Poll: Ryzen 7950X3D, 7900X3D, 7800X3D

[Audioboxer]

A lot of these users were running PBO too which is over clocking. The actual number running pure stock is likely much smaller than claimed. My CPU has issues with AVX2 with PBO +200 but runs it fine stock. That said there will always be a small number of faulty runs that fail at stock.

The AVX2 stuff I was talking about was just at stock, like the OC.net topic also has posts throughout 2023 with fails at stock. Seems to be an issue bubbling under the surface since launch but other things like chips going on fire in 2023 due to VSOC took precedent lol. Then there is the fact such a small number of people will ever run Prime95, let alone, specific AVX2 testing.

On that note, kind of sad I missed the days with VSOC going higher than 1.3v lol.

Whenever PBO is turned on ymmv, there isn't really any official complaining to be had at failures with it on, reduce the curve and deal with the silicon lottery lol. Hence me with 1 core at -1

Yeah the performance on the 7900X3D is a bit all over the place, but they have dropped the price a fair bit to try and compete with the 14700K with it. Really they need to drop the price on the 7950X3D a bit though as it makes up for the areas the 7800X3D drops back 1-2 generations in performance outside of gaming.

If you play games just best going 7800x3D. If you do some multitasking and/or want something outside of "just gaming" I'd save up for a 7950x3D/try and bag one second hand. Though, given we're on an OC forum I'd say my advice with a 7950x3D in this climate is buy from somewhere you can return easily! Might rule out private second hand sale, unless you approach the seller and they too are an OCing nutter and would be able to tell you how good the chip is before buying.

I think the AMD box warranty transfers if a private seller had all original packaging, but RMAing with AMD seems to be quite a lengthy process.

 

[LtMatt]

The AVX2 stuff I was talking about was just at stock, like the OC.net topic also has posts throughout 2023 with fails at stock.

Whenever PBO is turned on ymmv, there isn't really any official complaining to be had at failures with it on, reduce the curve and deal with the silicon lottery lol. Hence me with 1 core at -1

I'm not convinced tbf, most people don't even think PBO is overclocking so probably had it running at +200Mhz.

I have two cores at +3 with 200Mhz if I want to pass stability in AVX2. It's a shame that there is not a one size fits all 'test' for tuning CO would make lives a lot easier.

Still, this is 100% stable in everything. SHA3, AVX2, Football Manager 2024, y cruncher 00-x86, Karhu 8+ hours. Karhu is about 334.5MB average.



I saw your WRTS at 1 on oc.net, I might try lowering mine to see if I can get anymore from it.

 

[Rroff]

The AVX2 stuff I was talking about was just at stock, like the OC.net topic also has posts throughout 2023 with fails at stock. Seems to be an issue bubbling under the surface since launch but other things like chips going on fire in 2023 due to VSOC took precedent lol. Then there is the fact such a small number of people will ever run Prime95, let alone, specific AVX2 testing.

On that note, kind of sad I missed the days with VSOC going higher than 1.3v lol.

Whenever PBO is turned on ymmv, there isn't really any official complaining to be had at failures with it on, reduce the curve and deal with the silicon lottery lol. Hence me with 1 core at -1

A bit off topic for this thread but seen some interesting things with the mobile 7000 series chips/SoCs like the 7840u, etc. there seems to be as much as 25% performance variance at a given TDP on some models and a small but not insignificant number of chips which are faulty, usually which only manifests under specific, often synthetic, workloads but can also cause random problems under normal use.

 

[jigger]

I wouldn't touch a 7900X3D personally, get 8 3D cache cores either by going for the 7800x3D or spending a bit more on the 7950x3D.



While I don't doubt Windows installs that have had hardware changes/some who joined AM5 early and went through chipset/BIOS issues end up having jank (so clean install for hardware changes), the issue with the AVX2 testing is more about the way some cores have been binned on CCD1. It appears they don't take enough voltage and end up erroring. It's no surprise it tends to be the cores that are identified as the best cores.

I don't believe these issues are as widespread with 7950x's, hence, pinpointing the balancing act AMD have done to bring 3D cache to a 2 CCD processor (effectively sticking 3D cache on a 7950x).

Not too much uses AVX2 code so chances are lots of regular people will enjoy their 7950x3Ds as is, some won't even use PBO. But anyone overclocking/tinkering is unlikely to want to accept a chip that can't pass an industry standard like Prime95 at stock. Especially when other 7950x3Ds pass just fine.

The 7900X3D is somewhat of a bargain. I could make really good use of 24 threads and the system would double up as top tier gaming system. A 7950X3D is quite a bit more.

lol at industry standard.

 

[nlel1975]

Can you guys explain then why my CPU which failed OCCT Extreme Variable Large Test on Corecycler at stock when I first ran it, but now passes it for the full hour after only having done a fresh windows install.

I’ve not touched anything in the BIOS its totally stock in all my tests right from the start. I promise you I’ve not RMA’d my CPU, it’s the same one that used to fail the specific OCCT test but now passes.

To me this can only be a software thing, either windows itself or OCCT malfunctioning under certain windows installs. My advice to anyone with this problem is to try a fresh windows install before RMA.

 

[Rroff]

Can you guys explain then why my CPU which failed OCCT Extreme Variable Large Test on Corecycler at stock when I first ran it, but now passes it for the full hour after only having done a fresh windows install.

Marginal stability can be affected by even small differences in settings and system file versions, etc. while it could be a simple software issue it could be that differences in the OS environment simply aren't exposing the weakness, or not as much. Though that tends to be more common with RAM stability issues than CPU.

 

[nlel1975]

Marginal stability can be affected by even small differences in settings and system file versions, etc. while it could be a simple software issue it could be that differences in the OS environment simply aren't exposing the weakness, or not as much. Though that tends to be more common with RAM stability issues than CPU.

Mine now passes the specific OCCT test at stock, what more would anyone want?

 

[jigger]

Can you guys explain then why my CPU which failed OCCT Extreme Variable Large Test on Corecycler at stock when I first ran it, but now passes it for the full hour after only having done a fresh windows install.

I’ve not touched anything in the BIOS its totally stock in all my tests right from the start. I promise you I’ve not RMA’d my CPU, it’s the same one that used to fail the specific OCCT test but now passes.

To me this can only be a software thing, either windows itself or OCCT malfunctioning under certain windows installs. My advice to anyone with this problem is to try a fresh windows install before RMA.

Some kind of patch.

 

[nlel1975]

Some kind of patch.

Surely the others who've had or still have this problem are fully up to date with windows patches.

 

[jigger]

Surely the others who've had or still have this problem are fully up to date with windows patches.

Who knows. So many variables to take into consideration.

 

[LtMatt]

Can you guys explain then why my CPU which failed OCCT Extreme Variable Large Test on Corecycler at stock when I first ran it, but now passes it for the full hour after only having done a fresh windows install.

I’ve not touched anything in the BIOS its totally stock in all my tests right from the start. I promise you I’ve not RMA’d my CPU, it’s the same one that used to fail the specific OCCT test but now passes.

To me this can only be a software thing, either windows itself or OCCT malfunctioning under certain windows installs. My advice to anyone with this problem is to try a fresh windows install before RMA.

I can’t. I still think the test can be a bit ropey tbh so not sure how much value I place on it.

 

[Audioboxer]

Can you guys explain then why my CPU which failed OCCT Extreme Variable Large Test on Corecycler at stock when I first ran it, but now passes it for the full hour after only having done a fresh windows install.

I’ve not touched anything in the BIOS its totally stock in all my tests right from the start. I promise you I’ve not RMA’d my CPU, it’s the same one that used to fail the specific OCCT test but now passes.

To me this can only be a software thing, either windows itself or OCCT malfunctioning under certain windows installs. My advice to anyone with this problem is to try a fresh windows install before RMA.

A fresh Windows install can make a huge difference if something has been configured wrong or there has been driver/software jank built up over time.

Apart from AGESA sometimes being a mess AMD have been known to release broken or bugged chipset drivers. Heck, the latest couldn't even be installed for a while and flagged as a virus lol.

It's not just OCCT, that app simply uses Prime95 in the backend. Corecycler using Prime95 AVX2 can cause the same core crashes. I repeated it with Corecycler on my CPUs that wouldn't pass at stock. If you RMA with AMD they will accept Prime95 erroring as evidence of a busted CPU (as long as you're running stock/PBO is off).

Not bad advice for anyone, but if you're upgrading CPU a fresh install is always good advice. More so if it's a platform change. Heck if you're pushing OCing hard you can even corrupt Windows from errors/reboots lol, so it's not even bad advice once you find something stable to simply reinstall just in case

Press F for Steam/other apps logging themselves out when you get curve optimiser core reboots (especially if running SHA3 benchmark in AIDA64 lol). The amount of times I've had to log back into Steam or import my iCUE profile again due to stability testing!

 

[BRIGGBOY81]

Hi, just slapped in the 7800x3d and what expo are people using? Is it better use 1, 2 or tweaked. The ram is 64gb gskill trident z5 cl30

 

[nlel1975]

Hi, just slapped in the 7800x3d and what expo are people using? Is it better use 1, 2 or tweaked. The ram is 64gb gskill trident z5 cl30

I just used expo I it set the ram to 6000 with cl30

 

[nlel1975]

I can’t. I still think the test can be a bit ropey tbh so not sure how much value I place on it.

Surely my results only serve to support this theory.

 

[LtMatt]

Surely my results only serve to support this theory.

Yes they certainly do. If other tests fail, like Prime95 (AVX2) as Audio mentioned, then it's just a faulty CPU which should be RMA'd.

 

[nlel1975]

Yes they certainly do. If other tests fail, like Prime95 (AVX2) as Audio mentioned, then it's just a faulty CPU which should be RMA'd.

I'll never know as I never used the separate Corecycler program in AVX2 mode ignoring cores 0, 1, 2, 3, 4, 5, 6, 7 running for 2 minutes per core until now... and it passed.

I was under the impression running OCCT Extreme Variable Large Test on Corecycler mode was the same test and you only needed to run either of them to test for it.

Unfortunately I can't restore the system image which failed the OCCT test, to test in Corecycler program as I've deleted it and made a new image of my freshly installed system.

 

[LtMatt]

I'll never know as I never used the separate Corecycler program in AVX2 mode ignoring cores 0, 1, 2, 3, 4, 5, 6, 7 running for 2 minutes per core until now... and it passed.

I was under the impression running OCCT Extreme Variable Large Test on Corecycler mode was the same test and you only needed to run either of them to test for it.

Unfortunately I can't restore the system image which failed the OCCT test, to test in Corecycler program as I've deleted it and made a new image of my freshly installed system.

If it passed the other core cycler, it was probably okay tbf. Why it failed OCCT, who knows. Likely some software issue and not hardware otherwise it would fail now.

Perhaps the OCCT Test of cycling between cores in just a couple of seconds exposes a fault that any other workload does not, who knows.

All we know is the OCCT VVX2 is not the be all and end all of stress tests, as other apps pick up instability which it misses.

 

[nlel1975]

All we know is the OCCT VVX2 is not the be all and end all of stress tests, as other apps pick up instability which it misses.

It’s likely in my opinion that it would have failed both the OCCT test and the separate Corecycler test under my old system image but I’ll never be able to prove that as I’ve now deleted said image.

All I know is I had the OCCT errors and now don’t under a fresh windows install. There’s never been anything wrong with my CPU.

 

[BRIGGBOY81]

OK had a quick play last night and she seems happy with -30 on 6 cores and -25 on the 2 strongest. Will do a proper stress test later today but for now it's chocolate time