![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
I've had my sig system for about a year now.
When I first got it, the CPU (6600K) could reliably reach 4.4GHz without a significant amount of Vcore (It might have gone higher if I'd pushed, but it was 'good enough' and didn't run too hot. The XMP profile (3200 16-18-18-38) on my DIMMS (Corsair Vengeance) just wouldn't work (it wouldn't even post) but I was able to set the speed and timings manually and it worked (although I ended up dropping it to 2400 to get the CL down).
The board came with an F1 BIOS, I updated it to F2 during the build. An F3 version came out not long after, but it was beta so I decided to leave well alone.
A week or so ago I saw there had been three new BIOS Versions (F20-F22) since I last checked, including improvements for XMP (in hindsight - foolishly) upgraded the UEFI firmware from F2 to F22 (Which can't be reversed according to the support page, I guess because F20 introduced 7th gen support).
It's major cosmetic change as well as whatever is under the hood. Apart from the fact it looks ugly and is a really laggy interface, I'm having some issues with performance.
My 6600k that used to hit 4.4GHz easily, now can't get it above 4.3 (It will post, but bluescreens while loading into windows). Or at least it won't without an unacceptable amount of Vcore.
XMP now works (good!) but while the setting in setup says 3200, both the side popout in UEFI setup and CPU-Z say it is running at 2800 (See Photo) Same thing happens if the XMP profile isn't used, and timings are set by hand again.
The CPU thing is annoying, but it's possible it was coincidence? Either way it's outside spec so I know it's not guaranteed, but both the RAM and the MB have 3200MHz on the spec list, the setting is there and it say's its doing it, and then it just isn't.
It just seems less stable overall, regularly kicking back to setup after making changes saying there was boot failure, but powering off and on again with the same settings it runs fine and passes stress tests.
Anyone seen a bug like the RAM speed before?
Any suggestions for fixing?
I don't have a Kabby Lake chip, am I really unable to flash an older version?
Am I just stuck crossing my fingers for an F23 BIOS?
Edit:
Resolved it but posting in case anyone else has a similar issue.
After speaking to Gigabyte and further testing, I think it's due to voltage, still can't get XMP but manually setting voltage to 1.36v lets me manually tweak it up to 3100. I know the max for the Skylake MC is 1.5v but I didn't want to keep pushing it higher just to see if 3200 would eventually work.
It never failed POST, but it did take a lot longer than normal when restarting after setting a high memory speed with less than 1.36v, my guess is it was failing POST but then self tweaking until it passed, which is fine but an error message would have helped.
The DRAM voltage is in 0.02v steps, so when manually tweaking I'd had it at 1.34v.
In the end I found that 1.36v also let me run it at 3000MHz but keep the same timings as the JDEC speed (15-15-15-36) rather than the slightly more relaxed XMP speed (16-17-17-36) which is a satisfactory compromise to me.
Managed to get the CPU back up to 4.4GHz but it required 1.4v and still wasn't really stable. Dropped to 4.3Ghz, but while it seemed stable a prime 95 torture test reported that core 3 was making mistakes. 4.2 passes the prime test and is happy with just 1.28v which seems to be quite a bit cooler (and quieter), I think when I'd run it at 4.4 before I perhaps just hadn't been thorough enough.
When I first got it, the CPU (6600K) could reliably reach 4.4GHz without a significant amount of Vcore (It might have gone higher if I'd pushed, but it was 'good enough' and didn't run too hot. The XMP profile (3200 16-18-18-38) on my DIMMS (Corsair Vengeance) just wouldn't work (it wouldn't even post) but I was able to set the speed and timings manually and it worked (although I ended up dropping it to 2400 to get the CL down).
The board came with an F1 BIOS, I updated it to F2 during the build. An F3 version came out not long after, but it was beta so I decided to leave well alone.
A week or so ago I saw there had been three new BIOS Versions (F20-F22) since I last checked, including improvements for XMP (in hindsight - foolishly) upgraded the UEFI firmware from F2 to F22 (Which can't be reversed according to the support page, I guess because F20 introduced 7th gen support).
It's major cosmetic change as well as whatever is under the hood. Apart from the fact it looks ugly and is a really laggy interface, I'm having some issues with performance.
My 6600k that used to hit 4.4GHz easily, now can't get it above 4.3 (It will post, but bluescreens while loading into windows). Or at least it won't without an unacceptable amount of Vcore.
XMP now works (good!) but while the setting in setup says 3200, both the side popout in UEFI setup and CPU-Z say it is running at 2800 (See Photo) Same thing happens if the XMP profile isn't used, and timings are set by hand again.
The CPU thing is annoying, but it's possible it was coincidence? Either way it's outside spec so I know it's not guaranteed, but both the RAM and the MB have 3200MHz on the spec list, the setting is there and it say's its doing it, and then it just isn't.
It just seems less stable overall, regularly kicking back to setup after making changes saying there was boot failure, but powering off and on again with the same settings it runs fine and passes stress tests.
Anyone seen a bug like the RAM speed before?
Any suggestions for fixing?
I don't have a Kabby Lake chip, am I really unable to flash an older version?
Am I just stuck crossing my fingers for an F23 BIOS?
Edit:
Resolved it but posting in case anyone else has a similar issue.
After speaking to Gigabyte and further testing, I think it's due to voltage, still can't get XMP but manually setting voltage to 1.36v lets me manually tweak it up to 3100. I know the max for the Skylake MC is 1.5v but I didn't want to keep pushing it higher just to see if 3200 would eventually work.
It never failed POST, but it did take a lot longer than normal when restarting after setting a high memory speed with less than 1.36v, my guess is it was failing POST but then self tweaking until it passed, which is fine but an error message would have helped.
The DRAM voltage is in 0.02v steps, so when manually tweaking I'd had it at 1.34v.
In the end I found that 1.36v also let me run it at 3000MHz but keep the same timings as the JDEC speed (15-15-15-36) rather than the slightly more relaxed XMP speed (16-17-17-36) which is a satisfactory compromise to me.
Managed to get the CPU back up to 4.4GHz but it required 1.4v and still wasn't really stable. Dropped to 4.3Ghz, but while it seemed stable a prime 95 torture test reported that core 3 was making mistakes. 4.2 passes the prime test and is happy with just 1.28v which seems to be quite a bit cooler (and quieter), I think when I'd run it at 4.4 before I perhaps just hadn't been thorough enough.
Last edited: