E6320 / P35 overclock hitting the wall too early

[soulglo]

It's been a long night of reboots and I'm out of ideas. Here's the issue:

C2D E6320 (7X Multiplier)
Gigabyte P35-DS3R Rev2, latest BIOS
2GB OCZ Reaper PC2-6400 (800MHz) 4-4-4-12
Geforce 8600GTS
Windows XP SP2
Big Antec server case, big Thermaltake heatpipe coolers on CPU and NB w/Arctic Silver, 4 case fans, etc.

By all reports, even if I got the runt of the Intel batch I should be able to push this CPU to around 3.5GHz, which would be a 500MHz FSB. The OCZ DDR2 is widely reported to be happy at 475MHz or more. However, the highest stable OC I can get is @435MHz which gives a 3.05GHz CPU clock. Sometimes it will boot @440MHz, but it doesn't last long. Past 446MHz it won't even post. These numbers are at stock voltages : 1.35 Vcpu, 1.9 Vddr.


Here's the weird part: I get exactly the same results when I push all the voltages to semi-scary range:
Vcpu: 1.55 (BIOS), 1.50 (reported)
VDDR: +0.4 (BIOS), 2.21 (reported)
G(MCH): +0.2
FSB: +0.2
PCI-I: +0.2


I'm still absolutely stuck at 435-440 MHz, so it certainly seems like voltages are having no effect. I've also relaxed RAM timings and that has little to no effect. Also, I've confirmed that the Vcore, DDR voltage, and RAM timings are actually changing via CPU-Z, RightMark, and SpeedFan. The voltage rails look pretty darn clean in BIOS monitoring.

I thought it was crappy RAM but I get nearly identical results with Patriot DDR2-6400.
The only bit of troubleshooting I think I haven't done is to swap out the CPU, but seeing as I just blew rent money on new RAM thinking that was the problem...yeah.

CPU temps are reported around 24C @ 1.35V and 29C @ 1.55V. The sensor that I think is on the Northbridge runs at 44C.

Basically the overclock just won't go past what seems like a brick wall, it's happening way earlier than it should, and it seems unaffected by voltage.

Tell me I'm missing something stupid?


Bios Setup
-----------------
Robust Graphic Booster Auto
CPU Clock Ratio 7X
CPU Host Clock Control Enabled
CPU Host Frequencey (mhz) 435
PCI Express Frequency (mhz) 100
C.I.A.2 Disabled
System Memory Multiplier (SPD) [2.00]
Memory Frequency (Mhz) 800 870
High Speed Dram DLL Settings [option 1]
RAM 4-4-4-12, otherwise Auto

 

[Ika]

Your aim is too high. A lot more processors than you think won't hit 3.5ghz. I would tell you that you've just hit your CPU's FSB wall, but maybe others can help more.

The only thing I can offer is to change your voltages a bit - put Vcpu around 1.45, DDR around 2.1, MCH up one or two ticks and leave the FSB and PCIe voltages alone. Also, try lowering your memory's timings manually (CTRL-F1 in BIOS).

 

[tigersty1e]

I think 3.5 is hard for any processor... let alone a 6320.

For reference, I'm at 450 FSB and I'm not even at the limit. I just haven't had time to push it anymore. I'll do it this weekend and report back in this thread. This is at stock volts (except for ram. Mine is rated to run at 2.1v), so your system should be able to hit 450 FSB easy.

Have you tried locking your multi to 6 and upping the FSB?

Have you tried going to 450 FSB anyway in case there's a FSB hole?

 

[soulglo]

I've tried pretty much every iteration of voltages for the different rails. Doesn't seem to have an effect, though I think it crashes "harder" when the Vcore is pushed way up.

If there's a hole, it's a big one...I've tried most spots between 445-455 and it hangs before the mobo BIOS screen. Maybe I'll try higher just to be sure.

I think I'm just distressed because it's not cooperating at all...no tweaking makes a difference.

 

[SerpentRoyal]

Use 1:1 memory divider and increase NB/SB by a few notches.

 

[tigersty1e]

So what happened when you put the mutli to 6?

No post?

If so, then your CPU is fine and I would put that at Normal or Auto.



Try setting the SPD multi to 2.5 and set your FSB to 360.

That should put your ram to 900. See if you'll post and run stress tests.

If that pasts, then your ram is also fine... and we'll move from there.

 

[Xvyz]

I would set the ram voltage to at least 2.2v and loosen timings to 5-6-6-18. Then test with one stick of ram.

Your E6320 sounds like the E4300 I had. It was happy at 3.1GHz, but would top out at 3.4GHz no matter the voltage. It also would run really hot. On the other hand, my E6300 has unlimited o/c potential and is cool as a cucumber. It happily spun at 533FSB x 7 (3.73GHz) with my unmodded P5B vanillia on air. It's sweetpoint was 450FSB x 7 with that board @ 1.31v. It now runs a cool 490FSB x 7 @ 1.37v installed on my new P5K-D p35 board.

So in closing I would say that too aggressive of ram timings, even though they may be recommended by the mfg., is a common handicap in o/c'ing, especially the tRAS cycle time. Even my Crucial Ballistix need 15, so 12 tRAS is really pushing it. I use the default 5-5-5-18 @ 980MHz with 2.0v.

 

[soulglo]

@Tigerstyle: your suggestions slapped me on the forehead and said "duh." I should have thought of that, cheers.

Everything ran fine at 380FSB X (2.5) = 950 ram clock at 4-4-4-12, so it's not the RAM itself. (was quite pleased the memory ran this high, too)

Using the 450FSB (2.0 ddr multi) and forcing 6X cpu multi - it crapped itself immediately.

I'm not sure that makes the CPU core clock the problem, though...

 

[soulglo]

@ Xvyz: It is a Conroe core, so it bothers me that much more that I'm having these kinds of problems.

I tried booting at 5-6-6-18 @ 900MHz and it failed, so given my previous post I've now eliminated memory clock and memory timings as the source of the problem.

 

[Triton67]

Bioses F4 thru F6 suggested value "0" for "Refresh to ACT Delay", I've installed F7 and it suggests "22", but I've always used "42"..gave me stability with 4 sticks.
Also, this mobo don't need any more volts for MCH or FSB, I'm running 8x433 atm, left them at stock volts.

pic to early bios mem settings:
http://www.xbitlabs.com/misc/p...5-ds3r/timings.jpg&1=1

Many DS3R board owners have had problems with NB HSF, it easily loosens from the chip....I've re-set it twice already...

 

[soulglo]

I thought about the NB heatsink as a possible problem...I never even ran it with the stock heatsink in place, but immediately upgraded it with a heatpipe/fan combo. The mounting is a little weird so maybe it's not doing the job. I'll mess with that tomorrow...too bad I don't have the mounts for the original one anymore.

Also, I've put in 42 for Refrest to ACT, no change.

 

[soulglo]

Ok, I finally went in and reseated the NB cooler with fresh Arctic Silver 5 (been a busy week.) It does run a little cooler--38C now.
It didn't change anything with my results though--it's still crashing.

The only question that's I'm still stuck on is in regards to the FSB. Do I have any control over it's ratio to the CPU host frequency? (i.e. core clock). With the 450MHz that I'm trying to achieve, that makes a rated FSB of 1800MHz, which is quite far over the specs for the CPU/Mobo. I just can't find anywhere to change this ratio/multiplier, and I can't recall if that's even possible with these processors.


Oh, and does anyone know for SURE that what the BIOS calls "system" temp is the actual ambient temp sensor somewhere on the board, or is it the NB?