I bought an AMD 6800XT that runs worse than my GTX 1080. Am I going nuts?

[Craftyawesome]

46 minutes ago, Action_Johnson said:

I've never seen results from an OC that made me think "that seems totally worth it"

On what systems? Ryzen single core is often higher than multicore at reasonable voltage. Modern intel has a little more room, and older intels have quite a bit of overclockability. Even as recent as the 6600k, 25% up from the all core turbo was common.

[SeanTwig]

7 minutes ago, Mister Woof said:

Throw some positive airflow over those VRMs, and you should be okay. 

 

Even if your VRMs get really hot/unstable in those synthetic workloads, the reality is that for most gaming applications, even heavy ones, it won't get close to as hot as that.

I have a fan blowing at my case as it's open on one side.  I need to get a new case, so at the moment I am running this fan pointed at the Mobo, which I think is helping quite a bit.  Current ambient is like 20-25C.

 

I ran CB15 and got a score of 1516.  I watched my VRM in HWiNFO64 and it got to max 57C.  Is this too hot?  I feel like that's well cool enough.  

 

When you say to increase the limits in XTU, which ones would that be?  I have ones for Turbo Short Power Max, Processor Core IccMax, and Turbo Boost Power Max.  Do I touch the Boost power window?  It's set to 8 Secs.  Only thing I have changed is adjusted all cores to run at 50X multiplier (only the first 2 were set at that).  I haven't touched the Cache

[Mister Woof]

32 minutes ago, SeanTwig said:

I have a fan blowing at my case as it's open on one side.  I need to get a new case, so at the moment I am running this fan pointed at the Mobo, which I think is helping quite a bit.  Current ambient is like 20-25C.

 

I ran CB15 and got a score of 1516.  I watched my VRM in HWiNFO64 and it got to max 57C.  Is this too hot?  I feel like that's well cool enough.  

 

When you say to increase the limits in XTU, which ones would that be?  I have ones for Turbo Short Power Max, Processor Core IccMax, and Turbo Boost Power Max.  Do I touch the Boost power window?  It's set to 8 Secs.  Only thing I have changed is adjusted all cores to run at 50X multiplier (only the first 2 were set at that).  I haven't touched the Cache

I haven't used xtu in a long time; you can do that in bios

 

 

[SeanTwig]

3 minutes ago, Mister Woof said:

I haven't used xtu in a long time; you can do that in bios

 

My BIOS is super buggy for some reason.  I have a feeling I need to flash it, however I read some people had issues with some Gigabyte BIOS flashes and bricked their boards.  Knowing my luck, I will have the same thing happen to me, so I decided to just leave as is.  Problem is that some of the options don't display properly.  You can still change them, but the don't give you a visual indication that you are changing anything.

 

In other news, I have started to increase the Power (W) max from 95W to 97W, each time running a Cinebench15 and recording all my voltages and clock speeds.  I think I am going to do that incrementally until I am getting no power limit errors.  It keeps throttling and in XTU under the Mhz Max, it seems to sit at 4.2Ghz max

[Mister Woof]

I have primarily Gigabyte motherboards - and haven't had any issues with flashing.

 

I've probably flashed my Gigabyte Z370 Aorus Gaming 5 about 4 or 5 times without problems. Do it in BIOS, not in windows.

 

 

58 minutes ago, SeanTwig said:

My BIOS is super buggy for some reason.  I have a feeling I need to flash it, however I read some people had issues with some Gigabyte BIOS flashes and bricked their boards.  Knowing my luck, I will have the same thing happen to me, so I decided to just leave as is.  Problem is that some of the options don't display properly.  You can still change them, but the don't give you a visual indication that you are changing anything.

 

In other news, I have started to increase the Power (W) max from 95W to 97W, each time running a Cinebench15 and recording all my voltages and clock speeds.  I think I am going to do that incrementally until I am getting no power limit errors.  It keeps throttling and in XTU under the Mhz Max, it seems to sit at 4.2Ghz max

 

[Bombastinator]

2 hours ago, SeanTwig said:

The only thing that's a bit annoying, is that I can't seem to get my CPU to ever get above 60C during Cinebench15.  I have so much room to increase performance.  Even if only a little.  Just seems like such a waste, buying a K CPU and leaving it at stock.  Lordy I wish I was better at this sort of thing

The majority of k chips sold are never overclocked at all.  None the less @Mister Woof makes some points. It’s doable if you really want to.  You don’t need to right now though.

[SeanTwig]

23 minutes ago, Bombastinator said:

The majority of k chips sold are never overclocked at all.  None the less @Mister Woof makes some points. It’s doable if you really want to.  You don’t need to right now though.

I hear ya, but now my interest is piqued.  I need to learn lol.

 

Atm I have turned the max power up to 200W which got rid of the error "Power Limit Throttling", but I still get "Current/EDP Limit Throttling".  

 

The thing that I see that makes me think something is set wrong is when I run a benchmark, I can't get the CPU usage to stay around 100%.  It spikes up to 100% and then almost immediately drops down.  In the Screenshot you can see where I ran the bench, but it never stays up, yet it's not thermal limits because that only got to 77C that run.

 

Is Overclocking REALLY this hard?  I feel like either I am stupid (don't answer that), or I am overthinking things or misunderstanding some setting somewhere.

[Bitter]

You need to raise the CPU current limit in BIOS, if your Z370 board has good enough VRM this won't be a problem but keep an eye on VRM temps if they're reported to monitoring software like hwinfo. The fact that the board was setup for the default limits for CPU wattage is concerning, most enthusiast boards are defaulted with higher limits but maybe Gigabyte is different. I haven't owned a recent Gigabyte board for Intel. Anyway, a Z370 board should be 'smart' enough to throttle the CPU if the VRM gets too hot, you'll see XTU showing if it's throttling but it may indicate something like thermal throttle event even though CPU temps are fine. That would be the VRM overheating. 5Ghz with a good cooler isn't a problem, on the stock cooler at 100% fan speed you should be good to get all core to 4.5 maybe 4.6 if you're delicate with voltage and load line calibration settings and have exceptional case cooling with low ambient temps. Even a pretty basic $30-40 tower cooler is going to absolutely blow the stock cooler out of the water by a huge margin and easily allow all core clocks past 4.6 and likely let you run single core clocks to nearer 5ghz. If you find that you're really thermally constrained or the VRM isn't happy with running 6 cores at higher speeds you can disable a couple cores and run 4 cores at higher clock speeds and may see better results in some games doing that.

[Bombastinator]

25 minutes ago, Bitter said:

You need to raise the CPU current limit in BIOS, if your Z370 board has good enough VRM this won't be a problem but keep an eye on VRM temps if they're reported to monitoring software like hwinfo. The fact that the board was setup for the default limits for CPU wattage is concerning, most enthusiast boards are defaulted with higher limits but maybe Gigabyte is different. I haven't owned a recent Gigabyte board for Intel. Anyway, a Z370 board should be 'smart' enough to throttle the CPU if the VRM gets too hot, you'll see XTU showing if it's throttling but it may indicate something like thermal throttle event even though CPU temps are fine. That would be the VRM overheating. 5Ghz with a good cooler isn't a problem, on the stock cooler at 100% fan speed you should be good to get all core to 4.5 maybe 4.6 if you're delicate with voltage and load line calibration settings and have exceptional case cooling with low ambient temps. Even a pretty basic $30-40 tower cooler is going to absolutely blow the stock cooler out of the water by a huge margin and easily allow all core clocks past 4.6 and likely let you run single core clocks to nearer 5ghz. If you find that you're really thermally constrained or the VRM isn't happy with running 6 cores at higher speeds you can disable a couple cores and run 4 cores at higher clock speeds and may see better results in some games doing that.

I haven’t read all 5 pages of this thread, but the impression I get is one of the problems is that the VRM is weak.  It’s going to need active VRM cooling of some sort like a downdraft cpu cooler or a separate fan to get strong enough.  That’s not undoable though.  I don’t know what cpu cooler he is using.  If it is a stock cooler it’s a bit worrisome because stock coolers already are downdraft.

[SeanTwig]

24 minutes ago, Bitter said:

The fact that the board was setup for the default limits for CPU wattage is concerning, most enthusiast boards are defaulted with higher limits but maybe Gigabyte is different.

If you're talking about the 200W limit, that was 95W (which is stock).  I raised that in order to not throttle.  I find that at idle, I am running 4.9-5Ghz, but when I bench, it drops to 4.2Ghz?  That seems counter-intuitive to what I understand.  Raising the wattage to 200W seemed to get that to stay around 5Ghz or at least 4.8Ghz during load. 

 

In real world game testing, I ran Ac: O, which runs better for sure, however I still get CPU usage sitting around 70% and same with GPU usage?  Maybe that's just a poorly optimised title.  Fps generally sits around 70-90 now though so that's a 10-15fps improvement.  Thing is, for the money I spent on this CPU, I can't help but feel I'm leaving a little left in the box.

 

28 minutes ago, Bitter said:

You need to raise the CPU current limit in BIOS, if your Z370 board has good enough VRM this won't be a problem but keep an eye on VRM temps if they're reported to monitoring software like hwinfo.

I am going to try to find where that is in my BIOS and try increase it.  Is there a setting that's preferred?  VRM maxed out at 80C while running around in Ac: O.  It generally sits at around 45C at idle

[SeanTwig]

2 minutes ago, Bombastinator said:

I haven’t read all 5 pages of this thread, but the impression I get is one of the problems is that the VRM is weak.  It’s going to need active VRM cooling of some sort like a downdraft cpu cooler or a separate fan to get strong enough.  That’s not undoable though.  I don’t know what cpu cooler he is using.  If it is a stock cooler it’s a bit worrisome because stock coolers already are downdraft.

TLDR:  I bought a Corsair H100i and I am running a desk fan pointed at my Mobo to inject with some more air.  Ambient in my room is around 20-23C atm.  The thing is, I don't see any VRM throttling in HWiNFO64 when I bench or game.  Temps don't even go that high really

[Bombastinator]

1 minute ago, SeanTwig said:

TLDR:  I bought a Corsair H100i and I am running a desk fan pointed at my Mobo to inject with some more air.  Ambient in my room is around 20-23C atm.  The thing is, I don't see any VRM throttling in HWiNFO64 when I bench or game.  Temps don't even go that high really

So lots of cpu cooling but no VRM cooling at all other than whatever your case fans happen to hit it with.  No VRM throttling is no VRM throttling though.  I’m still tempted to point a spare random case fan at the VRM if there happens to be one sitting around.  Myself I’ve got a box of old ones in my closet but not everyone is like that. 

[SeanTwig]

3 minutes ago, Bombastinator said:

I’m still tempted to point a spare random case fan at the VRM if there happens to be one sitting around.

I have this huge one pointed right at it.  It's actually quite cold in my case.  VRM while in Tarkov is sitting at.... 47C  Like that's the thing, I'd understand if I am getting temps of like 90C+ but neit.

 

I would say it's strange, but at this point, everything is strange lmao

[Bitter]

Games are bad for looking at temps and power because the load is constantly changing. Use Prime95 torture test with small fft to establish power draw and temps under max CPU load, it'll never be that bad when gaming but it'll be a steady load and the same load each time you do testing.

[SeanTwig]

10 minutes ago, Bitter said:

Games are bad for looking at temps and power because the load is constantly changing. Use Prime95 torture test with small fft to establish power draw and temps under max CPU load, it'll never be that bad when gaming but it'll be a steady load and the same load each time you do testing.

Jesus!  This thing doesn't fuck around!  It's the first benchmark tool that has actually got my temps up.

 

This is a screen of HWiNFO64 after the stress.  I just used the small like you noted.  Not the VRM going up to 127C and CPU maxing at 93C.  Does this basically mean that I am limited by the Mobo?

 

 

[Bitter]

That's pretty hot, yeah you need more VRM cooling but 1.48 volts is a little high for an 8700K I think? I'm not super versed in 8th gen overclocking but I seem to recall you'd want to be more around 1.375 to 1.385 range? Like I said this isn't my area of expertise and I missed a few things between page 1 and page 5 but 126C on the VRM is pretty smokin hot and while it's not throttling it's not good. You might need to use the good old finger tip sensor to make sure the VRM heat sinks are actually getting hot just be careful as 126C will hurt you, I'd advise putting a finger on the heat sink and THEN running the stress test rather than poking it once it's already hot. Might jerk your hand away and boop something if you do that. You need to know if the heat is making it into the heat sink, it should be terribly hot pretty fast. May be a thermal pad issue there causing poor heat transfer if it never feels that hot or takes a long time to get hot. Circling back around, if you can get it stable with lower voltage that'll also drop the power draw through the VRM and lower the CPU temps as well and in general be easier on the system. 5Ghz is a nice OC but sometimes 4.8 or 4.9 will get you the same frames for A LOT less power/heat which is a better gaming experience.

 

I guess I should ask, what is the current problem since 5 pages is a lot to skim.

[Bombastinator]

45 minutes ago, SeanTwig said:

I have this huge one pointed right at it.  It's actually quite cold in my case.  VRM while in Tarkov is sitting at.... 47C  Like that's the thing, I'd understand if I am getting temps of like 90C+ but neit.

 

I would say it's strange, but at this point, everything is strange lmao

I see.  I thought the big fan was just pointed at the computer in general.  Perhaps the case cover is off and the fan is pointing through that.  so bumping the VRM a bit harder shouldn’t be no thing.  The bit that can be worrisome with temp sensors though is they only measure the point they measure at.  

[Bombastinator]

45 minutes ago, SeanTwig said:

Jesus!  This thing doesn't fuck around!  It's the first benchmark tool that has actually got my temps up.

 

This is a screen of HWiNFO64 after the stress.  I just used the small like you noted.  Not the VRM going up to 127C and CPU maxing at 93C.  Does this basically mean that I am limited by the Mobo?

 

 

Heh.  Yeah.  Prime95 is a power virus for CPUs.  It will eat everything you got no matter how much you got.  It was originally written by an academic as a sort of folding@home just for him.  He was greedy though and the thing is made to take every ounce of cpu power it can.  This wound up being useful for torture testing and he noticed this so he rewrote it to work better for that.   There is a not dissimilar app for GPUs called fur mark that will grind a gpu till there’s nothing left to grind.   The two had a habit of destroying hardware before safeties got put in to keep that from happening.  They’re both quite old.

 

I got the impression that the mobo being the issue was old news. 

[Craftyawesome]

55 minutes ago, SeanTwig said:

Jesus!  This thing doesn't fuck around!  It's the first benchmark tool that has actually got my temps up.

 

This is a screen of HWiNFO64 after the stress.  I just used the small like you noted.  Not the VRM going up to 127C and CPU maxing at 93C.  Does this basically mean that I am limited by the Mobo?

 

 

What voltage did you set? Was 1.5 vcore an accurate reading? That's way too much. Drop it to around 1.36v, maybe a little higher if you are running cool.

 

If vcore is reading higher than what you put in the bios, then try lowering LLC.

[SeanTwig]

53 minutes ago, Bitter said:

That's pretty hot, yeah you need more VRM cooling but 1.48 volts is a little high for an 8700K I think?

I have a 9900K now.  Also, VRM only gets really hot when I run Prime (verified with finger method).  Currently I am in Tarkov and I have 46C VRM and I can hold my finger on it quite comfortably.  So that seems to be reporting correctly

57 minutes ago, Bitter said:

Circling back around, if you can get it stable with lower voltage that'll also drop the power draw through the VRM and lower the CPU temps as well and in general be easier on the system.

Yea I don't even think I had any voltage added?  I think I just increased the wattage limit from 95W to 200W so maybe it just draws heaps more because I removed the limit too much?  I saw on a spec sheet on Tom's Hardware where they tested and said anything from 150W to 200W is normal draw under load.

41 minutes ago, Bombastinator said:

I see.  I thought the big fan was just pointed at the computer in general.  Perhaps the case cover is off and the fan is pointing through that.  so bumping the VRM a bit harder shouldn’t be no thing.  The bit that can be worrisome with temp sensors though is they only measure the point they measure at.  

Yup I have case cover off because I had a fan making a noise so I removed it.  Now to compensate, I just have it open with a fan.  I did give it a real good clean last night with some compressed air and some paper towels etc.  It's looking all clear and really quite nice, so it's not a problem I think with dust or something.

 

My guess is I have not configured the clock speeds correctly.  All I can say is I remember Overclocking with my MSI board a few years back with the i5 6600K and it was a dream.  So easy to navigate.  This Gigabyte board is garbage.  I think next I am going Asus.

 

16 minutes ago, Craftyawesome said:

What voltage did you set? Was 1.5 vcore an accurate reading? That's way too much. Drop it to around 1.36v, maybe a little higher if you are running cool.

 

If vcore is reading higher than what you put in the bios, then try lowering LLC.

I will endeavour to find out where LLC is in my BIOS lol.  I originally set something but it crashed so I set back to default, but maybe I still have it set to something high.  I will restart now and update you

[Bitter]

Got it. Still a pretty high core voltage to be running even for a 9900K I think. A bit lower would be much less difficult on the VRM and your cooling. If the board is on just all auto defaults it's probably juicing the CPU a bit harder than it needs to be.

[SeanTwig]

Ok I apologise in advance for the picture spam but this is what I see in BIOS. 

2 minutes ago, Bitter said:

Got it. Still a pretty high core voltage to be running even for a 9900K I think.

I can see now the core voltage up real high. That's the weird thing, unless you can see something screaming at you in the screenshots, I see the Vcore set to 1.3? Or am I reading that wrong? 

[Craftyawesome]

46 minutes ago, SeanTwig said:

Ok I apologise in advance for the picture spam but this is what I see in BIOS. 

I can see now the core voltage up real high. That's the weird thing, unless you can see something screaming at you in the screenshots, I see the Vcore set to 1.3? Or am I reading that wrong? 

Are you still using XTU? Either way, auto voltages can get crazy with overclocking. Please try manual 1.36v

[SeanTwig]

18 minutes ago, Craftyawesome said:

Are you still using XTU? Either way, auto voltages can get crazy with overclocking. Please try manual 1.36v

I was using XTU yes, but I set everything back to default and I will try some manual overclocking. The only thing I don't like about doing it manually is that my BIOS assumes I already know what I'm doing and if I set something that I think is decent, it just BSoD's and then I don't know what caused it lmao 

[SeanTwig]

44 minutes ago, SeanTwig said:

Please try manual 1.36v

Went into BIOS. Started by setting everything to default. Then changed the following :

 

CPU bclk to 47

CPU Vcore to 1.36

 

Saved the profile. Saved and booted. Got to Windows fine opened up a browser to download OCCT, opened OCCT and I hadn't even clicked a button yet, I was just looking at the voltages (which were somehow STILL running at 1.44V) then my pc hung and it restarted. Like, what am I doing wrong? Something reeeaaalll fucky is going on with the voltage