[SOLVED] Gigabyte B450 Aorus Elite

MetalMatty

Honorable
Apr 20, 2017
76
2
10,535
So, this is kind of a two part question but both deal with overclocking on this specific board.

So, firstly... I was running a Ryzen 2200G with this board for around two years. About a year-ish ago, I decided to overclock the CPU. Read a bunch of guides, blah blah blah.
Went into BIOS and realized there is no way to change the CPU Voltage, other than this Dynamic thing.
In any event, I left all of that alone, and the only settings I changed was XMP mode (which was on since day 1 for the RAM), and I bumped the CPU to 4.0GHz. It held stable, and CPU VDD in HWMonitor showed that it would hold right around 1.35-1.4V when it needed to. All was well, that overclock was stable and everyone was happy.

Fast forward to last week, I installed a Ryzen 5 1600AF. The CPU swap of course caused the BIOS to do its soft reset, so I reenabled XMP, and did the same exact thing I did with my 2200G. I first tried bumping speed up to 4.0GHz, which is where I would have been a happy camper.
Booted up fine, opened HWMonitor, and it's showing the CPU VDD is only 1.125 max. Start the P95 test, and CPU VDD drops down to 1.08 and system locks up. I had to drop down to 3.8GHz to get it stable, and CPU VDD is still 1.125 idle, 1.08 stressed.

I read the couple guides there was, and bumped the Dynamic Vcore DVID +.204, figuring this should put the VDD in the 1.3-1.4 range needed. It doesn't change the CPU VDD at all. It does change the CPU VCORE, but doing this doesn't seem to add any sort of stability, or it IS stable but overheating and crashing, not 100% sure. CPU temps stay under 84, but the motherboard temps get high, or what I perceive to be high.

So I guess my questions...
1.) Am I looking at the right monitors in HWMonitor (as in, I am using CPU VDD as what to monitor for voltage)?
2.) Why with the 2200G did the CPU VDD automatically go up to 1.4V when needed, but the 1600AF refuses to go over 1.125?
3.) Am I even doing this right at all, am I stupid, and is this specific Gigabyte board just not a good one to OC on?

Thanks y'all,

- Matt
 
Solution
So, this is kind of a two part question but both deal with overclocking on this specific board.

So, firstly... I was running a Ryzen 2200G with this board for around two years. About a year-ish ago, I decided to overclock the CPU. Read a bunch of guides, blah blah blah.
Went into BIOS and realized there is no way to change the CPU Voltage, other than this Dynamic thing.
In any event, I left all of that alone, and the only settings I changed was XMP mode (which was on since day 1 for the RAM), and I bumped the CPU to 4.0GHz. It held stable, and CPU VDD in HWMonitor showed that it would hold right around 1.35-1.4V when it needed to. All was well, that overclock was stable and everyone was happy.

Fast forward to last week, I...
So, this is kind of a two part question but both deal with overclocking on this specific board.

So, firstly... I was running a Ryzen 2200G with this board for around two years. About a year-ish ago, I decided to overclock the CPU. Read a bunch of guides, blah blah blah.
Went into BIOS and realized there is no way to change the CPU Voltage, other than this Dynamic thing.
In any event, I left all of that alone, and the only settings I changed was XMP mode (which was on since day 1 for the RAM), and I bumped the CPU to 4.0GHz. It held stable, and CPU VDD in HWMonitor showed that it would hold right around 1.35-1.4V when it needed to. All was well, that overclock was stable and everyone was happy.

Fast forward to last week, I installed a Ryzen 5 1600AF. The CPU swap of course caused the BIOS to do its soft reset, so I reenabled XMP, and did the same exact thing I did with my 2200G. I first tried bumping speed up to 4.0GHz, which is where I would have been a happy camper.
Booted up fine, opened HWMonitor, and it's showing the CPU VDD is only 1.125 max. Start the P95 test, and CPU VDD drops down to 1.08 and system locks up. I had to drop down to 3.8GHz to get it stable, and CPU VDD is still 1.125 idle, 1.08 stressed.

I read the couple guides there was, and bumped the Dynamic Vcore DVID +.204, figuring this should put the VDD in the 1.3-1.4 range needed. It doesn't change the CPU VDD at all. It does change the CPU VCORE, but doing this doesn't seem to add any sort of stability, or it IS stable but overheating and crashing, not 100% sure. CPU temps stay under 84, but the motherboard temps get high, or what I perceive to be high.

So I guess my questions...
1.) Am I looking at the right monitors in HWMonitor (as in, I am using CPU VDD as what to monitor for voltage)?
2.) Why with the 2200G did the CPU VDD automatically go up to 1.4V when needed, but the 1600AF refuses to go over 1.125?
3.) Am I even doing this right at all, am I stupid, and is this specific Gigabyte board just not a good one to OC on?

Thanks y'all,

- Matt
Scrap HWMonitor. It's never been very useful for Ryzen.

Instead get HWInfo64. The core voltage reading you want to pay closest attention to is CPU Core Voltage (STI2 TFN). That's the internal core voltage in the cores. The other one is VCore, in the section for the Super IO chip. That's the output of the VRM measured somewhere along the traces between the VRM and the base of the socket. The difference between the two is used by extreme overclockers to gage your load line to set an LLC value.

The 2200G is based on Zen1 (Ryzen 1000), the 1600AF is based on Zen1.5 (Ryzen 2000). The2200G has a GPU on the same die. These are very different processors, it's difficult and pointless to compare.
 
Solution

MetalMatty

Honorable
Apr 20, 2017
76
2
10,535
Scrap HWMonitor. It's never been very useful for Ryzen.

Instead get HWInfo64. The core voltage reading you want to pay closest attention to is CPU Core Voltage (STI2 TFN). That's the internal core voltage in the cores. The other one is VCore, in the section for the Super IO chip. That's the output of the VRM measured somewhere along the traces between the VRM and the base of the socket. The difference between the two is used by extreme overclockers to gage your load line to set an LLC value.

The 2200G is based on Zen1 (Ryzen 1000), the 1600AF is based on Zen1.5 (Ryzen 2000). The2200G has a GPU on the same die. These are very different processors, it's difficult and pointless to compare.

The comparison wasn't so much in the CPUs themselves, it was more wondering why the system would raise voltage to 1.4v with the 2200G, but refuses to do so with the 1600AF. There is no option on this board to manually raise voltage to a set amount, just to increase or decrease under Dynamic, so it wasn't done for either CPU.

HWInfo is showing the same numbers as Monitor was.

I honestly think the problem may just be heat. I think this board is getting too hot and that's what is causing the crash at 3.9GHz+. CPU temps are fine, but the motherboard temps (they are poorly labled on Monitor), were quite high.

But yes, with HWInfo, CPU Core Voltage is showing 1.125, and VCore is showing 1.2V (I believe I have it set to +.102 in BIOS, so that would make sense).

Which of the motherboard temp monitors are quite important to monitor, and what is their temp limit? I assume the VRM is the one that really needs an eye kept on it, but I'm not sure what a safe temp would be for it.

At 3.9GHz, it was running the stress test fine for about 35 minutes, then it locked up. I'm wondering if board temps simply got too high and it went into its safety mode.
 

MetalMatty

Honorable
Apr 20, 2017
76
2
10,535
So under a stress test using HWInfo, my VCore is showing 1.2V.
So, logically, assuming I can keep temps under control, I have .2v of room left to increase Dynamic Voltage setting in BIOS.
 
The comparison wasn't so much in the CPUs themselves, it was more wondering why the system would raise voltage to 1.4v with the 2200G, but refuses to do so with the 1600AF.
....

Comparing voltage requirements of one CPU to the other, or it appeared to be.

1.125V (SVI2/TFN) is very, very low especially if at load. I would expect something closer to 1.3V...like 1.275 or so. It's OK for voltage to rise up to 1.35 or even to 1.4 at low/no load.

Tdie is the important temperature to monitor; also Tctl but I believe they should be the same. It might be shown as Tdie/Tctl for a 1600.
 

MetalMatty

Honorable
Apr 20, 2017
76
2
10,535
You're comparing voltage requirements of one CPU to the other. 1.125V (SVI2/TFN) is very, very low especially if at load. I would expect something closer to 1.3V...like 1.275 or so.

Tdie is the important temperature to monitor; also Tctl but I believe they should be the same.
ooTb86w.png


This is after roughly, and only, 10-15 minutes on the hardest P95 stress test. Those are the two sections I'm paying attention to for voltage, and "VRM MOS" for motherboard temp, and obviously CPU TCTL temps for CPU temp.
 
..

This is after roughly, and only, 10-15 minutes on the hardest P95 stress test.
...
Temperature of both VRM and CPU (Tctl/Tdie) are way, way low for P95 stress test.

What cooling do you have? stock Wraith cooler?

Feel free to raise voltage to keep stable at higher clocks so long as temperature can be kept in control. By 'in control' I mean in the mid 80's for that P95 stress test, which is unworldly and never going to be seen in use.

Tjmax for the CPU is 95C, so you have that much margin there.

As I said, I'd expect to see SVI2/TFN core voltage to be around 1.275-1.300 Volts under stress test like P95, so you have that much margin there to play in.
 

MetalMatty

Honorable
Apr 20, 2017
76
2
10,535
Temperature of both VRM and CPU (Tctl/Tdie) are way, way low for P95 stress test.

What cooling do you have? stock Wraith cooler?

Feel free to raise voltage to keep stable at higher clocks so long as temperature can be kept in control. By 'in control' i mean in the mid 80's for a P95 stress test, which is unworldly and never going to be seen in use.

As I said, I'd expect to see SVI2/TFN core voltage to be around 1.275-1.300 Volts under stress test like P95, so you have that much margin there to play in.
So, switching to HWInfo was the best advice you could have given me. HWMonitor had to be showing me incorrect numbers, because now everything is going quite well at 4.0GHz and I can clearly see what I'm doing and what I'm changing.

I have an Enermax Aquafusion 240mm.

I'm bumped up to 4.0GHz now, gonna stress it for a half hour, if it holds and temps stay good, going to see how high we can get it.

Apparently the only advice I really needed was "Scrap HWMonitor and get HWInfo" lol.

Thanks for your help bud, I really appreciate it. That's why I love this forum so much.
 

MetalMatty

Honorable
Apr 20, 2017
76
2
10,535
That AIO should be overkill in the extreme for this CPU :) So you will have a lot of cooling margin to really push the CPU hard and still keep temps in control!

Have fun...
Fantastic AIO, but more importantly hands down the best customer service I've ever experienced. I sent their Facebook page a message at 1AM EST and they replied back within 20 minutes AND sent me some odd ball pictures I had requested.

I ended up crashing out after 20 minutes, so I bumped the voltage up a bit, it's holding pretty steady at 1.28 now, so far so good. CPU temp is and has been steady at 75c and VRM has finally (at least so far) flatlined at 90c. High, but I'll be happy as long as it's under 100c, though I think 95c is going to be my own personal limit.
 
.... High, but I'll be happy as long as it's under 100c, though I think 95c is going to be my own personal limit.
VRM's have a ton more thermal margin than CPU's, 125C is what most FET's are rated, so 95-100C is certainly 'safe'. But also Prime95 and especially small FFT's is so unworldy unreal; it's not what you'll ever see. I don't see that kind of stressy processing load even running Folding@Home which is a terror on Ryzen processors.
 

MetalMatty

Honorable
Apr 20, 2017
76
2
10,535
Decided to keep it at 4.0GHz. That was my original goal. I could have hit 4.1 I think, but I'm happy at 4.0GHz.

Fun fact:

CinebenchR20 results for...
2200G at 4.0GHz was 1439 points.
1600AF at stock clock was 2693.
1600AF at 3.8GHz was 2839.
1600AF at 4.0GHz is 3038.