GA-Z68X-UD3P-B3 F6 bios issues - write bottleneck with Vertex 3 ssd

[christillis]

Check Ebay for the very short SATA cables. Obviously you have to be careful what you are buying but I purchased some very nice ones a while ago and picked up a few once I had verified the quality.

Thanks for the kind words by the way.

I had forgotten the link to the SATA cables, my bad. I have updated the post, but if it's too late, here it is: Overclockers UK.

Thanks DM, I was considering the eBay way...but we all know what that place is like 

You're welcomo 

[Tinker]

@Dark Mantis, just a question, you mentioned a while back best to update firmware via IDE
mode for Vertex 3  but the link is for AHCI. Am assuming there are other ways of upgrading as can
always plug my drive into USB. No idea as yet how to find firmware used on SSD unless the update
software tells you.


Regards tinker
 

[Dark Mantis]

As with any firmware update it is best and safest to make sure that the drive being flashed is the only one connected as this saves any confusion or errors.

[GTZaskarLE]

I'm back.

With a nice response of GGTS (Gigabyte Global Technical Service)!
They provided me with the beta bios F7B which is not yet available on the Gigabyte website.  
They asked me: "Please run a test with the attached beta bios and let us know your test results."

So here we go with F7B.

TESTS concerning write bottleneck with sata 3 SSD and ATTO Disk benchmark:

M.I.T. -> Advanced frequency settings -> Advanced CPU core features -> C3/C6 State Support [AUTO]
F7B : Not OK, write bottleneck

M.I.T. -> Advanced frequency settings -> Advanced CPU core features -> C3/C6 State Support [Enabled]
F7B : Not OK, write bottleneck

M.I.T. -> Advanced frequency settings -> Advanced CPU core features -> C3/C6 State Support [Disabled]
F7B : OK, no write bottleneck

Conclusions:
-  C3/C6 State Support should still be set to [Disabled], not to [Auto].
-  It's not clear why [Auto] works for the F5 bios and don't for F6, F7A and F7B.


TESTS concerning CPU frequency 3.8 GHz with X.M.P. memory profile1 1600 MHz:

M.I.T. -> Advanced memory settings -> X.M.P. [Profile1] 1600 Mhz
F7B : M.I.T. -> Advanced frequency settings -> Advanced CPU core features -> CPU Clock Ratio: 34x
That's good.

There is still left one inexplicable item.
When I enter the M.I.T. menu (not entering any of the 5 menu's) it shows:
CPU Frequency  3812.32 MHz (??)
However, as showed above, within the Advanced Frequency Settings menu the CPU Clock Ratio is 34 which is expected.

Conclusion:
-  Switching to XMP memory profile1 (1600MHz) does not clock the processor to a minimum of 3.8 GHz anymore.
   It stays at 3.4 GHz which is what one can expect.
-  It is unexpected to see the CPU Frequency status at 3812.32 MHz at the M.I.T. start screen

So the overall conclusion would be there has been made some progress but we are not yet there completely.
It's looks like Gigabyte is taking this serious which I appreciate a lot.
Go for F7C?

(I informed GGTS about this test)

[christillis]

That's very useful GTZ, thanks for sharing your response from GGTS 

[Tinker]

Once again very impressive GTZaskarLE  . Have replied to your last PM re X58A.
Doubt that board will achive those scores  with Vertex 3 max iops unless a lot of
money spent.


Regards tinker.

[Dark Mantis]

Well it does appear the GGTS are doing their best to find a solution and are on the right trail. I think they should be applauded for this as even though it is what they are there for we are quick enough to castigate them for not trying at other times.

[GTZaskarLE]

Hi DM, Christillis & Tinker,

Thanks and yes, it's going into the right direction.

In mean time I got two other reactions from GGTS.
They have sent this thread to the head quarters.

Concerning the wrong CPU frequency status they told me it would be fixed in the next bios version.
There is a so called "running change" on this issue.

The C3/C6 State Support will probably be set to Disabled by default in the next bios version.
We will have to wait for a couple of days for the new bios release.

So I am interested what will happen.

In mean time I will be offline for some time because tomorrow my summer holiday starts.
I will be back in the second half of october.

See you guys later.

[leno]

Not sure if this was already said, but the reason F5 with AUTO doesn't have the issue is because F5 disables C3/C6 if set to auto. I flashed with F5 and the issue was gone, but cpu voltage now stays at 1.2V when idling, instead of dropping to 0.9V, meaning the suspend states are disabled.

If the fix in the new bios is going to be just disabling C3/C6 by default, it's going to be useless. They should make a proper fix.

[leno]

Correction to my previous post - F5 does disable c3/c6 by default, but cpu voltage doesn't stay at 1.2 at idle, but drops like before to about 0.9V. Not sure what I was looking at the first time .

Idle power consumption increases by just 5W with c3/c6 disabled so it's accceptable.

[Dark Mantis]

If the new BIOS has C3/C6 disabled by default that is fine as a workaround but it shouldn't be accepted as a permanent fix because some people might want to have the C3/C6 functionality. The actual problem still needs fixing.

[Flatline]

I use an GA-Z68XP-UD4 and Corsair Force 3 SSD and have this issue with write speed with C3/C6 State Support.

I just updated to the latest F5 Bios (From F4) and the issue is still there, just we now seem to have no C3/C6 State Support available at all. EG no matter what you put C3/C6 too, the VCore voltage doesn't drop on idle.

My Turbo Ratio's are 42-45 hence the higher VCore of 1.36. Some pretty pictures to prove it.

[Wigpo]

After reading this post I remembered that I got the same results with my OCZ Vertex 3 in the beginning. I have the GA-Z68X-UD7-B3 board with F8 bios
I never got around to test it and see what caused it, but the problem is gone now when I'm running 2 OCZ Vertex 3 in RAID0.

This seems to be a problem spread across many different motherboards by the look of this topic!

[leno]

Regarding the vcore not dropping when idle, I've noticed that if you enable/disable C1/C3/C6 on Z68X-UD3P in bios and just restart, for some reason the board always keeps the high vcore. Only after shutting down and powering on again, the board starts dropping clock and voltage properly. C1 enabled alone should be enough for this to work.

This is true if you are using Auto and Normal (Dynamic) vcore settings, and probably not if using Manual, in which case vcore is fixed.

And btw, in the new F7 bios, still have to disable C3/C6 to get rid of the bottleneck issue.

[GTZaskarLE]

That's true, I get the same results with testing F7:

TESTS concerning write bottleneck with sata 3 SSD and ATTO Disk benchmark:

M.I.T. -> Advanced frequency settings -> Advanced CPU core features -> C3/C6 State Support [Auto]
F7 : Not OK, write bottleneck

M.I.T. -> Advanced frequency settings -> Advanced CPU core features -> C3/C6 State Support [Enabled]
F7 : Not OK, write bottleneck

M.I.T. -> Advanced frequency settings -> Advanced CPU core features -> C3/C6 State Support [Disabled]
F7 : OK, no write bottleneck

Conclusions:
-  C3/C6 State Support should still be set to [Disabled], not to [Auto] or [Enabled].
-  It's not clear why [Auto] works for the F5 bios and don't for F6, F7A, F7B and F7.

I will inform GGTS about this, my ticket about this is still open.

I fully agree with DM who said:
"If the new BIOS has C3/C6 disabled by default that is fine as a workaround but it shouldn't be accepted as a permanent fix because some people might want to have the C3/C6 functionality. The actual problem still needs fixing."

So Gigabyte still has to figure this out.
Let's wait for what they can do.