Not stable at BCLK 200.

djoye

2[H]4U
Joined
Aug 31, 2004
Messages
3,116
I'm working with an i7-950, Gigabyte X58A-UD3R, and OCZ 1600MHz DDR3 RAM (OCZ3G1600LV6GK).

First off, I set the BCLK to 200 to make the RAM run 1600MHz and manually set the RAM timings then ran MEMTEST86+ and it passed so I think the RAM is fine.

Secondly, my current OC is 174×23=4.00GHz and it's stable. I changed only these settings:
Hyper-Threading: on
BCLK 174
QPI ratio: x36
uncore: x16
mem: x8
PCI: 100MHz (manual set)
CPU: 1.2500V
QPI: 1.295V
CPU PLL: 1.800 (manual set)
DRAM: 1.66V (rated 1.65, options are 1.64 and 1.66 in BIOS)

RAM is running at 1394MHz in this scenario.

To go to 1600MHz on the RAM I changed:
CPU: 1.25625V
QPI: 1.355
BCLK: 200
CPU multi: x20

I use Intelburntest and it passes but when playing Team Fortress 2 I get crashes. I played a game of Unreal Tournament 3 with no crashes. I haven't tried anything else.

When running 200 BCLK with same voltage as 174 BCLK I'd get these nvidia driver crashes; bump up the voltage and I'd get blue screens citing hardware issues. Go back to my 174 BCKL settings and I have no problems gaming.

Am I maybe leaving out another setting I need to tweak? Raising the BCLK appears to generate more demand but I don't know where to add power.

Random pictures!
bios1.jpg

bios2.jpg

cpuz1600a.png
cpuz1600b.png
 
I've noticed that hl2 engine games are picky about having a very stable proc.... that being said.

How are your temps? You could easily give that chip another bump or 2 of vcore.
 
i have the same problem with the source engine, i play counter strike source and i get random lock ups. im using an unlocked amd PII x2 545 @ x4 3.5ghz, it's stable in intel burn test and prime95, but random lockups still persist in source.
 
Temps usually stay under 70F with IntelBurnTest. I suppose I could add more CPU voltage, I just hate to. :)

Stupid Source engine games! :mad:
 
try a different multiplier. its a known issue that even multi's some times aren't stable.

try 21x191 or go for 4.2ghz with 21x200
 
When you decrease your multi, your CPU will require a bit more of a voltage. Im running 205 by 20 to get my I7 950 to 4.1Ghz. It required 1.375 on the CPU voltage. I also have HT on turbo off. I also think your Qpi/VTT is on the high side of things, mine is at 1.25. My board is an Asus Rampage III, passes 25 lynx and OCCT 1hour test. Enought for my needs.
 
Doing a google search a couple days ago I saw someone suggest that more speed be added to the PCI bus when running a high BCLK, I think this one person was running a 220BCLK and another suggested that they should maybe run 103MHz on the PCI bus. Is that typical to need to do that?

Also, I came down on my QPI/Vtt voltage; can't remember what I set it to but it's much lower and the system is running fine. Is there a specific test to run to determine how much QPI voltage I really need? RAM only tests?
 
Doing a google search a couple days ago I saw someone suggest that more speed be added to the PCI bus when running a high BCLK, I think this one person was running a 220BCLK and another suggested that they should maybe run 103MHz on the PCI bus. Is that typical to need to do that?

At a higher BCLK, yes. At 200, usually no.

Also, I came down on my QPI/Vtt voltage; can't remember what I set it to but it's much lower and the system is running fine. Is there a specific test to run to determine how much QPI voltage I really need? RAM only tests?

IBT/LinX using 5GB+, HCI MemTest, or Prime95 Blend using as much memory as possible.
 
I changed some settings:
QPI/Vtt: 1.255
vdroop control: Level 1
PCI-E: 101MHz
CPU @ 1.2625V

I haven't had any crashes playing Team Fortress 2. Maybe my QPI voltage was unnecessarily high.

edit: I was still crashing at 101MHz on the PCI-E so I went down to 99MHz with the same problem. Went up to 108MHz and I've been fine but I did also change the vdroop control to Level 2.
 
Last edited:
When you decrease your multi, your CPU will require a bit more of a voltage. Im running 205 by 20 to get my I7 950 to 4.1Ghz. It required 1.375 on the CPU voltage. I also have HT on turbo off. I also think your Qpi/VTT is on the high side of things, mine is at 1.25. My board is an Asus Rampage III, passes 25 lynx and OCCT 1hour test. Enought for my needs.
Seems you are correct. I was thinking too far into this. Raising the PCI-E speed didn't help but I just needed to bump up my CPU voltage one notch, I went from 1.25000 to 1.25625 and it got stable. I found that with Hyper Threading enabled I needed to manually set IntelBurnTest to 8 threads to get it to fully load the system so I went in and bumped up CPU vcore to something like 1.35 and came down to just a notch over where I was stable at 174 BCLK (1.25000V).

Games are stable now with the exception of Crysis. I played the "free" version of Crysis just fine in the past on my last OC'd E8400 system but I now have a legitimate copy I bought from Steam (that I never really played through, just purchased since I owed them, heh) and it won't run for more than 10 minutes without a crash. I searched and it seems that lots of people have problems with Crysis when no other games would crash. I've mostly played Mirror's Edge, TF2, and some UT3 and the system has been fine but I keep getting nvidia DLL crashes in Crysis similar to what I was getting in TF2 when I didn't have enough CPU voltage.

I also found that at 1.66V on DRAM that it was drooping to 1.63, not sure if that was hurting it so I set it to 1.68 (next up) and it drooped to 1.66 which I suppose is OK since it's rated for 1.65V.
 
Back
Top