Testing new build - thx to Soullessone21

Todays I recieved a gift from dear John, aka Soullessone21, that will somewhat update my computers into current state of things. In short, in a very nicely and carefully packed box arrived:
MSI P55A-GD55 - Intel P55 chipset, socket 1156 mainboard rev. 2.1
i5 750 1024k L2, 8192k L3, 2660Hz 133x20 1.120Vcore
2x 2048MB DDR3 1600MHz G.Skill RipJaws 9-9-9-24 1.5V (F3-12800CL9D-4GBRL)
2x 2048MB DDR3 1600MHz Kingston HyperX 9-9-9-27 1.9V (KHX12800D3K2/4G)
eVGA GTX 660 Superclocked 1046/1111MHz core, 2048MB GDDR5 at 6008MHz

So I get into work to test this out, hoping my old Enermax 620W Liberty will be up to the task. The most challenging thing was to work out the cables in the Bitfenix Merc Alpha, because I need the PATA HDD get connected and the abandoned connector is just at the end of the mainboard, bellow the last PCI slot:
…so it is really not well accesible in mid case, that put the PSU just on side with the end of the mainboard. But I pulled thru and get things working, the MSI mobo posted right away and at 39°C the CPU temp seems well reasonable. With no previous experience with such mainboard, I just went into the bios (old school one) and checked that everything is properly detected, etc.

Then the build was used to run the Sapphire A9RX480 mainboard (Opteron 148), so when the Windows tried to boot, it failed and reset come each time. Then I just disabled the Spread Spectrum (recommended for better O/C setting) and tried some Memtest and suddently the machine started booting into Windows 2000 SP4 Czech...!
(it probably just tried to boot from the PATA HDD and when I set the SSD as booting device, then things started working, the Spread Spectrum setting have nothing to do with it and I cannot see the PATA HDD in Windows, lol)

I was in shock, but these old Winblows booted up and started installing things to run...! It took many, many and many reboots till it install all it could and give me a mouse (USB mouse using for 1st time) ... but it works! I could not believe it, but it does work:

Sure, it disable the additional cores, effectively turning 4 core CPU into 1 core, but still… with Opteron 148 I get the SuperPi 1M test in about 39.7sec, now it took just 13.5sec 🙂
The graphic card is not installed yet, using a MSI GTX 210 as passive low-end testing card first.

So there is a proof, that Windows 2000 SP4 can work on MSI P55A-GD55 🙂 Not well (single core) and probably the 8G ram will be somewhat inaccesible, yet still it works.

About the overclock - well, I just pressed the OC genie, hit 2x the + button and suddently from 2.66GHz I got a 3GHz i5 750. So even the most lame users can do some overclock with this toy. 🙂 Definitively this is an interesting MSI mainboard - it even have a USB 3, so it does look very promising to me.

Next step - GTX 660 🙂 Can it run under Win2k? Probably cannot, as the drivers will not support W2k, but... that will get tested 😄

What I find most surprising is, that I managed the USB to work. I mean… come on. This is Windows 2000 and the mainboard have Intel P55 chipset, that was way more recent... that this piece of software. Yet still - mice work and USB data transfer too.
What does not work are 8G of ram:


3 398 708kBy only. Mind my slightly modified Windows 🙂 I do that for fun  😉 This screenshot was taken using the eVGA GTX 660 Superclocked card, witch itself have 2G of videoram, witch probably pose a challenge to Win2k itself alone, lol. So it is probably time to move on on a system, that could address the ram, the GFX card and… etc. 🙂 Even the build will still be limited in many ways, it will be something usable these days.

A big THANKS to John, of course  😉

And look, how fast I learned to overclock the mainboard - I just copied what the OC genie do, but this time I get into some manual tweaking and - 3.2GHz is not a problem:

Not enought? Okay, I push a bit more that 3.4GHz is stable too:

There is probably not much Win2k results on i5, lol. Never the less, the overclocking of this mainboard is straight-forward and w/o any troubles at all. Only problem I hit, when I forget to lower the ram dividers and tried to force the rams to crazy speeds the chips cannot take. Now they are running at their 800MHz (1600 effective) as they should, so I'm sort of happy.
Vcore +0.278V
VTT - 1.314V
Vdram - 1.701V

Only I did not understand the "turbo mode." On the bios screen it report 3.4GHz… and then on next line it suggest "OS speed" and with x21 multiplier it says 4200MHz. So should I expect the turbo taking me to 4.2GHz or what the hell is going on?

Anyway, 3.4GHz with one core give 12.2sec on SuperPi 1M test, witch is not bad at all.

The build is not looking any spectacular, but at least (given the PATA horror) I tried best with the cables:


I like it looks great

But don't thank me. The system is from the whole RHR team 🙂

Well, then I have to thank the whole RHR Team 😉 But you seems to be the one who do most of the job anyway 🙂 So kudos go to you first.

And well, hope you guys like the fun. I was amazed that it worked at all. Win2k did NOT seems to be even registred as OS by the CPU/mainboard. How do I know it? Because after that, I tried WinXP and the install crash/hang. Same Win7. And then I come to realize, that the turbo multiplier kick in and really I was forcing the poor i5 to run on low-profile small stock air cooler at 4.2GHz, witch it cannot take.

So for now, this turbo is disabled and I will controll the clocks from now on:

On WinXP SP3 (32bit) all the cores are present and utilized, so that is a step forward. About the ram = same story. 3.3G top. Still, looking at FurMark bench from 177 to 304fps is kinda awesome 🙂

Next step... Windows 98se? :))) Nah, that would be too insane. But it was fun... 🙂

Upon futher testing - the i5 750 Lynnfield CPU is quite good overclocker. The OC genie by MSI play it quite harshly, giving it +0.278V, terrorizing the poor CPU with 1.304Vcore ( http://s1.postimg.org/gmlwdz4lb/Win2k_i5_750_overclock2.jpg ) for stable OC.

When in manual mode, I just find out, that as slight bump as +0.074V is enought to 3.4GHz of stable 4 cores operation.

Difference between 1.304Vcore and 1.088Vcore is IMHO quite significant:


…and there is no telling, if the +0.074V (in bios, it show as +0.088Vcore voltage bump over 1V, assuming that default Vcore is 1.000V, in CPU-Z) is even necessary to get from 2.66GHz to 3.4GHz. Of course, the turbo clock 3.2GHz is what is officially supported with default Vcore, somaybe with good cooling it might do 3.6GHz stable w/o much voltage bump, witch is promising 🙂

Sadly I have to report that after 4 days of nice gaming the eVGA GTX 660 Superclocked died on me:

+0.074V is not enought past 3.4GHz, but +0.105V and 3578MHz is stable:

+0.129V is not yet stable for 3.65GHz, but very closely:
...maybe +0.136V will do stable? 🙂 Still far cry from +0.278V by OC genie.

Cool 🙂 Curses go to eVGA, tough.

That chip should do 4ghz without sweating. Although my experience with Lynnfield chips is they don't like odd # multipliers when you start ramping up the clock speed. Also these chips liked lots of VTT voltage vs Vcore voltage.

Well, tht will probably require a much better cooling first 🙂 I hit the wall with 3.65GHz cannot run stable - it always freeze, even with much higher Vcore, so the problem must be somewhere alse.
What VTT you suggest to try? Currently VTT - 1.314V and with Vcore 1.112V it do 3578MHz stable. Bumping the Bus speed from 210 to 215 resulting in 3663MHz, witch always freeze the machine during tests, even with lowered the memory speed to make sure, that this is NOT because of memory.

So you say more VTT? How much more? 🙂

So, little test. I bumped the Vcore to +0.204V (still far cry from +0.278V that the OC genie used) and bumped the VTT to 1.386V (last one before the color get red in the bios) - and x18 multiplier and hell, it booted right away! Whoa!
3780MHz is a bit step from 3578 … Holly, holly...


…it still freeze in the test, tough  :-[

I try to get some tips from guys that are sucesfull and know, what these chips want and how to get them to run fastest.

The basic specs of my i5 750 Lynnfield are 2.66GHz with turbo 3.2GHz, witch looks slow to me.

About the overclock - well, I just at first used the OC genie, but that won't get me far (3GHz base clock, IIRC 3.4GHz turbo).

First I just copied what the OC genie do, but then I get into some manual tweaking. OC genie pushed Vcore hard:
Vcore +0.278V
VTT - 1.314V
Vdram - 1.701V

I find out, that with Vcore just mere 1.112V it do 3578MHz stable:


However I hit the wall with 3.6GHz cannot run stable - it always freeze, even with much higher Vcore, so the problem must be somewhere alse.

Bumping the Bus speed from 210 to 215 resulting in 3663MHz, witch always freeze the machine during tests, even with lowered the memory speed to make sure, that this is NOT because of memory cannot cope with the speed, especially with tightened timings 9-9-9-18 😉 CPU-Z validation can be made, tough:

But it is not stable. I overheard the VTT might need a bit more "love", but I figured that I better ask fellow overclockers first 😄 Currently there is just a suxxking default intel noisy heatsink, so it is just for the testing:

Futhermore I would like to ask about some bios settings (features of the i5, for example), witch are suggested to be disabled/enabled in order to get maximum speed. IIRC one czech testing server always disable virtualization, so that probably should be a good start. Of course disabling the Spread Spectrum to get overclocking going, but there are far more options I'm interesed about, like:

Intel EIST - IMHO off, because this is not NTB that need baterry support and I want always full clock
Intel C-State - IMHO off, because conserving the energy is not the speed goal
C State package limit Setting - IMHO off, because I did not want any regulations
C1E Support - IMHO off, because sky is the limit and I don't want any other limits there
OverSpeed Protection - IMHO off, even MSI recommend that for OC
(between these CPU features are in manual mentioned Hyper-Threading Function. but I did not have it in bios)
Execute Bit Support - I leaved this on, but speed wise off is better, no?
Set Limit CPUID MaxVal to 3 - IMHO off, because who cares about old systems when overclocking, right?
Intel Virtualization Tech - IMHO off for speed?
Intel VT-d - IMHO off for speed?
Active Processor Cores - All - simple there, huh?

Mainboard manual: http://www.mediafire.com/?4vdh50m7ta0995w - https://www.oboom.com/RL8DUVCJ

And then there are the voltage settings:
CPU Voltage(V) - +0.105V
CPU VTT(V) - 1.314V
CPU PLL Voltage(V) - Auto
DRAM Voltage(V) - 1.701V
PCH 1.05(V) - Auto
…any suggestions about how to break the 3.6GHz wall?

And the there are the settings like:
PCI Latency Timer - IMHO the smaller the better, choosen 32 as lowest value
HPET - High Precision Event Timers - no idea what to set, maybe off for speed and less precision?
TCG/TPM SUPPORT - Trusted Platform Module can be enabled/disabled - what is best for speed?

…and don't even get me started about the memory settings. There is 8G consisting of:
2x 2048MB DDR3 1600MHz G.Skill RipJaws 9-9-9-18 1.7V (F3-12800CL9D-4GBRL)
2x 2048MB DDR3 1600MHz Kingston HyperX 9-9-9-18 1.7V (KHX12800D3K2/4G)

So far these seems stable at 860MHz (1720MHz effective) and 1.701Volts, so maybe later some possible finetuning could get handy too 🙂  Current ram settings:

Memory Timing - 2T (Suuuuuuxxx!)
CAS - 9
tRCD - 9
tRP - 9
tRAS - 18
tRFC - 88
tWR - 12
tWTR - 6
tTRRD - 5
tRTP - 6
tFAW - 24
B2B-CAS Delay - 0

tdrRdTRd - 6
tddRdTRd - 7
tsrRdTWr - 11
tdrRdTWr - 11
tddRdTWr - 11
tsrWrTRd - 16
tdrWrTRd - 4
tddWrTRd - 4
tdrWrTWr - 7
tddWrTWr - 7
tsrRdTRd - 4
tsrWrTWr - 4
Round Trip Latency - 59 (channel 1) / 61 (channel 2)

Surely something could be made faster there 🙂 Tips welcome  😉