*** AMD "Zen 4" thread (inc AM5/APU discussion) ***

[Chud_Jenny]

I can see some potential value by getting a RDNA3 card and put the saved money to a AM5 X3D and bridge the gap closer to a 4090 though I have no idea as a 4090 may still be miles ahead.

Hardware Unboxed (Steve) will do the cost per frame and performance per watt charts in this review. Depends on what you're looking for overall.

 

[LePhuronn]

I can see some potential value by getting a RDNA3 card and put the saved money to a AM5 X3D and bridge the gap closer to a 4090 though I have no idea as a 4090 may still be miles ahead.

It's actually quite a simple couple of questions to ask yourself.

1: do you want the very tippy top best? if yes, 4090
2: if no, do you want the very best ray tracing capabilities? if yes, 4090 or 4080
3: if no, but Ampere's ray tracing is sufficient, or raster is more important to you, then 7900XTX

Also consider that the reference 7900XTX is AMD aiming for efficiency, there's a boat load of performance left on the table for AIBs to exploit.

 

[McGraw]

Does make me wonder if there's much point in buying the 'X' versions of Zen 4 CPUs when using an air cooler. Air coolers can't really handle 5.4ghz on all cores (advertised boost speed of the 7700X), unless you are very lucky and get a chip that can handle a PBO curve setting of -30 on all cores with stability.

Does seem like the non X chips will be a no brainer for the majority of customers.

The Ryzen 7700 should be a great gaming chip, if sold for £300 (or maybe even less).

The 5.4ghz boost is single core and they seem to mostly boost to 5.55ghz. Mine all core boosts to 5 25ghz. I run an air cooler. The thermal right peerless assassin cost me £42 and keeps my cpu below 95c

 

[g67575]

The 5.4ghz boost is single core and they seem to mostly boost to 5.55ghz. Mine all core boosts to 5 25ghz. I run an air cooler. The thermal right peerless assassin cost me £42 and keeps my cpu below 95c

Are you running a PBO Curve Editor undervolt though? That can make quite a big difference - I suspect that many Zen 4 CPUs can barely handle any undervolting with complete stability (they are generally maxed out at default).

EDIT - having said this I keep thinking I've got a good setting for the PBO that actually works... At the moment I've got -20 all core set.

 

[McGraw]

Are you running a PBO Curve Editor undervolt though? That can make quite a big difference - I suspect that many Zen 4 CPUs can barely handle any undervolting with complete stability (they are generally maxed out at default).

EDIT - having said this I keep thinking I've got a good setting for the PBO that actually works... At the moment I've got -20 all core set.

Not yet, I'm running it standard with EXPO only to make sure it's behaving forms couple of weeks.

 

[g67575]

Not yet, I'm running it standard with EXPO only to make sure it's behaving forms couple of weeks.

Sounds like your cooler is working better than mine. I have a Deepcool AK620, which has performed well enough.

I didn't use that much thermal paste though.

 

[g67575]

Has anyone had much luck in getting stability with an infinity fabric (Fclk) of 2133mhz or 2167mhz?

Haven't managed it with my 7700X yet, even at 1.35v SOC.

 

[McGraw]

I seem to have got a -30 offset on my 7700x working ok. Couldn't be bothered waiting any longer to try it.

According to OCCT it's boosting all cores to 5.5ghz with the odd peak of 5.55.

I don't seem to be able to test single core only at the moment. Temps are in the 70s in a cold room on silent profile and it's using about the same power as my 4790k did.

What's the most reliable way to test single and multi core boost? Thanks

 

[g67575]

@McGraw - You could try Aida64, I usually test with the cache and memory options enabled.

Is the boosting code not stable for PBO, by default?

I've been getting random stability test fails, usually within 10 minutes in Aida64 when a PBO curve is set. Sometimes when I reset my PC these tests appear stable (for at least 5 hours).

I set Boost override to -200mhz (which affects all cores I presume), and the random crashing doesn't seem to happen.

I presume that some zen 4 chips just can't handle boosting above the official boost spec, which for a 7700X is 5.4Ghz.

Has anyone had similar problems when using a PBO curve?

Note - I am using a Deepcool AK620 air cooler.

 

[LtMatt]

I seem to have got a -30 offset on my 7700x working ok. Couldn't be bothered waiting any longer to try it.

According to OCCT it's boosting all cores to 5.5ghz with the odd peak of 5.55.

I don't seem to be able to test single core only at the moment. Temps are in the 70s in a cold room on silent profile and it's using about the same power as my 4790k did.

What's the most reliable way to test single and multi core boost? Thanks

What board are you using? If it's an Asus board, make sure you are on the latest BIOS and have an option called, Medium Load Boost It, enabled.

If and when that option is enabled, clock frequency will increase when more than four cores are under any sort of load. However, you may find that your -30 curve optimiser settings are no longer stable, at either idle or under load.

 

[McGraw]

Thanks.

Tried AIDA. Here's where I'm at with just EXPO and -30. Not seen the medium load boost. ASUS TUF Gaming. Cinebench 20,973

 

[Noxia]

Back to boot errors again. I cannot tell what is causing them but currently getting blue screen saying pc needs to be repaired. This was happening before I installed my GPU too, when I got them before I just reinstalled OS but it always came back.

As usual, bios reset solves all. Wondering if it's memory tbh, not sure if my memory is an approved one or whatever.

 

[Journey]

Back to boot errors again. I cannot tell what is causing them but currently getting blue screen saying pc needs to be repaired. This was happening before I installed my GPU too, when I got them before I just reinstalled OS but it always came back.

As usual, bios reset solves all. Wondering if it's memory tbh, not sure if my memory is an approved one or whatever.

Have you run Memtest on your RAM at all since you built the system? If not try that first, you can make a nice bootable USB with Memtest86+ on and just let it run for 1-2 full passes, assuming no faults show up before that.

 

[Noxia]

Have you run Memtest on your RAM at all since you built the system? If not try that first, you can make a nice bootable USB with Memtest86+ on and just let it run for 1-2 full passes, assuming no faults show up before that.

No, purely because the crashes only happen during boot and never using (even when I used all day), I will try that though thanks. The blue screen said something about driver certification but I recall googling this before and getting nowhere. Can go weeks with no issues too.

 

[alec]

No, purely because the crashes only happen during boot and never using (even when I used all day), I will try that though thanks. The blue screen said something about driver certification but I recall googling this before and getting nowhere.

Hmm, occasional crashes during boot, stable otherwise.
Is there a difference between cold boot and restart?

Smells like memory training isn't quite stable.

More volts or slower speed on memory could help. Guessing it will get fixed in later bios.

 

[Noxia]

Hmm, occasional crashes during boot, stable otherwise.
Is there a difference between cold boot and restart?

Smells like memory training isn't quite stable.

More volts or slower speed on memory could help. Guessing it will get fixed in later bios.

No and actually last time, it started up and then restarted itself, this was with DOCP 1 or 2 selected. Not sure if it happens if I don't select it but then I am gimping my memory I guess. Thing is, the error codes it displays on my mobo don't show anything useful, just cycles through the usual ones at start. Last time it froze was on initialising PCI bus something but its not consistent.

 

[Meddling-Monk]

Hmmm, when you first put the system together, with the default settings, did you experience these issues? Or was it only after you changed settings that these crashse randomly occurred/

 

[g67575]

Hmm, occasional crashes during boot, stable otherwise.
Is there a difference between cold boot and restart?

Smells like memory training isn't quite stable.

More volts or slower speed on memory could help. Guessing it will get fixed in later bios.

I suspect that my DDR5 / memory clock isn't stable. I tested my system with the memory controller set to half speed, and it still failed a stability test last night, after 5 hours. It doesn't seem to be related to memory timings either.

And it does seem to be affected by rebooting or waking from sleep.

 

[g67575]

I suppose there's a chance I might be able to avoid these issues with higher DRAM voltage, if ASRock ever releases their latest beta BIOS (1.11.AS06) on lower tier boards than the X670E Taichi.

New additions:
1. Add High Voltage Mode for VDDIO.
2. Optimize system stability.

Don't know what the hold up has been... VDDIO is limited to 1.45v at the moment, but other DRAM voltages can be set to 1.5v or above (probably not a good idea to go much above 1.5v though).

 

[Noxia]

Hmmm, when you first put the system together, with the default settings, did you experience these issues? Or was it only after you changed settings that these crashse randomly occurred/

Cannot recall which is why I am currently just running everything at totally default to see what happens,