AMD Ryzen 7 9800X3D

[IceWolf]

Interesting. I shall try it and find out.

The big Ryzen 7000 Memory and OC Tuning Guide - Infinity Fabric, EXPO, Dual-Rank, Samsung and Hynix DDR5 in Practice test with Benchmarks and Recommendations | Page 9 | igor´sLAB

AMD’s new Ryzen 7000 desktop CPUs, based on the Zen4 micro-architecture, still use the same chiplet design as their predecessors, with a few small but not negligible changes. Igor already gave details…

www.igorslab.de

Ok so I was wrong about FCLK: if your motherboard runs it higher than 2000 without instability you are fine, on DDR5-6400 memory 2133 FCLK can be 1% faster than the same settings at 2000 FCLK (although on X3D chips the margins are smaller with all memory optimisations).

What IS important is keeping UCLK and MemCLK in a 1:1 ratio. Since 6000 CL30 was within 1% of the best possible configuration and is supposed to work out of the box, I remembered those settings, which are 2000 FCLK, 3000 UCLK and MemCLK i.e. a 2:3 ratio. If you encounter any stability problems I would lower FCLK manually back to 2033 or 2000 and leave UCLK and MemCLK as they are: 3000 for DDR5-6000 and 3200 for DDR5-6400 memory. If all runs well then the higher FCLK is not a problem being out of sync from the 2:3 ratio.

[Edit] That being said, the Assetto Corsa 1080p results in that link to Igor's Lab show DDR5-6400 at 2133 FCLK being 1-2% faster than the same settings with 2167 FCLK, with 2000 FCLK falling in between. So maybe the 2:3 ratio DOES give a tiny benefit?

 

[matty007]

The big Ryzen 7000 Memory and OC Tuning Guide - Infinity Fabric, EXPO, Dual-Rank, Samsung and Hynix DDR5 in Practice test with Benchmarks and Recommendations | Page 9 | igor´sLAB

AMD’s new Ryzen 7000 desktop CPUs, based on the Zen4 micro-architecture, still use the same chiplet design as their predecessors, with a few small but not negligible changes. Igor already gave details…

www.igorslab.de

Ok so I was wrong about FCLK: if your motherboard runs it higher than 2000 without instability you are fine, on DDR5-6400 memory 2167 FCLK can be 1% faster than the same settings at 2000 FCLK (although on X3D chips the margins are smaller with all memory optimisations).

What IS important is keeping UCLK and MemCLK in a 1:1 ratio. Since 6000 CL30 was within 1% of the best possible configuration and is supposed to work out of the box, I remembered those settings, which are 2000 FCLK, 3000 UCLK and MemCLK i.e. a 2:3 ratio. If you encounter any stability problems I would lower FCLK manually back to 2033 or 2000 and leave UCLK and MemCLK as they are: 3000 for DDR5-6000 and 3200 for DDR5-6400 memory. If all runs well then the higher FCLK is not a problem being out of sync from the 2:3 ratio.

Certainly won’t have any issues of instability as it seems stable at 2200, so lower will be no problem. Saying that, FCLK stability is very difficult to test, so you can’t be 100% sure. For intrigue, I’m going to bench 2133/2200 and see the difference

 

[IceWolf]

Certainly won’t have any issues of instability as it seems stable at 2200, so lower will be no problem. Saying that, FCLK stability is very difficult to test, so you can’t be 100% sure. For intrigue, I’m going to bench 2133/2200 and see the difference

Sorry you replied before I finished editing my post. For gaming it does look like 2133 could be the sweet spot at DDR5-6400 so I will be interested to see what your results show

 

[matty007]

I remember a previous setup I had. I had FCLK a little too high but it was nightmare diagnosing it. Just used to get random audio glitches/tiny dropouts appear out of nowhere, and the odd crash occasionally. Took me forever to pin it down to FCLK. It’s very hard to test for.

I am pretty sure it is stable at 2200, but given my experiences, I wouldn’t mind running 2133/2167 if the difference is tiny/non-existent for peace of mind.

 

[matty007]

Sorry you replied before I finished editing my post. For gaming it does look like 2133 could be the sweet spot at DDR5-6400 so I will be interested to see what your results show

I shall bench it shortly and find out.

 

[daninthemix]

I remember a previous setup I had. I had FCLK a little too high but it was nightmare diagnosing it. Just used to get random audio glitches/tiny dropouts appear out of nowhere, and the odd crash occasionally. Took me forever to pin it down to FCLK. It’s very hard to test for.

I am pretty sure it is stable at 2200, but given my experiences, I wouldn’t mind running 2133/2167 if the difference is tiny/non-existent for peace of mind.

Damn, I had crackling on SPDIF out that went away...coincidentally I lowered FCLK from 2100 to 2000. Now I'm wondering if my weird LAN issue might also be solved (the NIC was refusing to go above 100mbps).

 

[matty007]

Damn, I had crackling on SPDIF out that went away...coincidentally I lowered FCLK from 2100 to 2000. Now I'm wondering if my weird LAN issue might also be solved (the NIC was refusing to go above 100mbps).

The crackling was almost certainly down to FCLK, yes.

 

[Zangoose]

Just installed mine - Seeing 60-70C temps on CPU (TCTL/TDIE) on HWINFO at pretty much idle, just discord/chrome open (1 tab). Is this ok or is it leading into a bad mount/paste job or just a cooler that isn't good enough? I have a H150i Elite installed currently - Maybe adjust the curves within iCue? Any help will be appreciated.

 

[Alex h]

I'd say that's a little high if it's idle mate

 

[tiberuk]

I had nothing but problems with my Aorus Elite so have swapped it for the Asus Tuf Gaming Plus. In my opinion you dodged a bullet. Gigabyte's BIOS has been broken since the safety update for X3D in October 2023. I am not the only one with problems on every BIOS since. Also according to Hardware Unboxed there was no benefit to performance going from B650 to X870 boards at 6000 CL30 EXPO, just other features

Well thats a very welcome response!

 

[Zangoose]

I'd say that's a little high if it's idle mate

Thanks - Going to take it off after food and see how the paste application looks and re-do and then go from there.

Update: Re-did the paste, its knocked off 10C and is now idling at 47-50C. I suppose thats better than what it was. Now at 60-77C while in Microsoft Flight Simulator, originally it jumped to 96C before I did the re-paste.

 

[matty007]

Thanks - Going to take it off after food and see how the paste application looks and re-do and then go from there.

Update: Re-did the paste, its knocked off 10C and is now idling at 47-50C. I suppose thats better than what it was. Now at 60-77C while in Microsoft Flight Simulator, originally it jumped to 96C before I did the re-paste.

A little high at idle still, but perhaps not if you aren’t running a curve optimiser?

 

[Narn]

btw in the bios at least for asus boards it gives the cpu ''sp'' rating.
does the number mean much when attempting to overclock? i only have a 91 rating

 

[Zangoose]

A little high at idle still, but perhaps not if you aren’t running a curve optimiser?

I don't think I am - Will look into it though.

 

[matty007]

btw in the bios at least for asus boards it gives the cpu ''sp'' rating.
does the number mean much when attempting to overclock? i only have a 91 rating

91??? It should be 110+

That being said, I have an ASUS board and can’t find the SP rating anywhere. On sites, I see it under ‘AI features’ but I don’t have that submenu for some reason.

You’ll soon know the capabilities once you start pushing it. However, 91 doesn’t sound right at all

 

[pBay]

Currently have my 9800x3d with 2x scalar, +200mhz, -35 CO. Hitting 24300 or so in Cinebench R23. All core maxing out under full load around 72C and a vcore of 1.2v.

Is that a decent chip or pretty much what most people are seeing?

 

[InVitriol]

Would a Noctua NH-U12S with dual fans be enough to cool the 9800x3d? Without overclocking. It works fine for my 5800x3d. Want to stay air and perhaps safe myself some money on the upgrade lol

 

[Cenuij]

Does anyone have advice in lowering temps a bit? I have a Noctua DH-15 dual tower, my idle temps are low-mid 40s, cinebench and gaming peaks at low 80s. And it's winter, so I assume summer will be much worse.

I am running at -25 Curve Optimiser.

These are not bad temps, your chip is designed to reach the maximum operating thresholds as soon as possible to get as much work done as fast as possible, then back off.

80c is small beers, you have 15c overhead.

If you want, you can test your cooling with artificial workloads that target the most energy demanding instruction sets on your CPU. y-cruncher bbp will make things nice and toasty for you.

 

[Mike25]

Managed to pre-order today, along with a bunch of other things
Now to wait on the new generation of GPUs

 

[Cenuij]

Would a Noctua NH-U12S with dual fans be enough to cool the 9800x3d? Without overclocking. It works fine for my 5800x3d. Want to stay air and perhaps safe myself some money on the upgrade lol

Yes