• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Question RE: OCing i5 750 with GA-P55-UD3R AND 8 gigs of RAM

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

Jidonsu

New Member
Joined
Oct 9, 2009
I didn't really want to ask this questions, but alas, I read all that I could over the last few days and searched as much as possible, so here I am.

There are tons of reviews / forum threads out there regarding overclocking the i5 750 and all of them seem to be very successful. The CPU seems to do very well with Gigabyte UD-3R as well. The thing is, most people seem to be overclocking with 4GB of RAM and I'm trying to do it with 8GB of RAM.

So here's the story. I've got the rig stable at 3.5GHz (175 x 20) at 1.20v. I'm running it with 8 gigs of RAM. I ended up leaving the load line calibration ON, the QPI/VTT on AUTO, and the VDIMM on AUTO as well. The memory multiplier is at 8x with timings at 9-9-9-24.

I know that some say I shouldn't leave those things in AUTO and I'm perfectly comfortable taking it off auto and entering values manually. I only want to get as high as 3.8 and I've got the CPU stable at 3.6 and 3.8. I had to increase the QPI/VTT and also the RAM voltage. I was happy with 3.6 and it was even passing 24 hours of prime95 on blend. What drove me nuts was that it would fail memtest86+ pretty quickly. I tested the system at stock settings and memtest passed. It passed as well at 3.5GHz with LLC on and all voltages set to AUTO except vcore.

So in summary:

3.5GHz - 1.2vcore, LLC on, AUTO everything else - Seems to do find with the stress tests.
3.6Ghz - 1.25vcore, LLC off, 1.29 - 1.31 VTT/QPI, 1.64 vDIMM (since I can't select 1.65 on the Gigabyte board, haven't tried 1.66 yet) - FAILS memtest and IBT right off the bat, but Prime Blend stable for 24 hours. Small FFTs stable as well. Giving the CPU more voltage didn't help.

So what's the question? I'm wondering high the MOBO is taking the other voltages while set on AUTO. It's my impression that because I'm running more RAM, more QPI/VTT is required. I'm certain the CPU isn't the culprit for the failures, nor do I suspect the RAM itself since no issues show up at stock speeds (or maybe I didn't test it enough). Has anyone tried OCing with more than 4GB and seen that a higher voltage is needed for the VTT/QPI? Note, increasing the CPU PCH to 1.12 from 1.05 doesn't seem to much either.

Anyways, this is my first time building a computer and overclocking it. I'm finding it all very fascinating and I'm more interesting in understanding and solving the problem than pure speed. Thank you for reading.
 
if your runing the ram at rated speed while the cpu is oced you should have no problems. now keep in mind the cpu now houses the memory controller like AMD cpus have for some time. using all 4 dimm slots will stress the memory controller more then just 2. a increase in the QPI/VTT might not be a bad thing but dont go to high as it can kill the cpu. the PCH only houses like the SATA and some PCIE lanes for the onboard nics and for those single 1x pcie slots.

my suggestion for starters, with 175 bclk, set the ram speed divider to be as close to its rated speed. manually set ram voltage to its rated, same for timings. if i can get 3.8ghz on a 860 on air, you should have no problems doing the same with the 750. as the 750 runs much cooler since HT is HW disabled by intel. sw HT disabled can actually cause a increase in cpu temps. while might be different for i7's but back in the day of P4's with HT they ran hotter with is bios disabled.

GL and report back! :thup:
 
Thank you for your reply.

The thing is, I'm find at the 175 blck since I left it at AUTO settings for voltages other than vcore. I'm CPU stable at 180 blck and at 190 as well. (x20 the whole way)

It's starting at 180 blck that I'm having issues. AUTO everything else except vCore is not working on at the blck of 180. I'm getting the CPU stable with 1.25v of VCORE (It's even stable at 1.225, actually), with RAM at 1440mhz (180 x8 multiplier) (Kingston HyperX 1600mhz) and the ram voltage at 1.64. VTT/QPI is as high as 1.31 and still I'm failing memtest and IBT. I don't know how much higher to push VTT/QPI. I'm certain that's the issue. Like you mentioned, I'm straining the memory controller with 4 DIMMS.
 
well your running the ram lower the rated.. plus, you didnt say what timings are set to and what is spec for timings. to much vdimm for a low speed can cause ram to corrupt data. another thing might be the bios needs to be updated for better ram support at higher speeds. that at times can be the cause for certain brands/models of ram.

have you also checked out www.xtremesystems.com in their forums. to see who is having what issues and how they resolved them? cause im willing to bet someone might be using the same ram as you and had the same issue. also some of the guys at XS get beta bios's from GB before they hit the FTP site at GB.
 
My timings are currently 9-9-9-9-24. It's rated at 8-8-8-8-24 1.65v at 1600. Maybe I'll do a switch and run mine my bclk at 200 and x18 to achieve 3.6. I'll play around this weekend.

I've done some reading at xtremesystems. No one seems to be running the same amount of RAM. Most seem to be running 4GB and cracking 4.0GHz on the chip. There is a new bios out for my board though. I'll do some more reading there also.

Interesting bit about too much vdimm for too low a speed though. Food for thought. Thanks!
 
Up to 1.35 QPI/Vtt ought to be well within the safe range, I think Intel specs the max even higher. You could also try oc'ing just with 4GB to see how things behave differently.

Also don't be fooled by which tests pass and what it means. I was going through an oc on an LGA775 recently (the older platform is actually a bit more complex with multiple related settings that need to be set in the correct proportions) and I could pass Memtest, IBT, P95 small FFT but fail P95 blend. The obvious answer there would be the northbridge voltage/GTL since I passed both memory and CPU-only tests. But it turned out I need more CPU vCore! That's the last thing I would have thought was the problem given P95 small FFT passing. The only way I figured it out was long iterative testing by changing one thing at a time. I'm not saying you will have to do that but don't dismiss things just because the obvious answer is something else.
 
The thing is, I'm pretty happy running it at 3.5Ghz right now. What bugs me is not knowing what it is doing with the voltages because of the AUTO settings, eventhough it works. The heaviest duty this computer will see is Photoshop and Lightroom, which is mainly the reason I'm using so much RAM. I'm only overclocking because it looks like it's fun to do. I knew it could be addicting.

This thing runs circles around my laptop, which is a Thinkpad T61 with a T7200 in it. I should just leave it be, but oh...the inquisitive mind...

So quite update. This things boots at 200bclk x 18, but only with memory multiplier at 6x. It doesn't like 8x at all, at least not with reasonble VTT/QPI values. So it's 3.6Ghz with 1200mhz RAM or 3.5Ghz with 1440 RAM.

Ooh, how about bumping it to 200 bclk x19 for 3.8Ghz @1200mhz RAM.

Haha. I'm sorry you guys have to put up with this. I hope I'm not wasting anyone's time.
 
Um, you're in an overclocking forum occupied by overclocking geeks :p We love it. :D

The BIOS probably has a section that tells you what the current voltages are even if they're set to auto. Usually it's near where the voltages are set. If not or it doesn't show them for auto try installing Gigabyte's EasyTune 6 in Windows, it's a program that shows lots of info including current voltages. It may turn out that some auto settings are actually not good and making the oc harder.
 
Awesome! I found a home. Good point about reading the voltages in BIOS. It's to the left of the column where I make changes, right? I always thought those were the default settings, but I assumed AUTO means it'll make changes on the fly.
 
Oh, what are your temperatures and what CPU cooler are you using? Get CoreTemp or RealTemp to monitor temperatures inside windows under program load.
 
I'm already monitoring with core temp. I'm using the Megahalems since they just came out with 1156 adapter. My temps at 3.5Ghz are in the lowers 50s with Prime and high 50's with IBT. At 3.8Ghz@~1.2875v, which is what I'm running as we speak (typing on my laptop, of course) is 65 with IBT and 55 or so with Prime Blend. Not happy about the 65, but then again, I'm never going to see those loads on a normal basis.

Idle temps are in low 30's for both. It was low 20's at stock. I think I got down to 21 one night....

Edit: I meant Real Temp. Kaspersky gave me some weird warning about the activity of Core Temp. Strange. Close enough, eh?

My Rig | i5 750 |GA-P55-UD3P | 4x2gb KHX1600C8D3K2/4GX |Prolimatech Megahalems| Antec EA750 | EVGA 1GB 9500GT (sigh) | 500gb WD Black | 750gb WD Black | DVD burner |Antec 902
 
Last edited:
well the author of coretemp is a long time member here. i have no problems with using coretemp at all...
 
Screenshot before I go to bed. Failed since the last residual was too big, of course. Will fine tune and test more conclusively later. Time for bed.

1.2875 vcore, LLC enabled (thought I had it disabled, which would explain the voltage shifts, suggestions on on vs. off?)
1.310 QPI/VTT

VRAM Voltage on AUTO for now since the RAM multiplier is on x6 right now.

3.8GHz.jpg
 
1.2875 vcore, LLC enabled (thought I had it disabled, which would explain the voltage shifts, suggestions on on vs. off?)
1.310 QPI/VTT

LLC off would be better for a 24/7 system THEORETICALLY...but my advice is to do what you need to get the OC you want. Theory usually just gets in the way :D

BTW, you're OC with 8GB is AMAZING! I've been following a few guys at XS who are all having issues with a bclock wall at 160-180MHz when running 4 mem dimms ;) Keep us posted and keep pushing it :) (I'll have to do some testing on the UD3R with 8GB, I'll get back to you on that)

Also, don't worry so much about temps...mid 60s is REALLY cool. You don't need to worry until you get up over 90C :beer:
 
My view on LLC is this -

If you are going to leave C1E, Speedstep and EIST etc ON then leave LLC OFF. In other words, if you're going to follow the Intel specs and design scheme follow them all.

If you turn OFF the power saving functions then turn ON LLC.

All the oc guides recommend to turn the power saving features off, I don't know if that's because of ease of oc'ing, going for 'max oc' or just a rule of thumb that gets perpetuated. Personally I leave power saving features on while even while testing oc's since I oc for settings I will use all the time so why would I want excess power draw. If I didn't and wanted the power savings features I'd end up costing myself more time to double test everything.
 
So far I'm more stable with LLC on, especially at these higher BCLKs, for some reason. Now I'm at 1.3v with LLC on. With it off, I keep inching higher without luck. Didn't really want to get past the 1.35v mark (though I know it's pretty safe). When I began OCing, I left all the power saving features on and it changed the multiplier depending on the load at any given time. I didn't think that it did a whole lot to save power though, since the voltage was manually set. I could almost swear that I read somewhere that the new beta bios from Gigabyte for the P55 boards has dynamic vcore up until the max you set, so it doesn't only reduce your overall clock speed at idle, it also reduces your voltage. I might've dreamt it though. I'll have to look again.
 
My view on LLC is this -

If you are going to leave C1E, Speedstep and EIST etc ON then leave LLC OFF. In other words, if you're going to follow the Intel specs and design scheme follow them all.

If you turn OFF the power saving functions then turn ON LLC.

All the oc guides recommend to turn the power saving features off, I don't know if that's because of ease of oc'ing, going for 'max oc' or just a rule of thumb that gets perpetuated. Personally I leave power saving features on while even while testing oc's since I oc for settings I will use all the time so why would I want excess power draw. If I didn't and wanted the power savings features I'd end up costing myself more time to double test everything.

I LIKE YOUR THINKING MADMAN...THIS MAKES A TON OF SENSE AND SOUNDS LIKE GOOD STUFF...I MIGHT HAVE TO INVESTIGATE FURTHER :beer:
 
Back