My GF’s getting tired of me sighing, lol.
I’m still on F21 so I figured I’d try 2t commandrate. Odd timings don’t work and are upped to the next even value.
Went back to the settings before and the different monitoring software would pick and choose which DIMMs to read. HWiNFO would find only DIMM2 and 3, CPUID typically 0-2 and never the 4th (i.e. “3”). Thaiphoon Burner would take ages to load, even longer to read SPD and would basically produce zeroes, thus reading nothing from the SPD chip.
Reset the BIOS and started testing voltages while OCing.
Everything Auto:
VCore maxed out at 1.380V with short spikes to 1.416V and went as low as 0.400V
SOC Voltage was anywhere from 0.8-1.104V.
Everything Auto at 40x multiplier:
The same as above
Everything Auto at 40.25x multiplier:
VCore spikes to 1.446V, floats between 1.390-1.416V
SOC Voltage stuck at 1.248V
Everything Auto, SOC Voltage 1.10625V and 40,25x:
VCore spikes around 1.355-1.380V, depending on workload, and switches between 0,400-0,887-1.325-1,369V (pr. core)
SOC Voltage at 1.100-1.104V
Now I’m back at my stable OC and everything is running perfectly stable. All DIMMs are read correctly and all else is well. My next restart will be to lower VCore a tad, 1.3125V seems to work fine.
A part of me is considering either going for the X470 motherboards once they arrive unless the MSI X370 Pro Carbon gets even cheaper.
"Dead" is a metter of perspective 
Things like U.2 ports aren’t expensive to add really so theres no harm in adding them IMO, PCI slots though are absolutely obsolete now just because it runs in Win10 doesn’t stop it from being ancient outdated tech 
I know you aren’t having a dig but lets not stomp GB in areas where they are no more at fault than other manufacturers and their designs. I’ve used front and rear audio on the Gaming 5 and both are fine for me weather that is down to me insulating more sources from EMI as standard or due to the updated modified drivers I use I couldn’t say but its just not been a problem for me. I might start to really dissect the firmware for the Gaming 5 but that would then mean development for each firmware would take a very long time and while I have historically done things like this for free even I would have to think about adding a paypal donate button or something because of the gigantic time sink it would be if I were to start replacing and modifying modules for all these boards. GB should absolutely be doing more and after easter I will chase up my contact and have some rather frank words for them if I think they are not putting enough effort in. If nothing else I’ve raised a few flags internally at GB now for just how piss poor communication is and it just might act as incentive for what few good people are at GB to jump ship.
F21 (still) on my K7.
LLC does make a difference.
VCore set to 1.31875V, 40.25x multi, Prime95 load:
Extreme: 1.368-1.380V, a few dips to 1.356V. Settles at 1.368V.
Turbo: 1.356-1.368V, a few spikes to 1.380V, a few dips to 1.344V, settles between 1.356-1.368V
High: Unstable, drops all the way down to 1.320V, tris to live with 1.332V and locks up. A few spikes to 1.344V but to no avail.
VCore dropped to 1.3125V, same multi and load:
Extreme: 1.356-1.368V with frequent spikes to 1.380V, still settles at 1.368V
Turbo: 1.332-1.356V with occasional spikes to 1.368V. Settles at 1.344V
As a whole, I’ve shaved 3-4 degrees off the CPU die-temperature and the VRM has dropped from 72-75 degrees to 67-70.
SOC stays between 1.104-1.116V
Still, it’s weird behaviour. Basically I’m using LLC to “overvolt” when compared to the VCore I’m setting up. Keeping a stable voltage seems unlikely but I’ll keep messing with it. I’m prepared to buy a new motherboard and even CPU, should both call it quits, so I’m just punching in numbers for the hell of it, now.
Edit/addition:
At least I can admit my mistakes, lol!
I’ve been looking at what the VRM’s send to the CPU, so screw my readings. Knowing this I raised the voltage further did let me OC to a stable 41xxMhz with no other issues than pushing higher temps.
LLC @ Extreme mereley reduces Vdroop to a minimum. VCore in BIOS is set at 1.3125V, VRM output registers as 1.320V and registers as 1.306 at the CPU. The latter drops to 1.294V under load.
Whether this is the board’s fault or not, I’m unsure of.
At least LLC works on the K7, definitely doesn’t on the Gaming 5 I have so GB seem to have broken something there. If anyone is looking for a good cheap replacement for their GB garbage the Biostar GT7 really is very good, its a bit lacking in some areas but I’m also working on a modified firmware which will expose some handy options and fix easily one of the biggest flaws the board has - the crappy 70c fan limit before all of the fans kick up to 100%. In fact I’ve already fixed that problem I have increased the maximum limit to 85c for proper fan control without the board just completely taking over whenever that old limit was hit. Zen doesn’t have an overly high recommended operating temp but up to 75-80c shouldn’t hurt it even long term.
LLC defintely works. I have no base of comparison so I can’t tell if a loss of 0,05-0,06V from the VRM reading to the CPU is normal. For reasons unknown, I now get a VRM-reading of 1.356V but CPU stays around 1.306 with unchanged VCore and LLC @ Extreme. Erratic behaviour to say the least, eh?
I’ve not experienced loss of fan control apart from the secondary CPU fan which upon restarting and during some cold-boots kicks up to 100%. This has been across all BIOS, so far.
Yep that matches up almost exactly with the erratic voltage behaviour I have measured which is up to around the 0.060v mark which is laughably bad. Considering all of the other problems the board has though in its firmware that voltage issue is actually a pretty minor issue in comparison to all the other bugs. Missing the GT7 but can’t make modded firmware for the Gigaballs boards not using it. Thats not to say the GT7 is perfect, its got bugs of its own some I’ve fixed others I still need top look at. If I can weed out the worst of the bugs the GT7 has it’ll be a nice board to use. Just have to try and isolate why the board can’t report an accurate CPU frequency in Task Manager when the CPU is under load and it’ll be good to roll. Few missing options I’d like to see like some more advanced memory timings but it’s not a major issue not having those options more just an irritation.
Hey ket do you think you could change the boot logo on my taichi?
As long as the 0.060V isn’t part of a stability issue or something worse … still not good enough, though. This is supposed to be a top tier card.
I’ve been playing around with the RGB-settings. My issue with the (lack of) synchronisation with Corsair’s RGB RAM seems to be as simple as timing related. Choosing solid colours, the colours sync up. If I choose any sort of lighting effect, be it flash, pulse or colour cycle the RAM un-sync within a few seconds. The RAM won’t flash, they instead start a pulse sequence. Apart from red, every colour is out of sync, where orange sticks out the most.
Basically, GB has designed and built a top tier motherboard with a quite capable VRM design, good looks and a nice feature set with lots of LEDs and a custom element. The problem is that most of the fun stuff doesn’t work very well, let alone as advertised, and in general the firmware hardly improves over time. Add to that the way the voltages behave which makes me question whether there may be a hardware issue of sorts, if not many.
I’m starting to sound like POE_UK … 
I’ll jump on X470 if something good launches. At the moment the MSI X370 Gaming M7 is on my short list, should my AX self-destruct.
ket, of the newer Raven Ridge/Pinnacle Ridge ready BIOSes (starting with F10, I think) have you looked at F22b yet?
The consensus on the Gigabyte forum seems to be that it’s the least buggy and most stable. (For example, the HPET bug isn’t there and supposedly P-States still work.)
For some reason Gigabyte removed it from all the motherboard product pages. Maybe because it was a beta version of F22?
Regardless, pretty much everyone is saying it’s better than the final F22 version. And it can still be downloaded from here: http://download.gigabyte.cn/FileList/BIOS/?C=M;O=D
I’m just thinking that if you base your custom BIOSes on older versions, you’ll be limiting yourself to Ryzen 1000 series users only. Owners of the Raven Ridge APUs and upcoming 2000 series won’t be able to use your custom BIOSes on Gigabyte B350/X370 boards at all. Which would be a shame, as they’d be missing out on your great work. 
All that being said: there are, of course, inevitably going to be more BIOS releases from Gigabyte in the very near future. What with the B450/X470 boards launching, and patches for the flaws discovered by CTS-Labs.
So I guess it’s best to wait a few more weeks and see how things go? Maybe with the release of B450/X470, Gigabyte will overhaul their AM4 platform BIOS code, and it will filter back down to the older B350/X370 boards?
I have tried F22b, its certainly not as bad as F22 but it is a far cry from being anywhere near as good as F9f, or F7a, etc, primarily down to the removal of such settings as CLDO_VDDP, and a handful of other memory related options. On top of that F22b also completely breaks XMP support in my testing which wouldn’t be such a big issue if it was not for the fact of the removal of CLDO_VDDP (which GB removed in the Gaming 5s case from any firmware from F10). I could just put all these settings back in myself but to do so would be time consuming and to be quite honest I’m not going to do Gigabytes job for them. I will mod stable, good firmware releases from GB when and if they are actually capable of making one which is why for now I’m modding older firmwares. Don’t fret over those CTS "findings", their BS non issues essentially and the whole situation reeks of corporate espionage with surprise surprise, the nefarious company behind it all seemingly being intel. THIS VIDEO will sum that situation up nicely. I wouldn’t hold my breath either if I were you regards hoping fixes trickle down to the X370/B350 boards, with GB, they almost certainly won’t because "Fuck you we got your money!".
Intel are full of shit, ill buy an intel system when nelson gets his eye back.
Yup thats easy. Takes like 10 seconds to do.
If anyone wants a good laugh check the GB beta bios thread and the last reply from someone at GB. That ignorant asshat plucked out TWO issues despite a plethora of easily verifiable problems with the firmware that GB CAN fix if they were to… well actually try.
@ket Hey ket… Any info on the broken GOP issue yet??? I’ve kept using your F5a modded firmware with no setbacks whatsover… Any firmware above this one still doesn’t give me the Windows 10 Options, FastBoot or UEFI Post… Thx in advances bro…
I can’t really look at that problem unfortunately as I don’t have an AMD GPU. What I can say is that there is no way AMD would break all of the stuff GB are trying to BS their way through and for proof of that all you have to do is look at any other board from any other manufacturer. I’ll update a version of F10m real quick for you to play around with.
I can’t really look at that problem unfortunately as I don’t have an AMD GPU. What I can say is that there is no way AMD would break all of the stuff GB are trying to BS their way through and for proof of that all you have to do is look at any other board from any other manufacturer. I’ll update a version of F10m real quick for you to play around with.
Ok… Sounds like a plan @ket !!! Let me know when when your F10m is ready for me to play around with… Thx again!!!
You have PM 
@ket Well so far so good!!! UEFI GOP works!!! Though now CPU-Z, AIDA64 & HWinfo all don’t detect ANY RAM MODULES or any SPD info at all…
Thats most likely due to the broken SPD data detection in the firmware. Thats been broken for months. Unsurprisingly, its only GB boards that have that problem.
ket king how to change boot logo?
I’m on Gigashit AB350 G3.