ICS9212AF-13 DRCG's... I should be getting 30 samples

Hi all. I am coming over here from the [H]ardOCP crowd. I am getting 30 of the new 600MHz ICS Patrs. The DRCG's everyone speaks of. They are on the new P4TE's but that is it. I had to go through a lot of crap to get them. I have a P4 1.6a and TH7II without the RAID.

I have the cold boot problem too. Will a bigger power supply aleviate the issue? Is it just a BIOS thing.

one more question. What in the hell does the differential curret optionfor the RAMBUS do? Is it like a voltage adjuster????

Thanks for any help/inquires

Hi and Welcome,
1st, the cold boot problem AFAIK has nothing to do with the PS. I only get this problem when all power to the TH-7-II is off, ie, no green light on the board. Just hit the reset button, might take a couple tries, but mine will always boot.
You could also do the "wire trick", that would make the default vcore higher and should alleviate the problem. Check out the posts here for details.
AFA "differential current", do a search at the Hocp/Strictly Intel Forum, *foo had a good technical explanation.
I've read that setting it to 4X has helped some with OCing, others say it doesn't matter.
And if you want to get rid(sell?) of 2 of the ICS 9212-13 chips, LMK
[email protected]
Mike G

the wire-trick will not solve the cold boot problem, i've tried this

I have read that some people claim it does, but have also heard it doesn't as you say.
I guess I'm confused about the definition of the "cold-boot" problem. When this came up in the past, with a different board and CPU(I think it was the Asus CUSL-2/i815 chipset), it was a problem where the CPU needed higher vcore to be boot and be stable, but the BIOS was only providing default vcore until it went thru initialization? If that's the case, and the wire-trick sets default vcore at the higher level, it seems like it should work?
As I say, I wonder if there are 2 diff. problems being defined as "cold-boot"?
In my case, 1.6A/2.1G/1.5 vcore(stable after many hours of Prime 95 torture Test and gaming) will always boot up fine if the green LED is on even though the computer is off. If I cut all power to the mainboard, and then boot, it stops before the video is initialized but will boot if I hit reset.
As I usually soft-off and there is always the green LED lit, it's not a problem for me.
For me it's another little quirk and a confusing issue as to what we are really seeing happen.

Read up on Mr Naturals' posts over the last two days and you will realise that there is no need to change the DRCG's. And another thing I would like to see the actual literature that states the -03 ones are only rated at 400mhz. So far the only specs at ICS are of them being mated with an ICS clock generator. The TH7II uses a different clock generator which is sending a different range of initial signals to the DRCG's.

From what *foo states and I have verifed by going and looking at all of the current DRCG manufactures' spec hseets in .pdf form. If you would do the same all DRCG's are the same. No different. If you go to the ICS website and look up the part number you will see that it cleary states that it is only rated to 400MHz. I have researched this subject for over two months. 8foo has proven his ability to clock higher with the new DRCG's as well as have others. Please don't knock me when your ignorance is at fault.

Hold on there Cheeseman. There is no need for for harsh words. I happen to agree with Gone_Fishin on a couple of points. With good cooling of both the RDRAM modules and the DRCG chips, lots of folks are getting amazing overclocks out of the current memory bus. You might want to look at Mr. Natural's efforts using pre-release PC-1066 RDRAM in recent posts like was suggested.

On the other hand, I agree with Cheeseman on a couple things too. I have also seen the ICS chip specs which are available on their website and it does clearly state the DRCG -03 chips are "rated" at 400 MHz. But, I can overclock them to 533 with a little extra cooling. Furthermore, my Northwood CPU is also only rated at 2 gig, but I can run it at 2.6+ gig. So, most DRCG -03 chips do overclock quite well.

That said, I have no doubt that the DRCG -13 chips which are rated even higher (600 MHz I think), probably overclock much better yet. I think what Gone_Fishin was driving at was that changing the chips from -03 to -13 is probably not really all that practical or needed for the typical overclocker at this current time since the RDRAM modules are more of the limiting factor right now, not the DRCG chips.

However, once the new PC1066 RDRAM modules are readily available, then that might all change. If Cheeseman or anyone else wants to go to all the trouble of changing out the ICS chips and risk damaging their motherboard for a little extra memory speed, then by all means, go for it. If you do, then please post your results here. Peace.

Cheeseman TT, I didn't mean to come off as insulting. Sorry you took it that way.
Specs for ICS 9212-03
http://www.icst.com/products/pdf/9212-03.pdf
Specs for ICS 9212-13
http://www.icst.com/products/pdf/9212-13.pdf

Now it states at the top of the specs 400mhz for -03 and 600mhz for -13 but if you look at the charts in each one they are identicle. The formula for the RDRAM speed is ( PLLCLK = REFCLK*A/B). Now the variables for A and B are identicle for both. REFCLK is the signal from the TH7II clock generator which we would plug in place of the ICS clock generator they used as stated in these pdfs. Their charts only go up to 533mhz for both (the -03 one is listed as reserve). The only listed difference is that the -13 has more bypass and test mode selections period, the only difference. And a logical explanation to all of this is that the -03 was released earlier and they couldn't confirm even the existence of a 533mhz operation. The additional bypass and test modes were most likely for use by motherboard manufacturers. So you see the -03 has the multipliers for 533mhz operation and it has been confirmed to work. Interesting to note is their claim of 600mhz operation would only work if Intel released a processor which ran on a quad pumped 150mhz FSB at default because they are only listing the defaults not overclocks. Maybe they know something we don't.

I don't know how many people I've told that I seriously doubt the dcrg's are their limiting oc factor. No one wants to beleive it though??
oh well to each his own. It looks like one hell of a tedious job and like batboy mentioned the risk of hosing your board is looming over head as well.

gone_fishin said:

Cheeseman TT, I didn't mean to come off as insulting. Sorry you took it that way.
Specs for ICS 9212-03
http://www.icst.com/products/pdf/9212-03.pdf
Specs for ICS 9212-13
http://www.icst.com/products/pdf/9212-13.pdf

Now it states at the top of the specs 400mhz for -03 and 600mhz for -13 but if you look at the charts in each one they are identicle. The formula for the RDRAM speed is ( PLLCLK = REFCLK*A/B). Now the variables for A and B are identicle for both. REFCLK is the signal from the TH7II clock generator which we would plug in place of the ICS clock generator they used as stated in these pdfs. Their charts only go up to 533mhz for both (the -03 one is listed as reserve). The only listed difference is that the -13 has more bypass and test mode selections period, the only difference. And a logical explanation to all of this is that the -03 was released earlier and they couldn't confirm even the existence of a 533mhz operation. The additional bypass and test modes were most likely for use by motherboard manufacturers. So you see the -03 has the multipliers for 533mhz operation and it has been confirmed to work. Interesting to note is their claim of 600mhz operation would only work if Intel released a processor which ran on a quad pumped 150mhz FSB at default because they are only listing the defaults not overclocks. Maybe they know something we don't.

Click to expand...

So what is the recommended procedure to get the 03's to overclock well? Just cool them better? I'm just wondering because I have a pair of samsung rimms that would do 133 on an old P4T that had the supposedly better drcg's, but now I've got a P4TE with crypress chips and I can't even get to 125 4X. I have to drop back to 3X to go any higher. Just kind of makes me wonder what's going on. I know the rimms will go higher, and I know the cpu is not the factor because it will run just fine at faster speeds as long as I slow the ram down.

Don't know anything about Cypress chips, the ones discussed here are ICS chips. You would need to look them up to see if they are capable of running at that frequency.

gone_fishin said:

Don't know anything about Cypress chips, the ones discussed here are ICS chips. You would need to look them up to see if they are capable of running at that frequency.

Click to expand...

Well I'm ditching the P4TE - have brand new TH7II sitting here waiting for some free time to do the wire wrap. So I will be using the th7ii with the 03's.

TC- the cypress chips are I think rated for 533. They have 2 parts. The NEW P4TE have been coming with the ne ICS chips rated for 600MHX. THat is my guess why you are stuck at 125. With the higher rated chips, there is less jitter in the clock, which RAMBUS is very critical of. That is proven from *foo's woprk over on [H]ardOCP.

Oh and gone_fishin, I am sorry bro. My new machine has had 2 casualties this week. The Diode on the P4 died because I fuxored a wire rap and now my floppy drive blows up (literally) and my WD 80 gig died today. So far today has sucked.

CheesemanTT said:

TC- the cypress chips are I think rated for 533. They have 2 parts. The NEW P4TE have been coming with the ne ICS chips rated for 600MHX. THat is my guess why you are stuck at 125. With the higher rated chips, there is less jitter in the clock, which RAMBUS is very critical of. That is proven from *foo's woprk over on [H]ardOCP.

Click to expand...

I looked up the part #'s for these cypress chips and the pdf says 400 - so I guess that's why I'm stuck. Maybe I'll have better luck with the abit

well I have the 30 some ICS 600MHz parts now, I got them yesterday. Over half are spoken for. Just incase someone wants to grab a couple.

I'm just curious about how folks are going about replacing those ICS chips. Looks like they are soldered to the motherboard to me. Is there a link to a post where someone decribes how they managed to do this mod?

"With luck and much patience"

Foo did explain somewhat how to do the mod, but of course, he's an engineer with the tools.

gone_fishin said:

Specs for ICS 9212-03
http://www.icst.com/products/pdf/9212-03.pdf
Specs for ICS 9212-13
http://www.icst.com/products/pdf/9212-13.pdf

The only listed difference is that the -13 has more bypass and test mode selections period, the only difference.

Click to expand...

In looking at these specs, I did notice one other difference, but I have no idea what it means. Under "Features" on the first pages and in the chart on the fourth pages it lists "Cycle to cycle jitter is less than 50ps" for the -03's and "less than 60ps" for the -13's. What does this mean? Is it anything that might account for a difference in performance?

The reason I ask is that there is an article today over at Hot Hardware called "Getting PC1066 Level Performance From PC800 RDRAM & PC800 Motherboards - A modification project on the Abit TH7II-RAID" where they replaced the stock ICS -03 DRCG's with TI's rated for 533MHz. Here is the link: http://www.hothardware.com/hh_files/Motherboards/abitth7iidrcg.shtml

A couple of things bumped me about the article, but this is so not my area of expertise, so maybe someone can help me out. In the article it says:

"With a 133MHz FSB and a 4X multiplier, you'll be hitting 533MHz SDR and 1.066GHz DDR on the RDRAM clocks. However, there is a good possibility you'll need an unlocked Pentium 4 for that because, there are very few chips with standard air cooling, that will hit a 33% overclock, with the exception of the 1.6GHz variety of the Northwood."

Maybe I've been absorbing the threads in this forum wrong, but I thought a number of people have been reaching a 133MHz FSB with P4's other than the 1.6GHz, like with the 1.8 and some even with the 2.0. Yes? No?

Also, the article says:

"the ICS DRCGs that come standard on this board, will only handle about 468MHz or 117MHz FSB for overclocking, without getting very flaky with the RDRAM."

Once again, has this been people's experience here? And when they use a phrase like "will only handle about" it makes it sound like it's just sort of a fact floating around out there rather than their experience with this particular setup. That seems to be the flaw with this article, unless I'm reading it wrong. To say:

"we were able to overclock our 2.4GHz Pentium 4 faster than we ever could before, with standard air cooling. We used a Thermaltake Volcano 478 to cool our CPU and this little trick on our P4 Northwood (the vcore mod,) to get the core voltage up a little higher than the TH7II-RAID allows, to 1.9V."

implies that they did other things differently this time in addition to changing the DRCG's that could have accounted for the better overclock. There are no comparison benchmarks with an identical setup using the original DRCG's to see a direct result of the better overclock coming purely from the change in DRCG's.

Sorry to go on about an article from another website, but it seemed relevant to this thread and to this Abit board forum. I'm curious as to what you all think about this latest information as I have all the parts here ready to put together an TH7II-RAID system with a 2.0a P4 and 1GB RDRAM and want to get every ounce of performance out of it I can. And the DRCG's may be a factor in that.

This is all new to me, so I would love to hear what you guys think. I'm a newbie, so forgive me for running off at the keyboard.

Peace,
Eggroll

Interesting article over there ... nice re-soldering.

but when I look at the sandra screenshots, I have to ask myself why they've been doing this mod.

my 2.68GHz@168FSB@504Mem is just a fraction behind at the CPU benchmark, and kick's their 3GHz@133FSB@532Mem's *** at the Memory benchmark.