5800X 3D WORSE than 5600 at gaming?!

[Nanoray]

Just installed my shiny new 5800X 3D and dialled in the RAM settings to precisely the same as I had with my outgoing 5600. Ran some gaming benchmarks... and it's WORSE - Superposition was down nearly 20%, Heaven was down from 4500 to 4250, and Shadow of the Tomb Raider came out within margin of error, but with admittedly much better CPU frametimes (and it went from 41% GPU bound to 70%).

 

BIOS has just been updated to latest stable, and aside from the RAM overclock, I have not touched any voltages - the 5600 and 5800X 3D are being compared stock for stock.

 

What on earth have I done/not done wrong?! I can't tell you how gutted I am to have spent £240 to get a WORSE gaming performance than a boggo 5600 I can keep hold of for nothing!

 

M.

[whispous]

So the one actual game you tried got better?

[Imakuni]

RAM speed kinda doesn't matter on the X3D, just enable XMP and call it a day to avoid performance drop due to non fatal instability.

 

At any rate, what was your mobo again?

[filpo]

10 minutes ago, Nanoray said:

Just installed my shiny new 5800X 3D and dialled in the RAM settings to precisely the same as I had with my outgoing 5600. Ran some gaming benchmarks... and it's WORSE - Superposition was down nearly 20%, Heaven was down from 4500 to 4250, and Shadow of the Tomb Raider came out within margin of error, but with admittedly much better CPU frametimes (and it went from 41% GPU bound to 70%).

 

BIOS has just been updated to latest stable, and aside from the RAM overclock, I have not touched any voltages - the 5600 and 5800X 3D are being compared stock for stock.

 

What on earth have I done/not done wrong?! I can't tell you how gutted I am to have spent £240 to get a WORSE gaming performance than a boggo 5600 I can keep hold of for nothing!

 

M.

Wait which gpu?

[Shimejii]

Also Thermals are a HUGE thing with 5800x3d. Also you need to use PBO2 Tuner, otherwise it wont be running at optimal settings. You can download pbo2Tuner or see if your bios has the it. -25 to all cores to start, then if you have a good air cooler 120W PPT, 80 TDC, and 110 EDC for the best results. 

 

What cooler do you have?

[Nanoray]

19 minutes ago, filpo said:

Wait which gpu?

4070, both times.

[Nanoray]

8 minutes ago, Shimejii said:

Also Thermals are a HUGE thing with 5800x3d. Also you need to use PBO2 Tuner, otherwise it wont be running at optimal settings. You can download pbo2Tuner or see if your bios has the it. -25 to all cores to start, then if you have a good air cooler 120W PPT, 80 TDC, and 110 EDC for the best results. 

 

What cooler do you have?

NH-D15S, with an extra 120 mm on the front.

 

Going to try CO undervolt tomorrow - my BIOS has Kombo strike, but I think MSI's latest BIOS lets you use CO for the X 3D anyway.

My plan is to use -25 for my "starred" cores and -30 for everything else, then kick it in the nads with some low-load OCCT core cycling.

[Nanoray]

30 minutes ago, Imakuni said:

RAM speed kinda doesn't matter on the X3D, just enable XMP and call it a day to avoid performance drop due to non fatal instability.

 

At any rate, what was your mobo again?

X570S Torpedo.

 

I've done 1000% pass with Memtest and no errors are showing up, not even WHEA. (I can't use XMP with my RAM because it's that crazy 4400 Patriot Viper stuff; I went as far as 3600 and kept everything 1:1.)

[filpo]

3 minutes ago, Nanoray said:

4070, both times.

How's cpu utilisation these times?

[Nanoray]

1 minute ago, filpo said:

How's cpu utilisation these times?

Significantly lower with the X 3D.

[filpo]

2 minutes ago, Nanoray said:

Significantly lower with the X 3D.

So both the cpu and gpu utilisation was lower with the 5600? Weird. Can u try a cpu reseat as maybe some of the pins didn't make good contact

[Nanoray]

25 minutes ago, filpo said:

So both the cpu and gpu utilisation was lower with the 5600? Weird. Can u try a cpu reseat as maybe some of the pins didn't make good contact

CPU is 100% in properly - I give it a wiggle before I locked the lever.

[pf100]

40 minutes ago, Nanoray said:

NH-D15S, with an extra 120 mm on the front.

 

Going to try CO undervolt tomorrow - my BIOS has Kombo strike, but I think MSI's latest BIOS lets you use CO for the X 3D anyway.

My plan is to use -25 for my "starred" cores and -30 for everything else, then kick it in the nads with some low-load OCCT core cycling.

I have an NH-D15S with a 120mm fan in front too.

I upgraded a 5800x to a 5800X3D because a family member needed a cpu so I gave them the 5800x, and the 5800x3d was noticeably faster. Something is obviously wrong with your system somehow. With gpu-z check to see if the pcie slot the gpu is plugged into is running in pcie4 mode and not pcie1.1 or something because I had that problem once with the 5800x. Also, since you have kombo strike all you need to do is set it to "kombo strike 3" setting and forget about curve optimizer. It works better and runs cooler than using curve optimizer but not by a huge amount. In fact, whether you use pbo, or curve optimizer, or kombo strike or a combination of all 3 or none at all, the 5800x3d should be noticeably faster than the 5600 by just setting xmp and that's all. Check everything you can - temps, speeds, everything.

[Nanoray]

25 minutes ago, pf100 said:

With gpu-z check to see if the pcie slot the gpu is plugged into is running in pcie4 mode and not pcie1.1 or something because I had that problem once with the 5800x.

Just checked the link speed in Device Manager and it says 4.

 

25 minutes ago, pf100 said:

Check everything you can - temps, speeds, everything.

It gets to 80 (at least) in CB23, but nowhere near that in games.

 

Tbh, I'm on the brink of selling all my AM4 kit and going to a 7600 on an AM5 board.

[Nanoray]

Just been in the BIOS and forced Gen 4 PCIe, disabled BAR, and enabled 1:1 UCLK (which I forgot to do).

No difference.

 

Tempted to put my 5600 back in to see how it does - I copied the BIOS settings precisely before I removed it, and I can key them all in to match...

[filpo]

8 hours ago, Nanoray said:

Tempted to put my 5600 back in to see how it does - I copied the BIOS settings precisely before I removed it, and I can key them all in to match...

Ye I'd try that

 

8 hours ago, Nanoray said:

Just been in the BIOS and forced Gen 4 PCIe, disabled BAR, and enabled 1:1 UCLK (which I forgot to do).

did you do an undervolt on your 5800X3D? if you did then I would revert it and see

[hollyh88]

11 hours ago, Nanoray said:

Just installed my shiny new 5800X 3D and dialled in the RAM settings to precisely the same as I had with my outgoing 5600. Ran some gaming benchmarks... and it's WORSE - Superposition was down nearly 20%, Heaven was down from 4500 to 4250, and Shadow of the Tomb Raider came out within margin of error, but with admittedly much better CPU frametimes (and it went from 41% GPU bound to 70%).

 

BIOS has just been updated to latest stable, and aside from the RAM overclock, I have not touched any voltages - the 5600 and 5800X 3D are being compared stock for stock.

 

What on earth have I done/not done wrong?! I can't tell you how gutted I am to have spent £240 to get a WORSE gaming performance than a boggo 5600 I can keep hold of for nothing!

 

M.

so you tried to actually bench your games as well right besides traditional tomb raider and the like?
it can be better than a 5600x but it doesnt have to be. in some games its just a limitation via the gpu.

anyway,

how i set it up in my pc also on an msi a b450 gaming pro carbon ac though

combo strike 3 (thats a -30)
cbs cppc preffered cores = disable
cbs global c state control = enabled
i set the max watt to 130
i first set my xmp profile to that of 3600 then i manually set the blck to 100 and then changed the ram speed and infinity fabric to be 1 to 1 aka 3600 and 1800.
 

it now runs at 4450 all the time during gaming and roughly 4350 all core during stuff like cinebench on cinebench i get a 14506 and single a 1467

 

but thats not why you get this cpu you get this cpu for its gaming perf. and more noticeably its 1% low gaming perf. games that used to have frame dips are now gone (coming from a 3800xt) and my framerate has almost doubled in many games with my 6950xt compared to my previous cpu.

with you it may be a bit less of a fps increase as you would like. also keep in mind your gpu. you will see more of a perf increase with this cpu with a better gpu. for those increased 1% lows. 

[Nanoray]

1 hour ago, filpo said:

Ye I'd try that

 

did you do an undervolt on your 5800X3D? if you did then I would revert it and see

That was stock; just did a -25/-30 and got roughly equivalent CB23, but 10 degrees cooler.

Left it core cycling on SSE for 90 mins while I went to town, with Youtube running alongside, and no errors reported.

 

I've made my mind up on AM5 - Prime Week plus my Amazon employee discount should save me a few quid.

[hollyh88]

1 hour ago, Nanoray said:

That was stock; just did a -25/-30 and got roughly equivalent CB23, but 10 degrees cooler.

Left it core cycling on SSE for 90 mins while I went to town, with Youtube running alongside, and no errors reported.

 

I've made my mind up on AM5 - Prime Week plus my Amazon employee discount should save me a few quid.

a 5800x3d will perform just as good as a 7600x in many titles or even better. so since you already have the 5800x3d i really dont see why you would want to spend more for no reason. especially when you only have a 4070 making the difference out to be even less.

of course it is your money at the end of the day but im telling you it wont make much of a difference or in some cases you would even lose performance. 

[Nanoray]

10 hours ago, hollyh88 said:

a 5800x3d will perform just as good as a 7600x in many titles or even better. so since you already have the 5800x3d i really dont see why you would want to spend more for no reason. especially when you only have a 4070 making the difference out to be even less.

of course it is your money at the end of the day but im telling you it wont make much of a difference or in some cases you would even lose performance. 

It's just frustrating me that it hasn't given me the bump I was expecting - I messed with the CO and it made little difference; multi-core improved a bit, but single-core went DOWN.

 

I'm going for a basic AM5 - 7600 (non-X), B650 Edge Gaming, DDR5 6000/32 (running at lower bandwidth for tighter timings).

 

Went mental and also ordered a 420 Liquid Freezer... and a Meshify 2 to put it in!

[hollyh88]

15 minutes ago, Nanoray said:

It's just frustrating me that it hasn't given me the bump I was expecting - I messed with the CO and it made little difference; multi-core improved a bit, but single-core went DOWN.

 

I'm going for a basic AM5 - 7600 (non-X), B650 Edge Gaming, DDR5 6000/32 (running at lower bandwidth for tighter timings).

 

Went mental and also ordered a 420 Liquid Freezer... and a Meshify 2 to put it in!

as i said it is your money but im already letting you know that you just spend a lot more for no performance gain at all. and putting a 420 on a 7600 is very overkill. i would suggest asking for advice from people next time so you know what the best products are to buy so you dont throw your money in the wind for no reason. 

regardless, i hope you will enjoy your new cpu! you should mark a comment as the solution so this forum post can be closed  whether its someone elses mine or yourself.

[Damocles Ti]

Have you tried reinstalling the AMD chipset drivers?
Can you check if your PBO has set the power limits the same on both CPUs? Multicore really uses power and that becomes the limiting factor

PBO curve primarily tunes frequency per given voltage, but you need to also extend the max frequency offset to actually reach higher clocks, if you don't it hits the frequency ceiling and lowers power draw.

 

[freeagent]

You should reinstall windows, but if you give up then so be it.

[Mark Kaine]

same thing here going from 3600 to 5800x3D most benchmarks got *worse* lol... i also see no diff editing videos and such...

 

thing is,  it should give you more stable frametimes and a bit better fps overall (most benchmarks don't show this)

 

Also i could swear 3600 feels "snappier" than 3200 for the ram (which also shows in benchmarks as lower latency) 

 

 

(ps: my 3600 also used more power, 130w vs 115w, which is surprising) 

 

i think the mistake with 5800x3D was that they didn't provide pbo out of the box as it really did so much on my 3600... i miss it lol

[Mark Kaine]

ps: best way to check if something is wrong is cb23.... i got a significant bump there compared to the 3600... (forgot how much  but it was like 1000 points more than in most "reviews" iirc)

 

ps2: also i got a hunch you messed up a some BIOS settings,  you really should leave it at default except XMP simply.

 

i did except i oc my RAM so i had to raise SOC a little bit too (iirc)