Ryzen Owners Zen Garden

[Calmmo]

New chipset, new benchscores!

 

[PerfectWave]

did u improve your bench? cos i guess the new driver are because of new tr4 mobo

 

[AlienIsGOD]

grabbed the chipset drivers also on the latest BIOS for my board (F50a).

 

[Calmmo]

did u improve your bench? cos i guess the new driver are because of new tr4 mobo

yup small increase in all of them

 

[GoldenX]

Was using a 3850MHz@1.35v OC, decided to test at the stock's 1.2v, 3750MHz. Fuck it, let's go green.

 

[robot zombie]

Was using a 3850MHz@1.35v OC, decided to test at the stock's 1.2v, 3750MHz. Fuck it, let's go green.

Haha, no shit? Another .15v for 100mhz seems pretty steep!

That's actually one thing I like about Zen and Zen+ (dunno with Zen 2 yet cuz I haven't gotten my hands on more than one.) They're generally not *amazing* overclockers, but in that sweet spot, the efficiency goes way up and the amount of performance you leave on the table is peanuts. There just is no need or reason to make them run super-hot and get deep into cooling. Also means the vanilla ones don't really need much of a mobo to prop them up.

I suppose you could look at it two ways. They don't have that 'limitless' nature to them where they can just keep sucking up watts and vomiting clocks. But for what they're meant to do, they don't really ask for much to shine and when they do, something just isn't set up right. I think it's probably a letdown to some people that you're not really *supposed* to push Ryzen CPU's, but to me it's like, what if you just don't need to?

Still seems that way. TBF I'm running a 3900X, but a fairly middling, single-fan tower cooler is getting the job done. Not really seeing it go over 100W often. But it scores over 7000 on CB, so I mean... for a 12-core chip to generally run under 100w but still have that ability is nice. As quirky as their boost system has always been, it is a nice compliment to what I might call their easygoing silicon. It likes to just chug along, sipping steady power.

It's pretty much a matter of bio-diesel or massive gas engine.

I've never had an Intel system myself, but I've built a good bit of them and one thing I always hated about working with the midrange and high-end SKUs was how they were just a *bitch* to cool if you wanted to really tap into them. I like silence AND performance in my builds (even if they're not MY builds.) So having a CPU that almost seems to want to run on the least power you can feed it to hit decent clocks almost makes more sense than an Intel power-hog. It almost doesn't matter if we're talking budget or high-end. Ryzen just is more manageable when it comes to tapping into the performance available without destroying 'power budget' as a concept.

Cruel irony that AMD's GPUs are the total opposite. Maybe they'll start to work that out one day.

 

[Calmmo]

So i've had 3 reboots @ 1867 , 2 in regular OS use and 1 testing out Outer Worlds.
Cant say i'm certain if it's memory or IF tho bumping ram voltage by 10mv didn't help since I got another reboot after that (no memtest errors in 2 1hr tests)
I'm assuming it's IF. Could upping SoC voltage a little have any impact on IF stability? running 1.1 atm
Need to figure out a way to stabilize system run before going into tightening timings like tct Trfc and the like
(i'm, running a pcie4 nvme mp600 also)

trying 1.125 for now see if it works out i guess

 

[okidna]

1.0.0.4 B is out for Strix X470-F : https://www.asus.com/Motherboards/ROG-STRIX-X470-F-GAMING/HelpDesk_BIOS/

 

[Wickedt]

What voltage is your ram at? I had same type of issues till i raised dram voltage to 1.46

 

[Mussels]

My STRIX B450-I still doesnt have it

 

[Lorec]

So I finally boot up my 3950x.
I have a vanilla C8H (0506 bios) mobo, and first what struck me was huge core voltage

I read through all posts from june to now in this thread and it seems its normal thing...
I have a custom loop, temps get kinda jumpy even in idle.
Aside from that had no problem with ram, running smoothly the DOCP profile etc.

Latest 1105 update states I cant downgrade from it, which is a bit scary.
I checked around internet and 0901 update (the one prior ABBA) seems worth trying (especially I can revert if needed).
There were PCH fan curve improvements etc...
Unless it wouldnt change much that is.

My goal is to get some manageable voltages here...
Any thoughts?

 

[TheLostSwede]

Anything higher is limited by the CPU.

You can't even lower the TRC? As that's quite high. What about the TRFC?

So i've had 3 reboots @ 1867 , 2 in regular OS use and 1 testing out Outer Worlds.
Cant say i'm certain if it's memory or IF tho bumping ram voltage by 10mv didn't help since I got another reboot after that (no memtest errors in 2 1hr tests)
I'm assuming it's IF. Could upping SoC voltage a little have any impact on IF stability? running 1.1 atm
Need to figure out a way to stabilize system run before going into tightening timings like tct Trfc and the like
(i'm, running a pcie4 nvme mp600 also)

trying 1.125 for now see if it works out i guess
View attachment 137666 View attachment 137665

Try increasing the TRFC, as if it's too tight, at least in my experience, you get the random reboots. Your TRFC timings are quite tight already.
Check with the DRAM calculator to see what it suggests.

So I finally boot up my 3950x.
I have a vanilla C8H (0506 bios) mobo, and first what struck me was huge core voltage
View attachment 138411
I read through all posts from june to now in this thread and it seems its normal thing...
I have a custom loop, temps get kinda jumpy even in idle.
Aside from that had no problem with ram, running smoothly the DOCP profile etc.

Latest 1105 update states I cant downgrade from it, which is a bit scary.
I checked around internet and 0901 update (the one prior ABBA) seems worth trying (especially I can revert if needed).
There were PCH fan curve improvements etc...
Unless it wouldnt change much that is.

My goal is to get some manageable voltages here...
Any thoughts?

That's normal. It'll hit 1.5V at times. This has been discussed at length all over the interweb...
Temps are jumpy when going from idle to any kind of load.
You really want AGESA 1.0.0.4 B for that chip.

 

[GoldenX]

You can't even lower the TRC? As that's quite high. What about the TRFC?

These sticks seem to be low quality B-die, so I can't complain much.

 

[TheLostSwede]

These sticks seem to be low quality B-die, so I can't complain much.

Just saying, those are two settings that you might be able to tweak a bit.

 

[Wickedt]

These sticks seem to be low quality B-die, so I can't complain much.

Did you import the rams settings from Taiphoon burner? I had some of those problems myself, i imported the settings into Ryzen dram calculator and poof, worked beautifully.

 

[GoldenX]

Yes, in manual with the html export it suggests lower values that don't seem to be stable at all (can't even post), so I'm using the "tRFC calculator" to find a lazy solution.

 

[robot zombie]

So I finally boot up my 3950x.
I have a vanilla C8H (0506 bios) mobo, and first what struck me was huge core voltage
View attachment 138411
I read through all posts from june to now in this thread and it seems its normal thing...
I have a custom loop, temps get kinda jumpy even in idle.
Aside from that had no problem with ram, running smoothly the DOCP profile etc.

Latest 1105 update states I cant downgrade from it, which is a bit scary.
I checked around internet and 0901 update (the one prior ABBA) seems worth trying (especially I can revert if needed).
There were PCH fan curve improvements etc...
Unless it wouldnt change much that is.

My goal is to get some manageable voltages here...
Any thoughts?

Easier said than done. It's more complex than this, but generally speaking, voltage, total power, temperature, utilization... all of the main ones, are locked-in on sort of a set curve. Or at least, that's the ultimate behavior. My experience is that any means you can use to try to lower voltage will also decrease boosts. Say your temperature is low, but your usage is high, but you've manually set a voltage offset. XFR and PBO don't acknowledge that you just want to lower the voltage. It sees the temp/util and wants to boost higher, but it can't because the voltage available isn't in-line with what it wants for those higher clocks. It wants what it wants to hit a given clock. Same thing happens if you go into your PBO settings and try to change the power limit. All you're doing by cutting voltage is nipping the upper straps off of the boost curve. It seems to slide all of the parameters back so that it won't boost as high as it would've otherwise. Basically, if all of the conditions aren't met, including voltage from the mobo, it absolutely will not hit its max frequency for any load. It'll always be lower in step with the voltage drop.

It's kind of crazy, how tight the power curve is. We're seeing different numbers because of mobo manufactures doing different things with their configs, but under the hood, the behavior from chip to chip is astonishingly uniform. Like, I could get another 3900x, drop it in this board, and I'm betting it'll run the same voltages, same clocks, same wattage. There really just is no wiggle to it. What it pulls out for you is pretty much all there is.

The general consensus is that it doesn't hurt anything. For overclocks, I think the common limit is 1.35-1.4v, or at least that's what I go by when I'm not okay with cooking a chip. But that's constant voltage. Ryzen's boost doesn't sustain it constantly. I'm betting it looks like it's flat-lining at 1.4-1.5 volts all of the time because the spikes and dips are too brief and slip between polling. But it's certainly not the same as running constant voltage. Guarantee you if you take that 3950x and pump 1.5v volts to it while putting it under load, the temperatures will be astronomically higher than it boosting on its own and hitting that same voltage.

The temperatures are kind of a similar thing. It's got those little spikes even under light loads... seems that when you give a Ryzen some meager tasks to much on, it likes to max the voltage out... probably because the current is so low. But again, if your cooling is up for it, not a problem. Worst case is that it stops boosting or even throttles down. But on the flipside, you might not even see much better temperatures with better cooling. If it had headroom outside of temperature, it might just eat up a good chunk of any temperature headroom you give it.. It's pretty much gonna keep going until it hits one of a few limits. You have the clock speed limit (they won't automatically run over advertised boost clocks.) And then you have wattage/current and temperature. Voltage is more of a soft limit. It has a max of 1.5, but that 1.5 doesn't limit clock speed... like it doesn't stop going higher because it wants the full 1.5. It does, however, seem to have a minimum, per a given clock speed and load. Again, to do certain clocks under certain loads, it requires that voltage. Mobo manufacturers could do better on their end, but from what I've seen the only ones could really change it much are AMD themselves. But it seems they've decided this is the optimal behavior, because they all do it

Nah, I think in your case a BIOS update might help. But the behavior you're describing is pretty typical. It's always gonna want those high voltages and little spikes. Over time I'm sure AMD will refine it and smooth it out so that maybe it doesn't spike so much and permits lower voltages... but it's not exactly a malfunction. Just really stiff rules on their end. They decide if it's a problem or not on their end. Is what it is... I feel a little conflicted, myself... but then I kinda think who am I to think I know better than AMD how their chips should run. I'm sure there are some issues with windows power management and just... all sorts of hiccups getting everything in this admittedly pretty complex boosting system to communicate and respond as intended. They've always liked to overvolt their stuff out of the box, but this is probably the first time you couldn't undervolt an AMD product that runs hot and high. There's gotta be a reason for that.

I'd say they're actually pretty good at self-regulating. They're nothing if not consistent, and while behavior takes some getting used to, I don't see one killing itself from it. It seems like their main concern is current and I assume there's a good reason for that. I've seen the voltage dip under heavy threaded workloads, when the current was the highest I've seen it. I think below a certain voltage, it's probably the combination of high current AND high voltage that kills many a CPU. You look at your voltage and see 1.5 and think gee, that's high! But when you look at your total wattage it could be like... 48. Which also means the current is very low. Hence why on the flipside, as current starts getting legitimately high, voltage will start to dip a little... more and more if the load keeps increasing. So it seems to me there is a voltage where even a small amount of current would be enough to fry a Ryzen, but 1.5v isn't it. I wish we all knew for sure, but it seems that IT at least knows how much of each it can take and will take steps to never hit a point of damage/degradation. I trust AMD has put in the footwork to figure out what voltage and current levels show signs of degradation and with their long-gathered knowledge of how CPUs behave found a safe line to draw for their boost system.

They really are a different beast, though. We're talking about a little mobile/SoC arch turned desktop powerhouse, here. All of the parameters and behavior are going to be different from what we're all used to with Intel, who's been rehashing the same chip for years. I think they're probably a lot more finicky to get the performance out of than those older, more refined architectures. Just a lot more going on with em, across the board.

And just so we're clear, that's me saying I don't really know what's going on with them but yes it is normal and no there's not much to be done

 

[biffzinker]

And just so we're clear, that's me saying I don't really know what's going on with them but yes it is normal and no there's not much to be done

I tried a -25mv offset, didn't bother the boost behaviour it was still hitting up to 4.2 GHz but when I tried a -100mv offset it didn't work out. Benchmark scores dropped, and the single core was only 4.150 GHz barely touched 4.2.

 

[Lorec]

Spoiler:

Easier said than done. It's more complex than this, but generally speaking, voltage, total power, temperature, utilization... all of the main ones, are locked-in on sort of a set curve. Or at least, that's the ultimate behavior. My experience is that any means you can use to try to lower voltage will also decrease boosts. Say your temperature is low, but your usage is high, but you've manually set a voltage offset. XFR and PBO don't acknowledge that you just want to lower the voltage. It sees the temp/util and wants to boost higher, but it can't because the voltage available isn't in-line with what it wants for those higher clocks. It wants what it wants to hit a given clock. Same thing happens if you go into your PBO settings and try to change the power limit. All you're doing by cutting voltage is nipping the upper straps off of the boost curve. It seems to slide all of the parameters back so that it won't boost as high as it would've otherwise. Basically, if all of the conditions aren't met, including voltage from the mobo, it absolutely will not hit its max frequency for any load. It'll always be lower in step with the voltage drop.

It's kind of crazy, how tight the power curve is. We're seeing different numbers because of mobo manufactures doing different things with their configs, but under the hood, the behavior from chip to chip is astonishingly uniform. Like, I could get another 3900x, drop it in this board, and I'm betting it'll run the same voltages, same clocks, same wattage. There really just is no wiggle to it. What it pulls out for you is pretty much all there is.

The general consensus is that it doesn't hurt anything. For overclocks, I think the common limit is 1.35-1.4v, or at least that's what I go by when I'm not okay with cooking a chip. But that's constant voltage. Ryzen's boost doesn't sustain it constantly. I'm betting it looks like it's flat-lining at 1.4-1.5 volts all of the time because the spikes and dips are too brief and slip between polling. But it's certainly not the same as running constant voltage. Guarantee you if you take that 3950x and pump 1.5v volts to it while putting it under load, the temperatures will be astronomically higher than it boosting on its own and hitting that same voltage.

The temperatures are kind of a similar thing. It's got those little spikes even under light loads... seems that when you give a Ryzen some meager tasks to much on, it likes to max the voltage out... probably because the current is so low. But again, if your cooling is up for it, not a problem. Worst case is that it stops boosting or even throttles down. But on the flipside, you might not even see much better temperatures with better cooling. If it had headroom outside of temperature, it might just eat up a good chunk of any temperature headroom you give it.. It's pretty much gonna keep going until it hits one of a few limits. You have the clock speed limit (they won't automatically run over advertised boost clocks.) And then you have wattage/current and temperature. Voltage is more of a soft limit. It has a max of 1.5, but that 1.5 doesn't limit clock speed... like it doesn't stop going higher because it wants the full 1.5. It does, however, seem to have a minimum, per a given clock speed and load. Again, to do certain clocks under certain loads, it requires that voltage. Mobo manufacturers could do better on their end, but from what I've seen the only ones could really change it much are AMD themselves. But it seems they've decided this is the optimal behavior, because they all do it

Nah, I think in your case a BIOS update might help. But the behavior you're describing is pretty typical. It's always gonna want those high voltages and little spikes. Over time I'm sure AMD will refine it and smooth it out so that maybe it doesn't spike so much and permits lower voltages... but it's not exactly a malfunction. Just really stiff rules on their end. They decide if it's a problem or not on their end. Is what it is... I feel a little conflicted, myself... but then I kinda think who am I to think I know better than AMD how their chips should run. I'm sure there are some issues with windows power management and just... all sorts of hiccups getting everything in this admittedly pretty complex boosting system to communicate and respond as intended. They've always liked to overvolt their stuff out of the box, but this is probably the first time you couldn't undervolt an AMD product that runs hot and high. There's gotta be a reason for that.

I'd say they're actually pretty good at self-regulating. They're nothing if not consistent, and while behavior takes some getting used to, I don't see one killing itself from it. It seems like their main concern is current and I assume there's a good reason for that. I've seen the voltage dip under heavy threaded workloads, when the current was the highest I've seen it. I think below a certain voltage, it's probably the combination of high current AND high voltage that kills many a CPU. You look at your voltage and see 1.5 and think gee, that's high! But when you look at your total wattage it could be like... 48. Which also means the current is very low. Hence why on the flipside, as current starts getting legitimately high, voltage will start to dip a little... more and more if the load keeps increasing. So it seems to me there is a voltage where even a small amount of current would be enough to fry a Ryzen, but 1.5v isn't it. I wish we all knew for sure, but it seems that IT at least knows how much of each it can take and will take steps to never hit a point of damage/degradation. I trust AMD has put in the footwork to figure out what voltage and current levels show signs of degradation and with their long-gathered knowledge of how CPUs behave found a safe line to draw for their boost system.

They really are a different beast, though. We're talking about a little mobile/SoC arch turned desktop powerhouse, here. All of the parameters and behavior are going to be different from what we're all used to with Intel, who's been rehashing the same chip for years. I think they're probably a lot more finicky to get the performance out of than those older, more refined architectures. Just a lot more going on with em, across the board.

And just so we're clear, that's me saying I don't really know what's going on with them but yes it is normal and no there's not much to be done

Gee whizz that was an extensive response.
Thanks!
I couldnt agree more actually.
AMD is doing something new so obviously we would be surprised with the actual outcome.
I mean its time to change habits right?
Maybe for me coming from r5 2600 to r9 3950x was just too much of a shock there I guess

 

[mstenholm]

Gee whizz that was an extensive response.
Thanks!
I couldnt agree more actually.
AMD is doing something new so obviously we would be surprised with the actual outcome.
I mean its time to change habits right?
Maybe for me coming from r5 2600 to r9 3950x was just too much of a shock there I guess

@Lorec Congratulation, you beat me to it with mouth I think. If you don't want to see the high voltage then load it big time with WCG . I can't wait to see some numbers (all core speed and temperature).

 

[Lorec]

Congratulation, you beat me to it with mouth I think. If you don't want to see the high voltage then load it big time with WCG . I can't wait to see some numbers (all core speed and temperature).

Sure, once I update everything I will get to testing WCG wise. Looking forward to seeing how it fares.
Not like I could do better than Your two 2700x pushing some serious PPDs
...wait You got 3900x recently aiming for 3950x as well? @mstenholm

 

[mstenholm]

Sure, once I update everything I will get to testing WCG wise. Looking forward to seeing how it fares.
Not like I could do better than Your two 2700x pushing some serious PPDs
...wait You got 3900x recently aiming for 3950x as well? @mstenholm

Sure but running my 3900x at 4.3 MHz all cores is not bad either. Since you have custom WC and winter is nearing then try the same if you intent to run it flat out.

 

[springs113]

Well to ad my .02. I've got several ryzen systems floating around, 1950x with zenith motherboard, 2700x with a crosshair vii, 3700x with strix itx x570 motherboard, 3950x and an aorus xtreme and finally a zenith 2 with a 3960x on order. So question away, I'll respond to the best of my abilities. Only trouble I really had was originally when the 1950x was launched where if I recall correctly installing ai-suite would decrease my fps in games considerably.

 

[TheLostSwede]

Sure but running my 3900x at 4.3 MHz all cores is not bad either. Since you have custom WC and winter is nearing then try the same if you intent to run it flat out.

That's an impressive underclock

 

[mstenholm]

That's an impressive underclock

I like to keep it under 75 C....one day when I fell bold I try 4.4 MHz