I got my 1.6a P4 from Newegg yesterday. With the P4T-E motherboard, Swiftech MCX478 with quiet panaflo fan, voltage at 1.70 (haven't tried anything else), and ram set to 3x I can run completely stable at 156x16=2496mhz. 156 is as high at the P4T-E will go. I'm currently running at 144x16=2304mhz with the ram at X4 (576mhz). The RDRAM I'm using is fairly old, so I've got more arriving tomorrow and hopefully I can get the ram to PC1200 speed (150 mhz fsb).
Here's the details from the box. Pack Date 01/25/2002, FPO/Batch 3149A726, SL668. Before installing on the motherboard I did the 'wire trick' to boost the default processor voltage to 1.70. Since I have to use the dip switches to set fsb frequency the voltage isn't adjustable in the bios.
I've got quite a few sticks of Samsung RDRAM to experiment with. The only ones which wouldn't do 133 (PC1066) were 4 new 256mb single sided parts I got a couple of weeks ago. These had date code 0152, which I'm sure means week 52 of 2001. These new dimms have the new 256 mbit parts rather than the old 128 mbit parts. In the past week I'd read that they wouldn't overclock as well, and I found that to be true. They could do 122 no problem, and I didn't try anything between 122 and 133. I had some 0136 double sided 256mb dimms, and they barely did 133mhz, but had problems at 136mhz. I wouldnt trust them long term at 133. I had 8 of the 128mb dimms with 8 parts (128mbit technology) on a single side. These had date codes between 0109 and 0125. Every one of these parts would do at least 138mhz. Surprisingly the oldest parts, the 0109's would do 144mhz!
A couple of likely conclusions can be drawn from this. First, if you've got Samsung RDRAM that was made in the last year or so, it should easily hit 120mhz+. If not the most likely problem is processor, rdram clock generator, or maybe pci or agp overclocking issues (less likely than processor or rdram clock gen in my experience). I guess a marginal power supply could be a factor. Secondly, if you've got new Samsung dimm's made with 256mbit parts you probably aren't going to have much luck at 133 or higher, unless maybe you boost the voltage to the rams or increase cooling. Those parts should still easily be good to 120+ . If you've got single sided 128mb Samsung dimms made in the past year (8 parts on a side not 4), you have an extremely good chance of getting at least 133mhz. Double sided dimms might make it to 133, expecially parts made in the past 6 months (from what others report).
There have been many posts about rdram clock generators over the past few weeks. If you've got a P4T-E with the ICS9212-13 ram clock generators (like mine) you're in luck, and the board should overclock like crazy with a 1.6a or 1.8a processor and Samsung RDRAM. If you've got a P4T-E with the Cypress clock generators (appears to be newer P4T-E motherboards or the old P4T's) you're probably out of luck hitting 133+. I understand some P4T-E motherboards have TI clock generators which should be good for 133mhz. They aren't rated for PC1200 like the ICS9212-13.
The ABit TH7II's with ICS9212-03 are only rated for PC800 use, but I bet the ICS9212-03 clock gen is the same as the ICS9212-13, probably either just not tested for the higher speed operation, or they didn't quite meet spec at the higher speed. My guess would be that they should have a good chance at 133mhz. I just got a new TH7II, but haven't tested it yet.
Bottom line, poor results overclocking Samsung rdram are most likely limitations of the rdram clock generators. The 1.6a and 1.8a processors combined with a good motherboard finally enable us to push P4 processors to 133mhz fsb and beyond and also enable us to push rdram to PC1066 and beyond without waiting until the official may release.
Here's the details from the box. Pack Date 01/25/2002, FPO/Batch 3149A726, SL668. Before installing on the motherboard I did the 'wire trick' to boost the default processor voltage to 1.70. Since I have to use the dip switches to set fsb frequency the voltage isn't adjustable in the bios.
I've got quite a few sticks of Samsung RDRAM to experiment with. The only ones which wouldn't do 133 (PC1066) were 4 new 256mb single sided parts I got a couple of weeks ago. These had date code 0152, which I'm sure means week 52 of 2001. These new dimms have the new 256 mbit parts rather than the old 128 mbit parts. In the past week I'd read that they wouldn't overclock as well, and I found that to be true. They could do 122 no problem, and I didn't try anything between 122 and 133. I had some 0136 double sided 256mb dimms, and they barely did 133mhz, but had problems at 136mhz. I wouldnt trust them long term at 133. I had 8 of the 128mb dimms with 8 parts (128mbit technology) on a single side. These had date codes between 0109 and 0125. Every one of these parts would do at least 138mhz. Surprisingly the oldest parts, the 0109's would do 144mhz!
A couple of likely conclusions can be drawn from this. First, if you've got Samsung RDRAM that was made in the last year or so, it should easily hit 120mhz+. If not the most likely problem is processor, rdram clock generator, or maybe pci or agp overclocking issues (less likely than processor or rdram clock gen in my experience). I guess a marginal power supply could be a factor. Secondly, if you've got new Samsung dimm's made with 256mbit parts you probably aren't going to have much luck at 133 or higher, unless maybe you boost the voltage to the rams or increase cooling. Those parts should still easily be good to 120+ . If you've got single sided 128mb Samsung dimms made in the past year (8 parts on a side not 4), you have an extremely good chance of getting at least 133mhz. Double sided dimms might make it to 133, expecially parts made in the past 6 months (from what others report).
There have been many posts about rdram clock generators over the past few weeks. If you've got a P4T-E with the ICS9212-13 ram clock generators (like mine) you're in luck, and the board should overclock like crazy with a 1.6a or 1.8a processor and Samsung RDRAM. If you've got a P4T-E with the Cypress clock generators (appears to be newer P4T-E motherboards or the old P4T's) you're probably out of luck hitting 133+. I understand some P4T-E motherboards have TI clock generators which should be good for 133mhz. They aren't rated for PC1200 like the ICS9212-13.
The ABit TH7II's with ICS9212-03 are only rated for PC800 use, but I bet the ICS9212-03 clock gen is the same as the ICS9212-13, probably either just not tested for the higher speed operation, or they didn't quite meet spec at the higher speed. My guess would be that they should have a good chance at 133mhz. I just got a new TH7II, but haven't tested it yet.
Bottom line, poor results overclocking Samsung rdram are most likely limitations of the rdram clock generators. The 1.6a and 1.8a processors combined with a good motherboard finally enable us to push P4 processors to 133mhz fsb and beyond and also enable us to push rdram to PC1066 and beyond without waiting until the official may release.
Facebook
Twitter