***The Official 5600X \5800X owners thread***

[danq]

Wow 5800X Price drop to 399.99!

 

[humbug]

Wow 5800X Price drop to 399.99!

That's MSRP, i paid £439 for mine.

 

[danq]

That's MSRP, i paid £439 for mine.

oh snap, I didn't know that lol. I was gonna pull the plug this morning from a competitor as I can buy discounted giftcards - but with the updated price, I can price match and get for approx £367!

 

[alec]

My Deepcool Gammaxx L120 is struggling with it during high load tho, 85c in Cinebench R23. Having said that 4.6Ghz to 4.7Ghz Ain't bad.

85 is not bad. I'm hitting 81-82 in Cinebench with Arctic 240.

 

[Guest2]

Try with PBO limits set to disabled. Auto / motherboard provided too much voltage for me.

I had a game crash (full reboot) at +200Mhz so have dropped it back to 0Mhz.
No point in multiple rounds of stability testing with R20, R23, Tomb raider bench, 3d mark benches then have games crash for the sake of a few extra fps (which wont be used anyway due to my fps cap)

Highly annoying when multiplayer games crash when you're doing well. I was about 15 mins into a Warzone solos game with 5 kills and 2nd to last circle when my machine rebooted. Wonderful!

Now running;
PBO - advanced
PBO limits - disabled
PBO scaler - auto
Max CPU boost clock override - auto
Curve optimizer - all -20, good cores 0 & 6 -5
CPU core control - auto
SMT control - auto
LN2 Mode 2 - Auto
NUMA nodes per code - auto

 

[humbug]

85 is not bad. I'm hitting 81-82 in Cinebench with Arctic 240.

It hit 90c after 4 minutes, temp was too good to be true, it still did a good job of keeping it to <4.6Ghz. I'm so impressed with these CPU's.

Try with PBO limits set to disabled. Auto / motherboard provided too much voltage for me.

I had a game crash (full reboot) at +200Mhz so have dropped it back to 0Mhz. No point in multiple rounds of stability testing with R20, R23, Tomb raider bench, 3d mark benches then have games crash for the sake of a few extra fps (which wont be used anyway due to my fps cap)
Highly annoying when multiplayer games crash when you're doing well. I was about 15 mins into a Warzone solos game with 5 kills and 2nd to last circle when my machine rebooted. Wonderful!

Played 2 hours of Star Citizen last night, i find that pretty well finds unstable CPU's, it was fine.

Its only boosting to 4.85Ghz ST, before i mess about figuring out how i get that up i might just see how low i can get the Negative Curve Optimiser, currently at 20, might go for 22 next, all core volts in R23 are 1.27v to 1.284v @ 4.6Ghz.

Sorry for how dark the video is, something is Wrong with Shadowplay / Youtube where anything uploaded to Youtube that was recorded with Shadowplay the image is really dark.

 

[haszek]

I managed to get my 5800X stable at even lower power limits (lower than disabled) to 120/75/110 (PPT/TDC/EDC) and those are my results. Increasing power didn't give me any benefits other than increased temps. This is with +150mhz, CO negative 30/15(best)/25/30/30/30/25/25. Adding 200Mhz also hasn't changed anything for me even with increased power limits, so not sure if it's working correctly on my motherboard/Bios.

Spoiler: CB23

 

[Guest2]

This is with +150mhz, CO negative 30/15(best)/25/30/30/30/25/25.

What does this bit mean?

I have all cores -20, except two best cores on -5. Do you mean you have -30 or -25 on all cores and -15 on the best core?

I might try reducing the power limits and re-testing

 

[DanielSon]

Hi all,

I have had the 5800x a couple of weeks now, supplied from another site just before the new year. Spent a nice couple of days re-building my setup and OS before work from home restarted! My parents then had the nice new year treat of my 6700k pass me down in a nice case as well and a full OS rebuild to boot!

lots of good info in this thread and I have been tweaking, testing and tuning now for a while. Still need to look at the RAM but will do that later as currently in the middle of critical work from home stuff etc.

This weekend I have done some performance and efficiency testing with cinebench R20 and R23. Results pictured below the spoiler.

For the the efficiency tests all I changed was the PPT limits in bios and nothing else. My testing method would not stand up to 'Gamers Nexus' but I'm happy enough! I was using HWInfo 64 and waiting for steady state on the power, temps and performance at each power level. Then reset the min + max recording at the start of a cinebench loop and recorded the average voltage, temp etc just before the run finished so not influenced by the spikes between runs etc. Overall I was not rigorous with this but the results look good enough. Then closed HWinfo and recorded a clean run without the HW polling that does effect score by a 100 points or something. I can also boost score by ~200 points by running the AIO at full pump and fan speed so a score over 16000 on R23 is easy if I want it.

You can really see why the 65-70W level is the sweet spot on these 7NM chips here! It's almost at the peek perf/watt and just before the raw performance was dropping off the cliff. This was with no additional tuning of the curve optimizer on the 8 core so you can see why the 5600x is doing so well! I'm sure with tuning at both ends of the septum things could be pushed further for performance or efficiency.

I'm currently running with 105W in bios, -20 Curve optimiser (prime95 stable) and +50 Mhz. This is giving 4900 Mhz in games when not under stress conditions or ~4550 Mhz on 100% all core. Roll on getting a mythical 3080 GPU one day and 4k VRR over 100 FPS will be sweet! When gaming now im using 70-80% render scale where needed with the 1080ti to keep above 60 FPS and while its still v nice there is a hit to image quality

You can also see an R20 comparison to my 4C 8T 6700k. The AMD chip is over 2.5x faster overall while matching the 6700k with only 4 threads of the potential 16!

5800x 8C 16T
Asus Strix B550-F Gaming
Corsair H110iGT 240mm AIO with retrofitted ML140 fans (well worth it - stock fans vibrated a lot)
16GB 3600 C14 8 Pack RAM - not tuned yet!
Seasonic Prime Ultra Snow Silent 650W 80 Plus Platinum
Nvidia 1080Ti - waiting on 3080 order!

key settings for below testing.
curve optimiser -25
freq tuning +50

Spoiler: Images of testing graph results

 

[haszek]

What does this bit mean?

I have all cores -20, except two best cores on -5. Do you mean you have -30 or -25 on all cores and -15 on the best core?

I might try reducing the power limits and re-testing

Yes exactly.

 

[haszek]

Hi all,

I have had the 5800x a couple of weeks now...

To me it looks 120 is the sweet spot like I have post above you. Decent drop in temps in comparison to default 140, plus small drop in performance while decent upgrade to lower limit

 

[alec]

graphs

curious how single core voltages go down with power limits. 0.084 mV difference for same 4900 clock
PPT limit shouldn't be in play for single core load, yet it still affects it

 

[Guest2]

I'm currently testing some power limited 5800X settings.
PBO limits (less than disabled/default)
125 PPT
75 TDC
110 EDC

Boost clock override - 150MHz
Scaler - Auto
Thermal throttle limit - Auto
Curve all -20, 2 good cores on -5

Results are quite good with much lower temps (Noctua U14-S)
All core boost is ~4.6Ghz, single core boost 5Ghz. Its about a 10c reduction for a small drop in multi core performance. IF still 1900Mhz, DDR still 3800MHz

 

[DanielSon]

curious how single core voltages go down with power limits. 0.084 mV difference for same 4900 clock
PPT limit shouldn't be in play for single core load, yet it still affects it

indeed its interesting how the power and temp goes down but remains stable at 4.9 Ghz single core turbo.

but if i reduce the curve optimiser more the multi core starts to be unstable so its a slight compromise overall.

anyway, all good fun!

Dan

 

[humbug]

Curve Optimiser = Voodoo Science. I don't understand how you get more for less but it seems to work.

 

[Guest2]

Curve Optimiser = Voodoo Science. I don't understand how you get more for less but it seems to work.

I thought it was voodoo at first but its quite simple. Reduced voltages (on weakers cores) = reduce heat. Reduced heat = more boost for better performance

 

[humbug]

I thought it was voodoo at first but its quite simple. Reduced voltages (on weakers cores) = reduce heat. Reduced heat = more boost for better performance

Ok that is quite simple, i'm no less impressed with it. And its good to see overclocking / Undervolting expanded to this level of graularity, its like they gave us what we wanted without us even knowing we wanted it.

 

[alec]

Curve Optimiser = Voodoo Science. I don't understand how you get more for less but it seems to work.

Offset undervolt causing clock stretching - now thats voodoo I still can't understand.
Somehow CPU notices it is getting too low voltage and fakes high clocks while actually running slower?

 

[Guest2]

Ok that is quite simple, i'm no less impressed with it. And its good to see overclocking / Undervolting expanded to this level of graularity, its like they gave us what we wanted without us even knowing we wanted it.

Exactly! Build it and they will come
I'm sure it'll get better as the bios updated come. The later bios works much better for me (MSI MEG Unify x570)

 

[Guest2]

Offset undervolt causing clock stretching - now thats voodoo I still can't understand.
Somehow CPU notices it is getting too low voltage and fakes high clocks while actually running slower?

yeh, this I dont get
There's no point in clock stretching. It's like its showing a faked cpu speed just so people may look at it and think "wow, look at my 5.5Ghz clock speed" only to realise a year later it's giving the same persormance as someone elses 4.5Ghz