• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

AMD Ryzen 7 9800X3D

As stooeh noted, I'm giving temps above ambient, not absolute. Fully idle for me is about 40 degrees absolute, 45-50 absolute in light tasks, about 75 under heavy load and up to about 80 under synthetic testing.

I see, so this is (very similar to) throwing a bit of old-fashion FSB overclocking into the mix and forcing the CPU to run at am (up to 44.5x) multiple of the ECLK?
Pretty much, but you can manipulate the V/F curve at specific points, for specific types of loads, based on MIN/MED/MAX Temp, Frequency etc. It's a lot to unpack at first, but can be very useful once you get the hang of it.
 
Pretty much, but you can manipulate the V/F curve at specific points, for specific types of loads, based on MIN/MED/MAX Temp, Frequency etc. It's a lot to unpack at first, but can be very useful once you get the hang of it.
I guess that because I'm asking the processor to up frequencies for a given ECLK multiplier, the basics are I'll need to reduce the negative Curve Offset or Slope. Which is what that portion of the video seems to be saying (cheers for sharing at that timestamp). So perhaps I'll dial back -25 CO to -10 or even 0 and see where I can get with ECLK.

It's intuitive how curve shaping and offset would fit in. Also upping permissible TDP etc. This reminds me of the good old LGA-775 days :)
 
Last edited:
Would the 6000-CL26 RAM kits be any good for people don't want to OC/finetune?

G.SKILL Releases Low Latency DDR5-6000 CL26 & CL28 Memory Kits
There's really no point. Cas Latency is not important on AM5 compared to intel setups.

There is a cookie cutter finetune guide which works for all A/M hynix based mem. It will lower your latency and improve on memory performance. Hell even I just copy and pasted and it worked for me and I never muck about with secondary timings.
 
Last edited:
There's really no point. Cas Latency is not important on AM5 compared to intel setups.

There is a cookie cutter finetune guide which works for all A/M hynix based mem. It will lower your latency and improve on memory performance. Hell even I just copy and pasted and it worked for me and I never muck about with secondary timings.
Got a link?
 

This is the one I used.

@LtMatt so after seeing your post about the curve shaper (vf curve?) I went about mucking about with it. I must have done some wrong as it no longer downclocks to low speeds. So it either runs at 4850 (idle!) or 5600 :P
Yes. I believe I noticed similar. At least, certainly higher clock speeds than normal. However, it doesn't seem to cause an issue as far as I can tell. Still figuring it out myself. tbf.
 
My 9800X3D has been up and running since boxing day without any issues so happy to say its all ok at stock.

What would be the best settings to use for PBO and Curve Optimiser?

Got a Gigabyte X870 Eagle, 32GB of C30 ram and the 980X3D is being cooled by an EK AIO (240mm Rad)
 
My 9800X3D has been up and running since boxing day without any issues so happy to say its all ok at stock.

What would be the best settings to use for PBO and Curve Optimiser?

Got a Gigabyte X870 Eagle, 32GB of C30 ram and the 980X3D is being cooled by an EK AIO (240mm Rad)
If you want an 'easy overclock' then try a PBO Curve Offset, all core, of negative 20, along with a max boost of +200. This should be stable on most chips; if not try setting Load-Line-Calibration to Level 4 or Level 3 (LLC does vary by manufacturer, my MSI has eight levels). That should boost performance to ~5.3 on all cores, and ~5.4 on 1-2 cores. To test stability, I recommend y-cruncher (option 2 then option zero) to perform a stress test of components for 2-3 cycles. Then follow it up with OCCT variable for 1 hour.

If you want, you can then play around with further reductions in PBO Curve Offset; I get around 5.4 on all cores using PBO: https://valid.x86.fr/4ymb32

There are also more advanced methods like LTMatt mentions here: https://forums.overclockers.co.uk/threads/amd-ryzen-7-9800x3d.18994330/page-128#post-37560090
 
Last edited:
do you guys think the 9900x3d will have a significant enough uplift over the 9800x3d in gaming?
Nah, at least not for now. Might be a bit more futureproofed if you don't plan to upgrade for a long time though.

Few games use more than eight cores, almost none really need more than eight. The 9900 will also only have six cores with 3D cache (although therefore more cores per cache) meaning there are a few titles it'll be a bit slower in. Alongside the thermal limitations this means it'll lag behind the 9800 X3D in games for the next couple of years.
 
Last edited:
Nah, at least not for now. Might be a bit more futureproofed if you don't plan to upgrade for a long time though.

Few games use more than eight cores, almost none really need more than eight. The 9900 will also only have six cores with 3D cache (although therefore more cores per cache) meaning there are a few titles it'll be a bit slower in. Alongside the thermal limitations this means it'll lag behind the 9800 X3D in games for the next couple of years.

You obvs haven't played Path of Exile 2. That will 100% any 8 core:cry:
 
You obvs haven't played Path of Exile 2. That will 100% any 8 core:cry:
In fair I did say almost none. There will be exceptions, but the 9800 X3D should outperform the 9900 X3D on average. Although with the X3D cores now being (almost) as fast as the normal cores, improving scheduling and with higher overall boost clocks (versus its likely power and thermal limitations) the 9950 X3D will be interesting.
 
Last edited:
Back
Top Bottom