Sign in with
Sign up | Sign in
Your question

How dangerous is (up)setting vdroop

Last response: in Overclocking
Share
January 7, 2011 10:59:37 AM

I don't want to spoil my new PC(my first build ever). But it appears that it is the only way I can achieve any type of overclock is to go against intel's vdroop recommendation setting. Is there a safer (relatively) way? If I try to increase Bclk even by 7 Mhz I get a failed boot not getting past the BIOS stage. Any advice is appreciated. Though if you have experience of Gigabyte H55M-UD2H with an i3-550 - 3.2Ghz, that would be excellent on settings advice, I am very new to this, so please not to technical unless necessary to the point. Some of the posts I have read may as well be rocket science at my current level of understanding.
I really want to get the CPU to 3.6Ghz via 150MHz Bclk.
Thanks, Nick
January 7, 2011 1:05:38 PM

I thought simply adjusting one variable was too easy. I'll give it a try.
Thanks for the direction.
January 7, 2011 3:27:48 PM

OK. After a bit of tinkering I've got the Bclk to 142 giving 3.4Ghz moved the memory multiplier down to 8 from 10 giving 8,8,8-21 from 9,9,9-24 vcore on auto appears to be knocking out 1.2v. Not really familiar enough with my new PC to be able to notice a difference. No doubt I'll be coming back for some more help soon.
Nick
Related resources
a b à CPUs
a c 100 K Overclocking
January 7, 2011 4:17:50 PM

You seem to need a bit of help with what you're doing... First off, your RAM should have come with some recommended timings. Most likey 1333mhz, CL8 or so, maybe 1600mhz CL8 or 9... For any RAM, there's a maximum speed they will run at any given latency.

As you now know, upping the base clock is going to modify the RAM speed. Ideal target RAM speeds are going to be around 1400mhz, 1600mhz, and 1800mhz depending on what you bought. 1333 is a JEDEC spec but I find you can generally run the same timings at 1400.

Most likely RAM was causing the issue moreso than voltage. If you leave your voltages on auto it will probably be able to maintain stability up to 160 base clock or so, although I never used an H55 board. But on two P55 boards I've used, Auto was fine. I'll mention, however, that Auto generally isn't going to give the ideal voltage for lowest power and heat.

Finally Vdroop - it is with Intel's spec because it is also within spec to have Turbo, EIST, C1E, C States. What I've found is allowing Vdroop (IE: LLC off) is best when using all these features. If you are going to turn off turbo however, then you should probably turn on LLC as well (no Vdroop).

Definitely read some guides about overclocking your system to get a better idea of what's going on.
January 8, 2011 3:28:01 AM

Thanks Gene 0 and Wolfram23

I'll be spending this weekend reading up, Appears that the cross-your-fingers method isn't working. I had altered LLC and the memory multiplier at the same time so incorrectly picked LLC as the issue.

I've set it all voltages to auto though moving the multiplier up to the standard 10x gives me reboots so moved the DRAM voltage from Auto to 1.6 but still the issue reboot persists unless multiplier is at 8x (1200Mhz). Running Prime95 stress test on heat and CPU setting gives a core temp hovering at 69C So won't be going for 160 without a better CPU cooler (CM TX3 probably). The voltage shown in CPUz is at 1.29 which appears not to bad but still a 0.1v jump for a 10 mhz increase in bus makes me want to try 148.

For now though anyone have any advice on what to adjust on memory issue to get back to the normal 10x multiplier I have tried moving the DRAM voltage up to 1.62 but havn't yet researched safe levels and may be on the wrong track anyway.The ram is corsair CL9 - the heatsinked variety.

Thanks again
January 9, 2011 4:17:13 PM

Right, I have absorbed all that I could. The auto setting on core voltage was giving too many extra volts. Manually and patiently adjusting got me 167 Bclk and [edit 1.25V] to give me 4Ghz with Prime 95 full blast at 69C - no errors (That temp was my 148 clock while on auto. Everest benchmarks look good and the increase is noticable compared with 3.2 esp. playing with Photoshop filters. I'll probably use core temp's overheat protection at 78C just in case..

So finally, thanks to Wolfram23, Gene O (who has helped me out on another thread also) and JSC, for while my questions may not have been taxing to you, it makes it all the more kinder that you took a moment out to help an OC'ing innocent. You have nudged me to exceed my expectations.
a b à CPUs
a c 100 K Overclocking
January 10, 2011 2:08:29 PM

Congrats on the OC. Just curious how long you tested in Prime? I find Prime to be too slow personally. Most OCers recommend overnight (8hrs) of Prime95 to determine stability. Personally tho I like Intel Burn Test. If you can do 20 passes at high stress, you're stable. That takes maybe 30 mins.

Did you figure out the RAM thing? Sounds like your timings are off to me.
January 10, 2011 3:42:18 PM

Thanks,I am still feeling chuffed about the 25% gain.

I tested for a half hour default stress test setting on P95. The most intense CPU work I do is Photoshop filters on 6mp raw files which take much less than a minute and maybe film conversions (2-3 a week) that take a lot longer but only appear to use a max 66% load at 4GHz .

I use EIST and all the other power saving funtions available
All of my programming compilations take a matter of seconds - there hardly ever 400 lines of simple maths. And is the main reason for overclocking as well as extracting full value for money from the pc. While typing this I am converting 3 avi's to divx which has been going on for the past 2 hours at 167 x 24 (4 cores it certainly ain't, I must admit) core temp shows the high as 58C but usually 47-52 while doing this. I am aware of possible errors in temp sensing tho air is blowing very cool evn when checking during the P95 test. I'm using the stock cooler with four case fans not inc. the PSU which sits old school at the top of the case.

On the Ram situation I still have the multiplier at 8x so 8x167 giving 1336mhz at posting. CPUz is showing the standard 999-24 with the command rate back to 1t it was at 2t when mentioning the 888-21. However, Everest shows a good result on memory reading and writing 15% faster and latency 5% quicker than it's benchmark for the non-clocked i5-650. Everest shows the timings as 899-24. It is beaten by the 650 on AES but thats to be expected.

I'm not a game player so maxing out my computer for extended periods won't happen. I have used Intel burn test but only 5(!) passes - looking at Windows task manager IBT didn't take the 4 threads up to the max and was giving cooler temps than Prime on max heat setting.

Once the avi conversions have finished I'll set the Intel Burn to go 20 passes on the maximum setting, now you have me thinking about it.
If it fails I'll post a follow up.

An odd observation I made that I am yet to figure is that while CPUz shows core voltage as 1.248, when running at 167x24 it drops to 1.232 and somtimes 1.216 just after the CPU fan makes it loudest grumble for 3-5 secs.

Any other queries feel free to ask I owe you and the forum and it may help you to help others as you have helped me out
a b à CPUs
a c 100 K Overclocking
January 10, 2011 4:06:22 PM

Sounds good. About the voltage, that's Vdroop and it's normal. Most performance mobos allow you to turn on Load Line Calibration which evens the voltage out, but that's for serious overclocking. You're fine :) 
!