Gigabyte GA-EP45-UD3P RAM overvolt issue

[sruland]

CPU: E8500 step E0
MB: GA-EP45-UD3P BIOS FE
RAM: F2-9600CL5D-4GBTD

After wrestling with Gigabyte for months and following these related posts closely, I flashed their latest BIOS which now allows me to populate DIMM slots 1 and 3. It also sets VDIMM to 2.0 and tRFC to 68.

I have my E8500 overclocked to 4.2GHz by setting the FSB to 443 and leaving the multiplier at 9.5. This finally runs stable, surviving Prime95 for several hours. The problem I am now having is when I change the memory strap divider from 1:1 to 4:3 (at 400MHz strap) which puts the RAM at 1188, the auto setting on the motherboard for VDIMM goes up to 2.2. When I manual reduce it to 2.1, the system becomes highly unstable. I was able to downgrade the strap to 333 and use the 6:5 setting which puts my RAM at 1063, but would very much like to get as close to the 1200MHz capacity I can out of this wonderful G.Skill RAM.

Any suggestions?

Here are my primary settings:
Vcore = 1.500
VTT = 1.40
VPLL = 1.59
VNB = 1.40
MCH/DRAM Ref = 0.95
VSBIO = 1.59
VSBCore = 1.20
VDIMM = 2.08 (as high as I feel comfortable with)
RAM = 5-5-5-15-2T
CPU Clock dDrive = 800mV
PCIE clock drive = 800mV
CPU skew = 50ps
MCH skew = 100ps
I also set the DIMM2 clock skews for both channels to +100ps; can't say I've noticed a difference there.

[genetix]

FIRST of DON'T FRY the good motherboard for my advices look for yourself where the voltages settle over, atm, values.

..and to point. Your issue is at North Bridge quite clearly not in RAM.

The NB cannot hold those kind of speeds it would output from RAM <-> MB(vNB) <-> CPU <-> GPU and no way at 1.4v. Recalculate it.. keep maximum of 1.528v ACTUAL (This means check the North Bridge voltage with Everest(or BIOS diagnostic/live voltage screens) what does 1.40v is ACTUAL in OS and adjust by that scale). Also, the 'Skew' is unnessessary. basically it doesn't matter either I would keep default as 150ps which is around 3 clicks, but that could create instability more than letting board decide which is more accurate. However, even IF you would get NB solved(which I doubt as it will not keep stable over 1.528v and 1188Mhz on memory = ~14x00MB/s total throughput, will sure need more) & Memory / System stable by this overvolt NB will put your system back to unstable(with exception that this is a very good motherboard).

Just in test methology, You need MemTest86+ run it several times (NOT completely just run it to see, if it starts). When you get it up from USB or DVD 2-3 times in a roll you know you begin to be close to stable NB, if you wanna do this at OS level which is bad I'd use IBT(Intel Burn Test v2.3) > Run test with 64MB of RAM 150 times > if this screws up > Back to BIOS(vNB is instant problem) > if not > continue with 2048, 4096, > full memory few times. Ideally also on memtest you are on good side in a sense that you cantake 'test #5' in memtest (keys: c 1 3 5 <hit enter>) > this basically will prove, if no errors that memory voltage is enough (however, with consider that any lowest memory voltage is always +0.10-0.20v to vRAM/vDDR to be stable in windows systems 'exception of Windows 7 build 18xxx/2xxxx which can function a hinch lower than XP or Vista').

[sruland]

I increased VNB to 1.52 (MB doesn't have fine tuning down to 1/1000s) and ran P95 just to make sure it didn't adversely affect stability. That worked out fine, but then I set the RAM back to 1188 and WIN7-64 wouldn't boot; in fact the welcome screen flashed rapidly. I reset the PC and dialed VNB down to 1.50 and tried again. This time it booted but failed P95 immediately, like last time when VNB was 1.40.

I think I'll put an Antec Spotcool fan in there to blow on the NB chip. I have no monitoring s/w loaded to detect the chipset temp, but wonder if that could be why WIN acted the way it did.

I also wonder if such a high VNB would require a higher VTT. I definately don't want to go beyond 1.40 VTT so maybe I'm stuck.

Can I feel ok running the RAM at 2.2v like the MB tried to run it at? What's the max for this RAM?

I'll try some more things tomorrow. Thanks for responding!

[GSKILL TECH]

You're pretty much at the max, you don't want to go too far past 2.20V.

Thank you
GSKILL TECH

[sruland]

Well, I allowed the MB to pump 2.2v VDIMM into the RAM, allowed it to auto-set timings (5-7-7-20), increased VNB to 1.50 and still can't even get it stable at 1066MHz. 1063 is as high as it will let me go. *sigh*

Would G.Skill recommend any other C2D boards that I could get my 1200MHz out of with this terrific RAM? Maybe it's a P45 chipset issue? Although from readings posts all over the internet, I'm pretty well convinced it's Gigabyte...

At least with their latest BIOS release, I can run at all without random restarts and BSODs.

[GSKILL TECH]

Post pictures of the BIOS and I can see what may be holding you back. We have no issues operating DDR2-1200+ with a similar EP45 motherboard.

Thank you
GSKILL TECH

[sruland]

Sorry it took so long to get back to you. Here are my BIOS settings for EP45-UD3P w/ E8500

Robust Graphics Booster [Auto]
CPU Clock Ratio [ 9 X]
Fine CPU Clock Ratio [+ 0.5]
CPU Frequency 420 Ghz (443 X 9.5)

CPU Host Clock Control [Enabled]
CPU Host Frequency (Mhz) [ 443]
PCI Express Frequency (Mhz) [100]
C. I. A. 2 [Disabled]

Performance Enhance [Standard]
Extreme Memory Profile (X. M. P.) [Disabled]
(G)MCH Frequency Latch [333Mhz]
System Memory Multiplier (SPD) [2.40B]
Memory Frequency (Mhz) 1066 1063
DRAM Timing Selectable (SPD) [Manual]
CAS Latency Time 5 5
tRCD 7 5
tRP 7 5
tRAS 20 15

Load-Line Calibration [Enabled]
CPU Vcore 1.30000V [1.50000V]
CPU Termination 1.200V [1.400V]
CPU PLL 1.500V [1.590V]
CPU Reference 0.760V [0.886V]

MCH Core 1.100V [1.400V]
MCH Reference 0.760V [0.886V]
ICH I/O 1.500V [1.590V]
ICH Core 1.100V [1.200V]

DRAM Voltage 1.800V [Auto]
DRAM Termination 0.900V [Auto]
Channel A Reference 0.900V [Auto]
Channel B Reference 0.900V [Auto]

CPU Clock Drive [800mV]
PCI Express Clock Drive [800mV]
CPU Clock Skew [50ps]
MCH Clock Skew [100ps]

All memory subtimings set to [Auto]

CPUID Hardware Monitor Results
CPU VCORE 1.47V
DDR 2.00V

CPUZ Results
DRAM Frequency 531.6 Mhz
FSB : DRAM 5:6
CAS# Latency (CL) 5.0 Clocks
RAS# to CAS# Delay (tRCD) 5 clocks
RAS# Precharge (tRP) 5 clocks
Cycle Time (tRAS) 15 clocks
Row Refresh Cycle Time (tRFC) 68 clocks
Command Rate (CR) 2T

btw, I am running WIN7 HP 64 bit

Thanks for the help!

[GSKILL TECH]

Try manually settings DRAM Voltage to 1.90V. You may want to try a different System Memory Multiplier, or leave the (G)MCH Frequency Latch [AUTO].

Thank you
GSKILL TECH

[sruland]

OK, I set the VDIMM to 1.9 and changed the strap to 2.66 (2.50), which gave me 1108 @ 443FSB. Prime 95 ran stable! So dropping the strap was definately an improvement @443.

Then I tried looking for ways to improve the RAM speed. I know from all of my previous tweeking that 4.2 is about as high as I can get the CPU to go. So I changed the multipier to 9 and raised the FSB to 466. I played with VNB (as high as 1.48) and VDIMM (as high as 2.1) for quite a while but couldn't get the system stable. This was at 1165. I tried increasing the strap back to 333, but the best multiplier I can get is 2.4 which yielded 1118. Still couldn't get the system stable. Tried the 400 strap but that only gives me 2.6 which was way too high. I even raised the VPLL up to 1.69 but anything over 1165 or so caused BSOD. Tried lowering the multiplier down to 8.5 with FSB at 495 but no luck there either.

Tweeked all these different combinations for a couple of hours but 1108 remains my new high. I'm thrilled that I could get it to overclock at all with this MB.

Any other ideas? I'd love to go higher!

Thanks again.

[GSKILL TECH]

Post your 1165 settings, must be lacking voltage somewhere.

Thank you
GSKILL TECH

[sruland]

Robust Graphics Booster [Auto]
CPU Clock Ratio [ 9 X]
Fine CPU Clock Ratio [+ 0.0]
CPU Frequency 4.19 Ghz (466 X 9)

CPU Host Clock Control [Enabled]
CPU Host Frequency (Mhz) [ 466]
PCI Express Frequency (Mhz) [100]
C. I. A. 2 [Disabled]

Performance Enhance [Standard]
Extreme Memory Profile (X. M. P.) [Disabled]
(G)MCH Frequency Latch [266Mhz]
System Memory Multiplier (SPD) [2.50A]
Memory Frequency (Mhz) 1066 1165
DRAM Timing Selectable (SPD) [Manual]
CAS Latency Time 5 5
tRCD 7 5
tRP 7 5
tRAS 20 15

Load-Line Calibration [Enabled]
CPU Vcore 1.30000V [1.50000V]
CPU Termination 1.200V [1.400V]
CPU PLL 1.500V [1.590V]
CPU Reference 0.760V [0.886V]

MCH Core 1.100V [1.400V]
MCH Reference 0.760V [0.886V]
ICH I/O 1.500V [1.590V]
ICH Core 1.100V [1.200V]

DRAM Voltage 1.800V [1.900]
DRAM Termination 0.900V [Auto]
Channel A Reference 0.900V [Auto]
Channel B Reference 0.900V [Auto]

CPU Clock Drive [800mV]
PCI Express Clock Drive [800mV]
CPU Clock Skew [50ps]
MCH Clock Skew [100ps]

All memory subtimings set to [Auto]

CPUID Hardware Monitor Results
CPU VCORE 1.47V
DDR 1.92V

CPUZ Results
DRAM Frequency 582.5 Mhz
FSB : DRAM 4:5
CAS# Latency (CL) 5.0 Clocks
RAS# to CAS# Delay (tRCD) 5 clocks
RAS# Precharge (tRP) 5 clocks
Cycle Time (tRAS) 15 clocks
Row Refresh Cycle Time (tRFC) 76 clocks
Command Rate (CR) 2T

WIN7 boots up fine, everything seems to work fine, but it can't handle Prime 95 for more than a few minutes. I tried knocking the FSB down to 465, then 460 to make sure it wasn't just the 266 strap being on the verge of stressing out. That didn't help. Best luck I had was when I adjusted VPLL up to 1.61. Ran P95 for 13 minutes before failing, so I upped it to 1.63, but it failed almost right away.

I also tried changing multiplier to 8.5 (FSB to 495) and tried both 333 strap and 400 strap but with no luck. I played with VDIMM (as high as 2.08), VNB (up to 1.50), and VPLL (up to 1.71) in different combinations (starting with voltages you see above and increasing one step at a time).

[GSKILL TECH]

What are advanced timings set to in BIOS?

Thank you
GSKILL TECH

[sruland]

All advanced memory subtimings are set to auto, so I can't be sure what they actually are, but here is what the BIOS has next to each one:

tRRD 4
tWTR 4
tWR 8
tRFC 68
tRTP 4
Command Rate (CMD) 0

Channel A Subtimings

Static tRead Value 6 [Auto]
tRD Phase 0 Adjustment 0 [Auto]
tRD Phase 1 Adjustment 0 [Auto]
tRD Phase 2 Adjustment 0 [Auto]
tRD Phase 3 Adjustment 0 [Auto]

Trd2rd (Different Rank) 8 [Auto]
Twr2wr (Different Rank) 8 [Auto]
Twr2rd (Different Rank) 7 [Auto]
Trd2wr (Same/Diff Rank) 9 [Auto]

DIMM1 Clock Skew Control [Auto]
DIMM2 Clock Skew Control [Auto]
DDR Write Training [Auto]

Driving Strength Profile [Auto]
Data Driving Pull-Up Level [Auto]
Cmd Driving Pull-Up Level [Auto]
Ctrl Driving Pull-Up Level [Auto]
Clk Driving Pull-Up Level [Auto]

Data Driving Pull-Down Level [Auto]
Cmd Driving Pull-Down Level [Auto]
Ctrl Driving Pull-Down Level [Auto]
Clk Driving Pull-Down Level [Auto]

Channel B Subtimings
***Same as Channel A***

I know these values (where they are shown) aren't accurate becuase CPUZ shows tRFC at 76, for instance.

[GSKILL TECH]

Command Rate 2

Static tRead Value should be 8-10

The higher this is the less performance, but can stabilize higher frequency. Vice versa.

Thank you
GSKILL TECH

[sruland]

I set CR=2 and tSR=10, dropped the mutiplier to 9 and set the FSB to 465. Gave me 1165 RAM. Ran a quick Prime 95 but failed after 5 minutes. Dropped the FSB to 460 and it ran 10 minutes.

I also tried dropping multiplier to 8.5, changing the strap to 333 (2.40 divider) and upping FSB to 495. This put the RAM at 1188. (These are all settings I tried earlier. I wanted to see how the changes I made today affected them.) Graphics card drivers failed to load on that one, but I didn't increase any other voltages. Changed the FSB : RAM divder to 1 : 1 just to check that it wasn't the NB not liking the high FSB, but everything ran stable (P95 for about an hour) with RAM = 990.

I remember reading once about static read affecting stability like you said. Tomorrow I'll play some more.

One thing I keep meaning to mention is that there is no Vdroop at all but there is VDIMM droop from 1.92 to 1.90. Don't know if this means anything.