vCores please [1700x/TaiChi] |
Post Reply | Page 12> |
Author | |
VUMeter
Newbie Joined: 14 Sep 2017 Location: UK Status: Offline Points: 148 |
Post Options
Thanks(0)
Posted: 25 Sep 2017 at 10:37am |
Still in the 'side off / testing' phase.
I installed the latest Ryzen drivers 17.30, with the "Ryzen Balanced Power Plan" for Windows 10. Whilst I haven't started getting all of the content software installed, the system doesn't feel sluggish - why would it 32GB RAM and 8x cores at 3.4GHz? Anyway, I put the 'tweaked' plan on and HWinfo, Ryzen Master and anything that I could see all indicated maxed out cores. Sure, OK, a screw up in the reporting perhaps. It's not really going that quick, the temps aren't climbing, the fans are going nuts. But importantly the voltages looked like they bouncing u to 1.45v on vCore. Under the regular stock balanced plan, vCore is reported in the 0.87v region - this is idle, browsers open, monitoring tools, nothing taxing. Playing with the Power Saver plan, things look even nicer in terms of vCore bouncing to 0.84 but dropping to ~0.7. Just what is going on here? If anyone has a guide to keeping things stock and dropping voltages, I'd like to hunt for longevity and low temps as much as possible here. I guess it's mostly follow the OC guides, omitting the chip-thrashing.
|
|
parsec
Moderator Group Joined: 04 May 2015 Location: USA Status: Offline Points: 4996 |
Post Options
Thanks(0)
|
You're using the default UEFI/BIOS VCore settings, right? They are always set very high by design, on purpose. Why? Simple, the VCore must be high enough to guarantee that any and all processors used with the board will complete POST, allow the user to get into the UEFI/BIOS interface, and run Windows without failing. Even if that VCore is much more than will really be needed by most if not all processors. Again, why? Imagine a UEFI/BIOS whose default VCore was set too low for a processor to POST, and get into the UEFI interface. Meaning no way to adjust the VCore to a higher value. How could you work around that if the default VCore was to low to allow a processor to POST and run the UEFI? You can't, and the board would be worthless, unusable. To avoid that situation completely, the default VCore is set to a value high enough that it will never happen. It's a compromise biased towards guaranteed operation, rather than a refined operation. Yes, the default VCore when one or more cores Turbo up to their maximum speed is ~1.45V, I see the same thing my different ASRock X370 board and a 1700X. At idle and with different Power Plan Processor State settings, the VCore is much lower. My VCore with C6 enabled will go down to 0.384V. My OC to 3.9GHz is stable with a VCore that slightly exceeds 1.350V occasionally. I have the VCore set to 1.350V for that speed. Reduce your VCore by setting OC Mode to Manual. Don't change the base processor speed, and you won't be over clocking. Set your Vcore in a now revealed VCore option to ~1.3V for a start. You can also change the VCore mode now, using Offset Mode for example. Also the LLC option will be available. As long as the don't change the base processor/core speed, you'll stay with the stock clock. Your board's UEFI layout will be different than mine, so other X370 Taichi users can provide more exact details, but you should be able to figure it out if you have any experience with configuring the VCore in a UEFI/BIOS. The different Power Plans have different Minimum Processor State settings, that only allow the processor to reduce its clock speed by different amounts, or not at all. That is in the Advanced settings of a Power Plan. |
|
datonyb
Senior Member Joined: 11 Apr 2017 Location: London U.K. Status: Offline Points: 3139 |
Post Options
Thanks(0)
|
why when you can have both
at 'locked' 3.9 my vcore still drops to less than half a volt at idle my temps are 23 degrees amd and ryzen with some asrock jam filling HAVING YOUR CAKE AND EATING IT ! |
|
[url=https://valid.x86.fr/jpg250][/url]
3800X, powercolor reddevil vega64, gskill tridentz3866, taichix370, evga750watt gold |
|
VUMeter
Newbie Joined: 14 Sep 2017 Location: UK Status: Offline Points: 148 |
Post Options
Thanks(0)
|
Thanks for the info. Sorry for perhaps being dense here. Here is a screen grab of HWinfo readout. Which vCore are we supposed to be looking at? OpenHardwareMonitor, which I know cannot access the CPU Tdie, can only provide the vCore data at the bottom, off of the motherboard(?) UEFI is all set to AUTO everything. These results are just idling with Win10 x64 Creators power plan set to" PowerSaver" - "AMD Ryzen Balanced" gives me the willies seeing all the stats flat out. I know what they say, but the temps do actually increase on that setting. The reason for the 1.475v is because I had run Prime95 on 2 threads. That caused the 3.9GHz on those two threads and thus the added voltage. When all cores are maxed (16 threads) in Prime95 the vCores all sit at 1.2v or just a hair under. Additionally, I set the UEFI vCore to 1.2v LL3 and SoC to 1.0v LL1 and everything was fine until I did a 2 thread test on Prime95, then the screen went black, but I couldn't get back to UEFI without holding the case power button. I've a feeling I have found that my CPU won't run on that low voltage when boosting 2 threads. Should that be surprising, probably not. Annoyingly when I did set the vCore manually in EUFI it did not drop back down when the CPU wasn't being used, but rather stayed at that same voltage all of the time. Again, I apologise that I am probably making you folks roll eyes here. I'm having a tough time finding a concise and easy to follow guide on overclocking that I can apply to simply under volting. It seems I can find different settings but there isn't so much explanation behind them or what to expect - eg. the vCore becoming static. Admittedly, I did this on the old C2D system and the vCore as static, except for droop on load, hence LLC or whatever ASUS called it then. Thanks as always. |
|
parsec
Moderator Group Joined: 04 May 2015 Location: USA Status: Offline Points: 4996 |
Post Options
Thanks(0)
|
So what VCore mode did you use when the VCore did not drop? Fixed or Offset? Sounds like Fixed, or using the Ryzen power plan. Both the VCore and power plan determine the VCore in Windows.
A VCore of 1.2V is too low for stress testing, as you know. If you're using you C2D CPU as a reference, don't, your Ryzen CPU is different. I suggested 1.35V for the VCore, but you didn't apparently didn't like that value. I suggested it because I have a 1700X. As you have seen, the power plan used affects the VCore in Windows. The Ryzen so called Balanced plan sets the Minimum processor state to 90%, and will keep the VCore high at idle. The reason the VCore is 1.475V is because the "UEFI is set to AUTO everything", as you said, while stress testing. Use the "Vcore" reading in HWiNFO64 at the very bottom of your picture. The max is 1.424V, and the minimum is 0.432V. That is clearly with VCore on Auto, right? It seems you tried a manual VCore of 1.2V, which failed running Prime95. Then you want back to Auto, resulting in over 1.4V. Set the VCore to 1.35V, using Offset mode. Set the Offset voltage below Offset Mode to one step above Auto, by pressing the + key. Then run the stress test and see what you get. You may be able to reduce the VCore from 1.35V without an OC. |
|
VUMeter
Newbie Joined: 14 Sep 2017 Location: UK Status: Offline Points: 148 |
Post Options
Thanks(0)
|
Muchas Gracias! Yup, I used fixed mode in UEFI. I didn't quite understand how offset would work. I'll go read a bit more about that and then poke around in that again. I picked 1.2v based on what HWinfo was showing with all cores working. I didn't really think about the high-boosted low-thread count work. As such it totally failed at that point. I wasn't using the C2D as a reference, no for a minute, entirely different silicon! AMD Ryzen Balanced power plan = 90% min CPU state, and a high vCore. Well that totally explains why it was a bit warmer than regular Balanced or Power Saver options! vCore from the motherboard sensor, that's the one at the bottom of the image, and the one that shows in OpenHardwareMonitor, SpeedFan, CPU-Z and most other software. HWInfo just gives a more comprehensive set of data as it taps into the CPU too. But we look at the motherboard sensor for this, cool, that clears things up some. Yes, vCore was on Auto in UEFI, it does a lot of jumping around. Normally sits around 0.848v when doing something trivial like web browsing, but is constantly moving below this. So far so good, your detective skills are very sharp. You knew what I had done here from the data I have provided. OK, so UEFI vCore to 1.35v using offset mode, not fixed (I think this might be a value I have found AMD recommend for long term use). I'll follow your suggestions and report back. I'll leave SoC on auto for now, as it's frankly not gone above 1.0v
|
|
VUMeter
Newbie Joined: 14 Sep 2017 Location: UK Status: Offline Points: 148 |
Post Options
Thanks(0)
|
Ah, now this looked more like I was expecting.
vCore (Motherboard) 0.384v - 1.456v Idle vCore sits around 0.544 more often now. The NH-U14s did it's job in keeping the Tdie below 60°C (~75-80% fan speed). vCore was around the 1.2v again, slightly over, with all cores. The high vCore is when taxing the two threads only. The first + option after Auto on the Offset mode is 0.00625 This seems a strange number being that 1.35+0.00625 is 1.35625 not 1.456, so I wonder how I attempt math on that. So now, I just reduce the main 1.35v to the next lowest, and so on, until things become unstable, then bump them back up a little. Thanks, this does make some sense to me.
|
|
VUMeter
Newbie Joined: 14 Sep 2017 Location: UK Status: Offline Points: 148 |
Post Options
Thanks(0)
|
After some fiddling around and numerous reboots (using the ASRock to UEFI tool), I've gotten somewhere.
I had to reset the power plans to default which has removed the Ryzen one. All of the cores clocks were pegged in HWinfo and I just couldn't seem to find a way to make them not be. OK, I appreciate this maybe a cosmetic issue, but it was also not really reflecting vCore movement. Open cmd.exe as admin and type: powercfg.exe -restoredefaultschemes That'll reset all of the power plans to defaults, including the drive sleeping and such. "AMD Ryzen Balanced" will be gone. I have left everything at auto again, NOT setting CPU to 3400 and Core to 1.35000. I have just change the voltage mode to offset and hit - key 4 times. It's now -0.02500. The vCore doesn't go above 1.408v with LLC3, and sits mostly at 1.392v. It's a matter of tweaking and this will come down more. When hitting at 16 threads with Prime95 it's still at vCore 1.2v Loweset measured vCore is 0.352v So, thanks Parsec. I'll poke about some more, I think I know what I am doing now.
|
|
datonyb
Senior Member Joined: 11 Apr 2017 Location: London U.K. Status: Offline Points: 3139 |
Post Options
Thanks(0)
|
or just use the same old sh1t i have used since bios 1.6
and kindly enough somehow asrock have assisted in bios 3.1 and now without even asking for it i get vcore drop to sub 0.500volts when idle and im fixed at 1.3 volts and 3.9ghz my bios setup take me mere minutes to do and never creeps above 1.3 volts the taichi board has been a real smooth ride for me since start of april when i built it |
|
[url=https://valid.x86.fr/jpg250][/url]
3800X, powercolor reddevil vega64, gskill tridentz3866, taichix370, evga750watt gold |
|
VUMeter
Newbie Joined: 14 Sep 2017 Location: UK Status: Offline Points: 148 |
Post Options
Thanks(0)
|
@datonyb,
So you are saying you have vCore in EUFI set to "Fixed", but it's dropping the voltage lower when not doing anything? That's definitely not my experience with using the fixed setting - parsec seems to agree that fixed means just that, static, unchanging no matter what the load. Obviously under heavy stress it could 'droop' which is where more aggressive LLC (lower numbers) comes in. I notice your post here: http://forum.asrock.com/forum_posts.asp?TID=6175&KW=&PID=36821&title=x370-taichi-oc-tips-for-begginer#36821
But you are using P-state options, which is why I think the 1.3v is coming in. The rest is all based off of that. Without fiddling with that stuff yet, I think by default vCore is going to be higher when 2 threads are stressed. I am playing with offset negative in hopes of reducing the rather high short-term vCore when single threaded tasks occur. I've also got to work on seeing if that buried Fan control hysteresis will actually work. So far it did nothing. It's the one in the Advanced tab > AMD CBS > (?)
|
|
Post Reply | Page 12> |
Tweet
|
Forum Jump | Forum Permissions You cannot post new topics in this forum You cannot reply to topics in this forum You cannot delete your posts in this forum You cannot edit your posts in this forum You cannot create polls in this forum You cannot vote in polls in this forum |