[yzonker]

So I've been playing Jedi:Survivor to test the claims of "low CPU usage leads to massive GPU bottleneck", all I can say is....

Seems to be an Nvidia issue or PC configuration issue not matching the engine demands well.

Stock settings 6950xt + OCd 7800x3d, all settings at High with Textures at Epic, 3440x1440, RT ON, Freesync ON, FSR2 Quality, Blur OFF, Filgtaim OFF and seeing 85+/-5fps most areas. Heaviest load areas dropping to 75ish, light load areas in the mid 90s.

If I set all to Epic it's about 7 FPS worse across the board.

This is roughly the same performance I observed in Jedi:Fallen Order my 2080ti when it released, but Survivor looks much better, so meh, not seeing the problem.

It has pretty-even frametime pacing, only stuttering on scene change while loading. GPU pegged at 98-100% usage. It does use a lot of VRAM at 14.8GB though, and with all at Epic it does fill VRAM, so it's spilling over in system memory.

The only other setting that is likely to be changing this from other people's experience is PSS disabled, which ignores core quality and schedules on next available instead of shuffling tasks to load best cores hardest. It also mitigates scenarios where the scheduler loads a thread per logical leaving unused cores idle because of lower CPPC value in an attempt to exploit best cores most.

Edit: will test SMT off later today as well as OCd GPU performance. I do wish it ran at 120+ FPS but it's perfectly playable as is.

A couple of users in the 4090 thread posted vids showing similar framerates. Must be people with older CPUs or badly configured systems.

 

[PensionMan]

Hello, my 7800X3D and asus X670e tuf plus wifi just arrived, do i need to wait for the new AGESA update to prevent burn problems or i can start with the latest bios ?

 

[tcclaviger]

A couple of users in the 4090 thread posted vids showing similar framerates. Must be people with older CPUs or badly configured systems.

Good to know. So, seems it's a configuration error or expectations are too high. Certainly the game has optimization issues, but I can't believe a 4090/7800X3D is incapable of 60fps as some are reporting on Reddit and the tech journos have starting running with.

@PensionMan Youre fine, just keep voltages in check if you're nervous (don't disable precision boost, crank BCLK to 120+, and add a 250mv positive offset....). I would also not use EXPO on anything Asus at the moment, Asus has a very long history of hammering high voltages with their presets.

AMD is not telling the whole story and released a mitigation without explaining the causation; even then, they only provided 1 voltage among many they're adjusting, SOC capped to 1.3.

 

[PensionMan]

Good to know. So, seems it's a configuration error or expectations are too high. Certainly the game has optimization issues, but I can't believe a 4090/7800X3D is incapable of 60fps as some are reporting on Reddit and the tech journos have starting running with.

@PensionMan Youre fine, just keep voltages in check if you're nervous (don't disable precision boost, crank BCLK to 120+, and add a 250mv positive offset....). I would also not use EXPO on anything Asus at the moment, Asus has a very long history of hammering high voltages with their presets.

AMD is not telling the whole story and released a mitigation without explaining the causation; even then, they only provided 1 voltage among many they're adjusting, SOC capped to 1.3.

I'm going to tune the memory manually so i don't use expo, but i want to use CO and PBO no problem with them ?

I mean if the update release in the next days it's fine i can wait 2/3 days.

 

[tcclaviger]

If the leak that attributes the CPU failures as a result of early degradation of thermal sensors in SOC due to too much voltage is correct, then yes PBO/CO is still safe. For that failure mode to occur multiple failures need line up


1-Too high SOC voltage.
2-Too little cooling allowing thermal runaway.
3-Nothing shorts in a way that triggers VRM OCP protection before blisters form.

Personally this is the causation I believe as it is something AMD would absolutely try to keep hidden by releasing the nothing burger they released and it fits the rare intersection causing events to be very infrequent.

Overall though we still don't know. AMD pointed a finger at a marketing term they made up which has zero technical meaning (EXPO). Most chips had no issues. Chances are non-zero of an thermal failure event, but they are very low.

Personally, I'm never adopting the gimp AGESA. Were I on an air cooler bouncing off tjmax, perpetually, yeah I would probably update to it.

 

[ZealotKi11er]

So I've been playing Jedi:Survivor to test the claims of "low CPU usage leads to massive GPU bottleneck", all I can say is....

Seems to be an Nvidia issue or PC configuration issue not matching the engine demands well.

Stock settings 6950xt + OCd 7800x3d, all settings at High with Textures at Epic, 3440x1440, RT ON, Freesync ON, FSR2 Quality, Blur OFF, Filgtaim OFF and seeing 85+/-5fps most areas. Heaviest load areas dropping to 75ish, light load areas in the mid 90s.

If I set all to Epic it's about 7 FPS worse across the board.

This is roughly the same performance I observed in Jedi:Fallen Order my 2080ti when it released, but Survivor looks much better, so meh, not seeing the problem.

It has pretty-even frametime pacing, only stuttering on scene change while loading. GPU pegged at 98-100% usage. It does use a lot of VRAM at 14.8GB though, and with all at Epic it does fill VRAM, so it's spilling over in system memory.

The only other setting that is likely to be changing this from other people's experience is PSS disabled, which ignores core quality and schedules on next available instead of shuffling tasks to load best cores hardest. It also mitigates scenarios where the scheduler loads a thread per logical leaving unused cores idle because of lower CPPC value in an attempt to exploit best cores most.

Edit: will test SMT off later today as well as OCd GPU performance. I do wish it ran at 120+ FPS but it's perfectly playable as is.

I have been looking for AMD GPU benchmarks especially 7900 XTX but cant seem to find any. Most people simply have Nvidia these days so the blame falls on the game. Nvidia is perfect.

 

[tcclaviger]

I have been looking for AMD GPU benchmarks especially 7900 XTX but cant seem to find any. Most people simply have Nvidia these days so the blame falls on the game. Nvidia is perfect.

I'm holding out for the next bigboy AMD card, but the 6950xt is such a good deal at ~650 it's hard to pass up IMHO. The 7900xtx I find uninteresting because overclocking semi-lockdown, though stock it's still like 20% ahead of my OCd 6950.

 

[tcclaviger]

In short: for a tiny boost disable SMT. For an even tinier boost also disable PSS. 0.2% lows and less suffer when swapping with system memory (we already know that).

Some Testing was done. Manual play of a short section of game I can reproduce with fighting included and a group of enemies to get more than a scenic walkthrough performance. Keep in mind, I'm capturing using OBS and CapFrameX, so there is some amount of performance drop vs what I see playing normally, seems to be about 9 to 10% lower FPS while capturing:

Below are the stats I got testing PSS ON/OFF and SMT ON/OFF and High vs EPIC settings and RT ON/FSR2 Quality/Freesync ON/163 FPS frame cap via RTSS

detailed graphs (bottom 4) are from the SMT OFF/PSS OFF capture.

CPU is 108 ECLK with positive CO offsets (I laugh in the face of thermonuclear X3D chips), 6400 tight C30, 2133 FCLK. Passes any amount of time in TM5 Anta777 and Y-cruncher all-test testing.

 

[pj530i]

Haven't really gotten into the game yet but I noticed right away that Jedi Survivor runs on the non-vcache CCD on my 7950x3d. Other games run on the vcache CCD as I'd expect

Anyone else see this?

 

[PensionMan]

I'm holding out for the next bigboy AMD card, but the 6950xt is such a good deal at ~650 it's hard to pass up IMHO. The 7900xtx I find uninteresting because overclocking semi-lockdown, though stock it's still like 20% ahead of my OCd 6950.

Yeah, price perf wise the 6950XT is unbeatable now you can get a nitro+ one for 650€ here, amd kill the sell of the 7900XT and 7900XTX with that price
And if you don't play 4K 120+ or RT games no reason to buy more

 

[ZealotKi11er]

Yeah, price perf wise the 6950XT is unbeatable now you can get a nitro+ one for 650€ here, amd kill the sell of the 7900XT and 7900XTX with that price
And if you don't play 4K 120+ or RT games no reason to buy more

I play at 4K which 7900 XTX bring me just above the threshold that my heavy OCed 6900 XT could not.

 

[yzonker]

I'm holding out for the next bigboy AMD card, but the 6950xt is such a good deal at ~650 it's hard to pass up IMHO. The 7900xtx I find uninteresting because overclocking semi-lockdown, though stock it's still like 20% ahead of my OCd 6950.

Ah come on over to the green side. 😉

4090 TUF/Strix + 1kw bios + EVC2 = fun

 

[ZafirZ]

Haven't really gotten into the game yet but I noticed right away that Jedi Survivor runs on the non-vcache CCD on my 7950x3d. Other games run on the vcache CCD as I'd expect

Anyone else see this?

You need to press windows key + g while playing it to get up the game bar, press the little config icon and then tick the box that says something like remember this is a game. After that you can reboot and it'll correctly put it on the right CCD. Since it's new game bar just doesn't know it's a game yet.

 

[skyline385]

GN video incoming, sounds like its more than Vsoc

We have been able to reproduce a catastrophic failure resulting in the motherboard self-immolating while we were running external current logging, thermography, and direct VSOC leads to a DMM. The issue involves incompetence on many levels. Video script being finalized now.
This video has taken us all week to do. We stepped off a plane from AMD HQ (unrelated, but convenient timing) on Tuesday and immediately set to work. It has been a time-consuming, expensive, physically exhausting process and we've loved every minute of it. The team is thrilled to have a real result and successful root-cause analysis that goes layers beyond "it's SOC voltage," because it's not that simple. We appreciate your support in viewing the video when it launches, because that's the real reward for our tired team. Keep an eye out and share it wide.

 

[Tactix]

Looking forward to the GN results, really hoping the results don't end in my returning my new rig
Although i suspect if that were they case they'd at least given some kind of warning to prevent further destruction.

 

[Tactix]

Is it odd
that my CPU VDDCR_SOC Voltage literally never reads any variance from 1.250v ?
Huh and neither does the _Misc.

This only started after entering a manual CPU VCORE SoC of 1.22v when on auto it fluctuated

 

[Tactix]

I just realized i think i plugged the 8 pin CPU cable in backwards on the Corsair SF750, meaning i plugged the "CPU" Labeled end into the motherboard and the "Type-4" Labeled end into the PSU.
Im guessing this should not matter as the pins should connect to the same place on each end right?
I mean its been running fine for days and all.

Normally id just swap it but this is an SFF build and id rather not pull **** apart, from what ive seen of the pin out on the 8 Pin cpu connectors they are the same on each end.

haha ok Nevermind its correct and I need sleep

 

[Steve1991]

I just realized i think i plugged the 8 pin CPU cable in backwards on the Corsair SF750, meaning i plugged the "CPU" Labeled end into the motherboard and the "Type-4" Labeled end into the PSU.
Im guessing this should not matter as the pins should connect to the same place on each end right?
I mean its been running fine for days and all.

Normally id just swap it but this is an SFF build and id rather not pull **** apart, from what ive seen of the pin out on the 8 Pin cpu connectors they are the same on each end.

haha ok Nevermind its correct and I need sleep

Yeah you cant plug them wrong even if you wanted to, at least not on corsair, they have these little plastic things attached preventing you.

 

[the_aeon]

buy in 7800x3d, Asus Rog strix b650e-f and g.skill Expo 6000 CL 30, it's been a week of blue screen reboots and freezes in Windows, it crashes even without Expo profile, bios on the latest version and chipset too, which can be ?? please help! I'm going crazy! is it the ram? the cpu? motherboard?? please!

 

[Tactix]

buy in 7800x3d, Asus Rog strix b650e-f and g.skill Expo 6000 CL 30, it's been a week of blue screen reboots and freezes in Windows, it crashes even without Expo profile, bios on the latest version and chipset too, which can be ?? please help! I'm going crazy! is it the ram? the cpu? motherboard?? please!

It’s usually ram, but hard to say… this is a fresh install of windows ?
can you try and run memtest86 through USB at optimized defaults (not expo) and see how that goes.
Link to the one I use free edition is fine MemTest86 - Official Site of the x86 Memory Testing Tool!
Does 4 full passes took around 3.5 hrs on my 6000 CL 30 mem, but you’ll get errors far sooner if there’s an issue.
Select boot menu at startup then your USB, don’t touch anything and the test will automatically start after a few seconds, good luck!