OC log: Penryn Quad on GA-X38-DS4

[Mathis]

Intel has decided that you can bleep off if you want to gain results that you didn't pay for I guess...

Anyhow, I think that people have pretty much agreed that a 3.6 GHZ Q9450 will most likely be more useful in the long run than a 4.0-4.2 GHZ E8400 or a 3.4-3.6GHZ Q6600 GO.

But it makes the overclocking seem so much less fun since no matter what we do we have to deal with that (purposely?) built in failure.

 

[Albuquerque]

Sorry, my fiance has been soaking up all of my time since my last post. She's in the shower now, so I've temporarily returned to the PC.

Interestingly, 455FSB (3640Mhz) is seemingly stable at bone-stock voltages across the board; it's been dual-Orthosing for the last 50 minutes. I haven't had much chance to fight with it yet, I was more interested in determining where the "stop" was.

Anyway, now that I know it's relatively stable at 455, obviously my next step is to make 456 stable

 

[JCLW]

Do the 65nm chipsets (4X series) overclock the quads any better then the 90nm chipsets (3X series)? If so it would point towards a chipset limit, if not it would point to a chip limit.

Are the Q9450s C0 or C1 stepping?

 

[th3]

Anyhow, I think that people have pretty much agreed that a 3.6 GHZ Q9450 will most likely be more useful in the long run than a 4.0-4.2 GHZ E8400 or a 3.4-3.6GHZ Q6600 GO.

"when more apps support quads", we will be using Westmere and Gesher before that happens, Yorkfield will be older than Methuselah by then.

 

[Albuquerque]

The X38 and X48 are identical silicon; it is only a speed-binning process that seperates them. My hope was that the newer X38 may not suffer as much from the FSB wall; seems that is not the case

Edit:
The fiance is pestering me to go to bed as we both need to wake up early tomorrow. So, unfortunately this thread and my tweaking must wait until 5:30pm EST tomorrow

G'nite.

 

[Mathis]

"when more apps support quads", we will be using Westmere and Gesher before that happens, Yorkfield will be older than Methuselah by then.

Agreed, but I'm speaking in terms of what will be most economical if I had to make a choice now and live with it until it gets recycled for scrap. If I have to completely replace all of my hardware for a new architecture/socket then I am going to go for a fresh build and keep my current rig permanent. The Q9450 looks like a good candidate for the long-term LGA775 IMO. Furthermore, if you do more than gaming with your computer, such as video encoding or any CS3 app, quads are awesome already. If gaming is your only measure of a CPU then you should be happy, but I think I will be replacing my E8400 for the Q9450.

 

[JCLW]

The X38 and X48 are identical silicon; it is only a speed-binning process that seperates them. My hope was that the newer X38 may not suffer as much from the FSB wall; seems that is not the case

That seems to be the rumor here but I'm pretty sure the 40 series chipsets are on the 65nm node.

The big feature is that all of these new chipsets are built on Intel's 65nm process for the MCH vs. 90nm for the older 3-series generation. The smaller manufacturing process should mean that they draw much less power.

http://www.anandtech.com/tradeshows/showdoc.aspx?i=3250&p=2

That's not the only process that Intel is pushing though, because the entire 4-series range is now built on 65nm technology, as opposed to 90nm with the 3-series. So it turns out that X48 will be different to X38 after all...

http://www.bit-tech.net/news/2008/03/07/intel_talks_p45_g45_and_x48/1

Intel's P45, G45, G43, and X48 chipsets, due to launch in Q2 this year, are built on 65nm process and offer less power consumption compared to previous offerings (90nm).

http://www.cdrinfo.com/Sections/News/Details.aspx?NewsId=22690

etc, etc...

 

[satandole666]

I'm terribly interested in the quad results but can I back this up a bit?

You are running 4338 daily on an E8400 at 1.4v? I saw you mentioned water...is it just ambient or are there TECs involved?

Also...what stability measure are you using for those daily settings?

I can run my E8400 around 4300-4400 daily but it won't pass a lick of Prime.

Sorry for the derail...just amazed at your gem E8400.

 

[Albuquerque]

Not sure which one to keep now...

The Q9450 is still running four primes (2 x Orthos) at home at 3,640Mhz (455 x 8) with stock volts across the board (except DDR voltage, which needs to be 2.1 at this speed for stability), temps when I left this morning were in the mid to high 50's with my radiator fans at my preferred 7v (far easier on the ears versus the WHOOOOOOOOOSH at 12v)

The E8400 is 12hr priority 10 in-place 10k FFT's prime stable; I don't need any more than that. It's under "ambient" water, temps usually sit in the high 50's with the fans at 7v.

So, even though I have about 16% less processor speed per core, I've got 200% more cores And funny enough, the quad with it's lesser speed is certainly running cooler than my E8400 -- likely because I don't need any extra voltage anywhere to get it stable. While I was relatively certian 4Ghz was a forgone conclusion, I'd still prefer to see it at 3.8Ghz even if I had to lean on it a little.

I've got a few more hours to tinker with it tonight, and if I cant get it to 475, I'll have a hard decision to make...

 

[Albuquerque]

Ok, so the good news:

3.64Ghz is 2x Orthos stable for 17h 30m when I rebooted my box after coming home from work. Highest core temperature was 51*c with my fans at their usual 7v, which is freakin awesome

Ok, so now the bad news:

I can't get this thing to budge 1FSB over 455. I've tried every voltage and clocking option in my bios, but 456FSb is simply not even bootable no matter what.

Ok, so now here's the good news:

I've broken out the big guns, and I'm rolling my own bios for this board. There are options in here for GTLREF that are not exposed in the user interface, so while I may or may not be able to directly modify them through the user interface, I can certainly adjust their BIOS-default values I can also finally change a few of the annoying bios defaults, like Turbo mode and 64-bit timer versus 32-bit timer, et al.

And of course, I'll let ya know how it goes

 

[phu]

Wow, 455 FSB on stock voltage!!!

I only try 410 so far at stock voltage and it's been good so far. Will play with it more tomorrow.

 

[phu]

Up to 440 FSB now. BTW I got the chip from tankguys too and the VID is 1.1250v.

 

[Albuquerque]

This may take a while until I can get the GTLREF settings exposed in the X38-DS4 bios. It's not as easy as it was back in the Award 4.50PG days

But once I get this nailed down and get a working bios up and running, I'll be able to post some good stuff for my other OCForums bretherin who need a better X38-DS4 bios

 

[Marvelous]

This chip has some serious FSB holes. I can not post at 400fsb for anything, but I am running 433x8(3.46ghz) 8hr 2x orthos stable. This is with a at 1.22v. I could probably go higher but I am on stock cooling and giving it more volts isn't worth it until I have a better heatsink.

 

[phu]

I got to 444FSB last night. This morning I tried 446 and couldn't boot up. Now I can't even boot up at 410FSB. I already cleared CMOS. Any idea what's wrong? It's the EX38-DS4 board

 

[Badussy]

I got to 444FSB last night. This morning I tried 446 and couldn't boot up. Now I can't even boot up at 410FSB. I already cleared CMOS. Any idea what's wrong? It's the EX38-DS4 board

Had the same issue happen to my DS4, but the bios defaulted back to the original since it has a dual bios. I had to redo my HD RAID changes in the BIOS...might not be your issue though.

 

[Albuquerque]

Not sure, mine doesn't have any "holes" that I can determine; I've tried posting in 5mhz increments from 500 to 600 without luck.

Gigabyte appears to use a CRC value not only on the entire bios file, but also on the SYSTEM.BIN module of their bios -- which makes unhiding things like GTLREF rather difficult. I've probably flashed a dozen bioses over the last 36 hours with very little success. I haven't given up yet though...

 

[satandole666]

Keep on chugging man...

It seems like if there is a hope for a 4ghz quad (minus the QX's) on this board it is you so far.

 

[phu]

Back up and running again. I don't understand why I have to set a lot of other settings manually while I didn't have to do that initially. Anyway at 440 FSB I can run prime for a while. I tried 450 and it only lasts 4 minutes in prime. This board seems to be very picky on the memory.

 

[Coreyhm1]

Did you find the GTLREF's? on the main screen you press ctrl+F1