Which MB supports overclocking with speedstep?

All about them.

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

Post Reply
ojg
Friend of SPCR
Posts: 34
Joined: Thu Aug 07, 2003 8:08 pm
Location: Norway

Which MB supports overclocking with speedstep?

Post by ojg » Thu Nov 20, 2008 7:05 am

Hi all,

I'm looking at getting a new motherboard to go with an E8500. I would like to do some light overclocking to use the potential of this CPU. However I would not do it if it meant turning off EIST.

After some extensive googling I can not figure out for sure which motherboards support this or not. Most of the oc forums always disable EIST out of principle so there is not much help to get from them.

So could we put together a list here of motherboards (and bios versions) that we know for sure:
A: Supports changing FSB with EIST enabled and working, and if so
B: Supports changing Vcore with EIST enabled and working.

Personally I would be mostly interested in P45 motherboards but let's make the list as complete as possible. First hand experience is best, but a link to a trusted source is also good.

Thank you!
OJG

QuietOC
Posts: 1407
Joined: Tue Dec 13, 2005 1:08 pm
Location: Michigan
Contact:

Re: Which MB supports overclocking with speedstep?

Post by QuietOC » Thu Nov 20, 2008 8:26 am

ojg wrote:After some extensive googling I can not figure out for sure which motherboards support this or not. Most of the oc forums always disable EIST out of principle so there is not much help to get from them.
Generally EIST/C'n'Q works with fine overclocking. Most motherboards will lock a CPU voltage if you set a specific one in the BIOS. Unfortunately, voltage reductions saves a lot more more than mere clockspeed reduction.

So it is a little tricky to both overclock a lot and reduce power, but it is possible. It generally looks like this:

1. Enter BIOS

1a. Lower the CPU mutlitplier in BIOS and set CPU to default/automatic voltage
1b. Lower memory ratio
1c. Raise FSB/HT clock
1d. Boot into Windows

2. CrystalCPUID/RMClock Overclocking

2a. Raise CPU voltage to highest setting
2b. Raise CPU multiplier

3. Orthos or some other CPU stress test
4. Repeat Steps 2b-3 until Step 3 fails
5. Program overclocked state into CrystalCPUID

6. CrystalCPUID/RMClock Undervolting

6a. Lower CPU multiplier to lowest setting
6b. Lower CPU voltage

7. Orthos or some other CPU stress test
8. Repeat Steps 6b-7 until Step 7 fails

9. Program undervolted state into CrystalCPUID

10. Have CrystalCPUID run at Windows startup and manage EIST/C'n'Q

frank2003
Posts: 332
Joined: Sun Jun 24, 2007 3:35 am

Re: Which MB supports overclocking with speedstep?

Post by frank2003 » Fri Nov 21, 2008 5:05 am

Aren't overclocking and EIST contradiction in terms? My understanding of EIST is that it goes by some table of predefined vcore and multiplier pairings. If you bump up the FSB clock and the EIST logic uses the pre-defined vcore in its table wouldn't your system become unstable (i.e. the vcore is lower that what it needs to be for the increased clock frequency)?

In my experience with AMD, the BIOS only turns on voltage and frequency controls if you turn off CnQ. I'd imagine Intel boards would do the same. Because not doing so just doesn't make sense. Just my $0.02.

QuietOC
Posts: 1407
Joined: Tue Dec 13, 2005 1:08 pm
Location: Michigan
Contact:

Re: Which MB supports overclocking with speedstep?

Post by QuietOC » Fri Nov 21, 2008 5:32 am

frank2003 wrote:Aren't overclocking and EIST contradiction in terms?
Definitely not!

I think most people would like to have both working. I have been doing both undervolting and overclocking using C'n'Q for many years.
My understanding of EIST is that it goes by some table of predefined vcore and multiplier pairings. If you bump up the FSB clock and the EIST logic uses the pre-defined vcore in its table wouldn't your system become unstable (i.e. the vcore is lower that what it needs to be for the increased clock frequency)?
Both EIST and C'n'Q are mechanisms for lower CPU frequency and voltage to save power when the CPU doesn't need as much performance. The preprogrammed proccesor states normally use more voltage than is neccessary. So, yes, even without tweaking the voltage levels, EIST can work fine with an overclocked processor.
In my experience with AMD, the BIOS only turns on voltage and frequency controls if you turn off CnQ. I'd imagine Intel boards would do the same. Because not doing so just doesn't make sense. Just my $0.02.
None of my AMD motherboards have required turning off C'n'Q. At worst they lock the CPU voltage to whatever value is set in the BIOS.

The great thing is that you can roll you own C'n'Q/EIST states using software. This allows both overclocking and saving more energy than running the AMD/Intel preprogrammed C'n'Q/EIST states. But it does take some work.

ojg
Friend of SPCR
Posts: 34
Joined: Thu Aug 07, 2003 8:08 pm
Location: Norway

Post by ojg » Fri Nov 21, 2008 7:26 am

Ok, so from what I understand now is that the motherboard or bios doesn't matter when using software based "speedstep" such as CrystalCPUID.

The need to have some software running in the background is a compromise I could live with. However It would be nice if the bios could handle it by itself, that way it would work the same in both windows and linux.

My prefered chioce would be the Asus P5Q-Pro or the Gigabyte GA-EP45-DS3R. Anyone tried overclocking these with EIST enabled?

maf718
Posts: 247
Joined: Sun Jul 13, 2008 7:25 am
Location: England

Post by maf718 » Fri Nov 21, 2008 9:00 am

I overclock with EIST enabled on a Gigabyte P35 DS3R (shouldn't be too different to a P45) and it drops the multi at idle but not the voltage. As QuietOC suggests above, I think this is the case with most if not all boards.

Out of interest I measured my idle power consumption at stock with EIST working properly and compared it to my results when overclocked at fixed voltage. The CPU was an E2140 (M0 stepping), stock speed 1.6ghz (8x200), and overclocked at 3ghz (8x375) 1.3v, in both cases the multi was dropped to 6 at idle. The overclocked configuration only consumed 7w more than the standard, measured at the wall. Disabling EIST entirely so the multiplier was forced to remain at 8 added maybe another watt at most.

I was surprised that the difference was relatively small compared to total system power consumption, and could easily be outweighed by making savings elsewhere.

Even so, I am considering using CrystalCPUID to try and get back some of those 7 watts at idle, but have been put off by the lack of easily available information on configuring CrystalCPUID. (I've seen a couple of guides but they are probably two or more years old and concentrate on AMD processors). Maybe it's easy to use and I could get by without a guide or maybe there are some good sources of info I don't know about? any suggestions welcome.

ojg
Friend of SPCR
Posts: 34
Joined: Thu Aug 07, 2003 8:08 pm
Location: Norway

Post by ojg » Sun Nov 23, 2008 5:52 am

Thanks maf718, that's the kind of info I like to get.

I just read a review of the EP45-DS3 on xbitlabs and while reviewing the voltage settings in the bios they said: "if the CPU, chipset and memory voltages are set to Auto, the mainboard will increase them on its own during overclocking: the more you overclock, the higher will they be."

and: "you can set each and every parameter to a specific value or Auto, but also to Normal. In this case, the voltage will remain default, no matter how far you are trying to overclock and hence the power-saving technologies will keep working just fine."

Have you tried setting the voltages to Normal to see if it drops the vcore automatically at idle?

maf718
Posts: 247
Joined: Sun Jul 13, 2008 7:25 am
Location: England

Post by maf718 » Sun Nov 23, 2008 9:43 am

Sounds interesting, I'd always thought it was the same as auto but I'll have a check later if I get time, and see if it works.

maf718
Posts: 247
Joined: Sun Jul 13, 2008 7:25 am
Location: England

Post by maf718 » Sun Nov 23, 2008 4:14 pm

Thanks for the tip ojg, it works!
I set "system voltage control" in BIOS to manual and then set CPU voltage to normal, (all other voltages were already normal) rather than a specific value, set the fsb for a modest overclock and it worked. Speedfan and CPU-Z both report 1.12v at idle. The nominal voltage of this chip is 1.2375v, reported voltage at load is 1.184v.

[I may even leave it on this setting if I can get it stable at a semi-decent overclock. So far its 1 hour Orthos stable at 2.66 GHz (333fsb), which is good going for 1.2375v because I already know the chip needs 1.3v to be stable at 3GHz, and silly voltage (1.4v+) to be stable at 3.2.]

Edit: And power consumption at idle is only about 1-2 watts more than with stock clocks.

gabeyd
Posts: 52
Joined: Fri Apr 30, 2004 1:12 pm
Location: San Jose, CA USA

Post by gabeyd » Sun Nov 23, 2008 9:54 pm

Wow, great, that works on my Gigabyte EP45-UD3R too! Set all CPU voltages to "Normal" in the BIOS, and now I'm running my E8400 at 3.6GHz 400x9 stock voltage (1.25) and it drops to 2.4GHz 400x6, 1.1 V.

I have an E0 stepping C2D E8400.

Works great! Prime 95 testing at 3.6GHz now. I wasn't going to overclock when EIST didn't work.

So, I can recommend the Gigabyte EP45-UD3R.

ojg
Friend of SPCR
Posts: 34
Joined: Thu Aug 07, 2003 8:08 pm
Location: Norway

Post by ojg » Sun Nov 30, 2008 5:17 am

Thanks guys! This is why I like this forum so much!

I also found this link: http://forums.tweaktown.com/f69/setting ... ble-27640/ where a guy oc's a Q9550 on the EP45-DS3R while having EIST enabled.

Since we haven't heard anything from the ASUS camp, I am going to go for the EP45-DS3R board.

Here is my strategy:
1. Boot at default settings. Note the vcore voltages at idle (x6) and full load (x9.5)
1. Set multiplier to x6 and the vcore to the x6 voltage noted in step 1
2. Increase fsb until it becomes unstable. Note last stable fsb
3. Set multiplier to x9.5 and the vcore to the x9.5 voltage noted in step 1
4. Increase fsb until it becomes unstable. Note last stable fsb.
5. Choose the lowest fsb of the two found in step 2 and 4. This should then determine the highest fsb to safely use with eist enabled and vcore set to "normal".

ojg
Friend of SPCR
Posts: 34
Joined: Thu Aug 07, 2003 8:08 pm
Location: Norway

Post by ojg » Fri Dec 12, 2008 12:01 pm

Hello again!

My new system has arrived and is being tested as we speak. The major parts:
Gigabyte GA-EP45-DS3R with F10 bios
Intel E8500 E0 stepping
Thermalright HR01 with Nexus PWM fan
Crucial Ballistix PC6400 2x2GB which failed memtest86, so I had to use my old Crucial standard PC6400 2x1GB instead :(

The stock settings were 1.056V Vcore at idle and 1.200V at load as reported by CPU-Z. This is about 0.05V lower than what is set in bios. (Is this what is called Vdroop?)

As long as Vcore is set to Normal in the bios EIST reduces both multiplier and Vcore when FSB is increased.

With multiplier set to 6 and Vcore reading 1.056 it was stable at 380MHz and crashed at 390MHz.
With multiplier set to 9.5 and Vcore reading 1.200 it was stable at 380MHz and crashed at 390MHz. A bit disappointed as I was hoping for more.

Right now I'm writing this while running at 420MHz/4.0GHz 8) with Vcore set to Auto which CPU-Z reports as 1.315V.

No problem with heat, the HR01 seems great. Speedfan detected the fan speeds correctly but cannot control them even when I disable fan control in the bios. I have to look more into this later, but for now I am letting the bios control the fan and that works pretty well. In Easytune6 it's possible to set the range of speeds and temperature that the fan shall operate over.

discopig
Posts: 26
Joined: Sun Jun 15, 2008 8:00 pm
Location: Edmonton, AB

Post by discopig » Sun Dec 14, 2008 9:11 pm

I have an MSI-P6NSLI-FI paired with an e6600. I have it oc'd to 3.4ghz with a little bump in volts for stability and have EIST enabled. Both the voltage and FSB vary according to load seen in CPUz.

ojg
Friend of SPCR
Posts: 34
Joined: Thu Aug 07, 2003 8:08 pm
Location: Norway

Post by ojg » Tue Dec 16, 2008 11:43 am

That's interesting. Does the MSI reduced the Vcore to the original value when idle or is the idle Vcore slightly higher as well?

An update on my progress:
- When Vcore is at "Normal" -> EIST adjusts Vcore
- When Vcore is at "Auto" -> EIST adjusts Vcore as long as FSB=333MHz
- When Vcore is at "1.25V" -> EIST adjusts Vcore with any FSB setting! So this is what I have been using.

I was only partially correct about the Crucial Ballistix. It passes memtest but at slower than specified settings. It was my memory that was holding me back at 380MHz. When I relaxed the memory timings I could get the FSB up to 480MHz * 9.5 = 4.6GHz with lots of extra Vcore.

Even though I spent ekstra money for PC6400 CAS4 ram they are not stable at the SPD setting at even 840MHz. So I am trying different settings to find a stable one. It's running prime95 at 4.0GHz at stock voltage, but it isn't 24/7 stable. I'll be tweaking some more because 4.0GHz is an important psychological milestone :-)

SebRad
Patron of SPCR
Posts: 1121
Joined: Sun Nov 09, 2003 7:18 am
Location: UK

Post by SebRad » Mon Dec 22, 2008 8:42 am

Hi, high performance RAM often needs higher voltage to achieve it's rated settings. The default voltage for DDR2 RAM is 1.8V but your kit is rated for 800MHz 4,4,4 at 2V. You should be able to change the RAM voltage easily in the BIOS. Maybe even go to 2.1V to see if will work at stated speeds, if not Crucial will should replace with minimal hassle.
Regards, Seb

alecmg
Posts: 204
Joined: Thu Mar 13, 2008 5:56 am
Location: Estonia

Post by alecmg » Tue Dec 23, 2008 2:23 am

I had a good experience with Gigabytes AI Overclocking (creepy I said that)
Basically it runs on default FSB and Voltages (EIST working) until some serious load, then ups FSB (keeping the default load voltage). Given a 45nm cpu, it can run with max available turbo setting (333 to 405 FSB) without any problem. At idle it still retains 333x6 and default EIST idle voltage. And with small bonus of underclocking the memory in idle.

flapane
Posts: 257
Joined: Fri Dec 19, 2008 2:55 am
Location: Naples, Italy
Contact:

Post by flapane » Tue Dec 23, 2008 7:33 am

I have a P5QL PRO and there isn't any way to adjust vcore in idle, while overclocking cpu. It always remain at 1.096v while at stock and with asus epu 4 engine, it dropped at something like 0.8v
Furthermore I have only 3 possible voltages, but it doesn't happen anything, the voltage always remain at 1.096 in idle and full load
Image

Post Reply