OS and enabled.

I'm also running a relatively fresh Windows only like 4 weeks old, W10 22H2 fully updated, AMD chipset drivers are installed with 6.10.17.152 version, game launchers and such all disabled from auto start. Only things on auto start on boot are MSI AB, HWINFO64, FanControl, nVidia app, OneDrive, ASUS Aura (standalone).

I just did a run on PBO +200 -30 CO with the above config and I got 23836. Exactly what I'd expect my system to do. It did not hold full clocks at all, bouncing between 5350 and 5390 mostly. Temps are totally fine 75.5c max tdie cores in the high 60's.

Could it be chipset drivers maybe?? What version if any are you guys running?
 
@Arizor what CPU-Z scores could I expect for example? Or LinPack GFLOPS?
 
Yeah that is what I tried to explain. Eclk drops it even further unless you drop PBO fMax down a bit. So PBO +200 + eCLK should be slower than PBO +0 + more eCLK. That's just how it chooses a multiplier against the v/f curve. It makes sense. eCLK makes it run a higher frequency than the CPU thinks it is so it wants to drop boost bins down to for example 51.5 multi so it does that but actual frequency stays higher with eCLK so you can trick it into boosting higher.

@donnieyeen the BIOS is freshly flashed about 3 days ago, only things touched are RAM + FCLK all manual + SVM, TSME, SMEE, data scramble, Thunderbolt, SATA Controller, iGPU, ASUS armoury crate installer disabled. Secure boot is set to Windows OS and enabled.

I'm also running a relatively fresh Windows only like 4 weeks old, W10 22H2 fully updated, AMD chipset drivers are installed with 6.10.17.152 version, game launchers and such all disabled from auto start. Only things on auto start on boot are MSI AB, HWINFO64, FanControl, nVidia app, OneDrive, ASUS Aura (standalone).

I just did a run on PBO +200 -30 CO with the above config and I got 23836. Exactly what I'd expect my system to do. It did not hold full clocks at all, bouncing between 5350 and 5390 mostly. Temps are totally fine 75.5c max tdie cores in the high 60's.

Could it be chipset drivers maybe?? What version if any are you guys running?
I wouldn’t worry about it @Imprezzion , I have the same thing with CB23, it’s just weird behaviour, probably something going on with that specific app and our comps.

Just try other benchmarks and make sure they’re what you expect.
I don't think its 'weird behaviour', clocks are simply not hit in multi core load. I can tell you I also had such a sample out of 5 I tested which would also not hit 24000 with pbo -30, was simply a trash can sample and it went for the sale. You can check doms binning also some were simply trash.. happens. If everything else was checked that is as user states.
 
It's basically identical SP to mine lol. Same amount of cores at 112 113 and 114 just different cores. I just also did a CPU multi core Cinebench 2024 run and I can see why it's more consistent maybe. The load is much much lower than R23. Temps and power draw are significantly lower in 2024 so it allows it to boost more consistently and higher.

@Arizor I am at 105.3 eCLK + PBO Enabled motherboard limits, 0 (disabled) fMax, 0 CO (auto) which is 5501.9Mhz all core clocks and it ran right at 1.217v which is what I want it to be at 75c tdie high 60's low 70's for the cores at ~136w average. Score was 1454 so right what I was expecting with what you told me.
 
  • Rep+
Reactions: Arizor
It's basically identical SP to mine lol. Same amount of cores at 112 113 and 114 just different cores. I just also did a CPU multi core Cinebench 2024 run and I can see why it's more consistent maybe. The load is much much lower than R23. Temps and power draw are significantly lower in 2024 so it allows it to boost more consistently and higher.

@Arizor I am at 105.3 eCLK + PBO Enabled motherboard limits, 0 (disabled) fMax, 0 CO (auto) which is 5501.9Mhz all core clocks and it ran right at 1.217v which is what I want it to be at 75c tdie high 60's low 70's for the cores at ~136w average. Score was 1454 so right what I was expecting with what you told me.
I just run cinebench r23 with pbo -30 no limits 10x scalar fmax 200 and I'm under 24k

23865k very similar to yours
 
I just saved a CMO file for my motherboard and it won't let me load it with a new CPU installed Duh :(. Lost all my ram timing settings and my training bios settings -_-
When I have stable settings, I make photos with the cellphone of every page. Its quick and fast and you can get your settings even if the PC doesnt work.
 
I just saved a CMO file for my motherboard and it won't let me load it with a new CPU installed Duh :(. Lost all my ram timing settings and my training bios settings -_-
That's why i always take photos to phone before flashing a new bios to get back easily my earlier cpu/ram settings.
 
It's basically identical SP to mine lol. Same amount of cores at 112 113 and 114 just different cores. I just also did a CPU multi core Cinebench 2024 run and I can see why it's more consistent maybe. The load is much much lower than R23. Temps and power draw are significantly lower in 2024 so it allows it to boost more consistently and higher.

@Arizor I am at 105.3 eCLK + PBO Enabled motherboard limits, 0 (disabled) fMax, 0 CO (auto) which is 5501.9Mhz all core clocks and it ran right at 1.217v which is what I want it to be at 75c tdie high 60's low 70's for the cores at ~136w average. Score was 1454 so right what I was expecting with what you told me.
SP score is pretty much irrelevant, its a gimmick, you should not give it any much attention. We've seen dk_mic have 116 worse then 113 or something like that.. pretty much 4 of 5 my samples hit 24500+ on pbo -20\-30 in cb23. If the cpu doesn't hit that, and there's no weird things going on in Windows, the scores sub 23? then I would return it\swap it.
 
Hi Gentlemen. I have my 9800x3D and it is behaving weird. My effective clocks are always lower than regular clock. On my CPU clocks are always -25 MHz lower. I noticed that it is related to power draw. If power draw is bigger, effective clocks go down. I think this is not a bug but just poor binned CPUs which cant hit boost clock effectively. Should I return/swap my CPU?
 
Finally seems like the first couple batches of 9800x3d are going out in canada. Got mine a few days ago. Got the system up and running last night. It's an impressive little 8 core cpu for gamming even out of the box.

But of course now is time to do some tunning and tweaking :). Is the skatterbench guide a good starting point to get started???


One other question i see alot of back and forth on scalar. Some say it will degrade the cpu, other says it's pretty minor in the voltage increase even at x10. Some say it helps hold higher peak boost clocks more longer and stable. So what the community consensus at the time on it?
 
2,481 - 2,500 of 10,009 Posts