Bleah :( 648 chipset doesn't oc well, see review links...

[lookin4dlz]

ABIT SR7-8X Review

Try as we might, we found it impossible to push the FSB of this board over 143MHz with any kind of stability.

ASUS P4S8X SIS 648 Review

At the unofficial DDR400, we are unable to complete most of the tests and it is very unstable even with the Corsair DDR400 CAS 2.5 module set to run in SPD mode. Overclocking wise, we managed to overclock up to 150Mhz FSB using our P4 2.4B processor. If you are into overclocking, the FSB is only selectable up to 166Mhz FSB.

 

[Sunny129]

there are only 3 SiS 648 boards on the market right now... the ABit SR7-8X, ASUS P4S8X, and Shuttle AS45-GTR. this means that there are many more SiS 648 mobos to come, and just as it is with any new product, it will improve over time. it is without a question that this particular P4 DDR solution is the closest to P4 RDRAM solutions in performance using PC2700 memory, and will surpass the the performance levels of P4 RDRAM solutions once stable with PC3200 memory. so not only do we have yet to see a majority of SiS 648 boards on the market yet, but with BIOS revisions on the ones on the market thus far, we are sure to see increases in FSB overclockability, and maybe even some new CPU : DRAM ratio dividers or even an AGP/PCI bus lock (which the ABit SR7-8X already has, but is pretty much useless with a board that OCs like a broken down school bus). revisions will also be made to the chipset as time goes by, and in no time we will be seeing the SiS 648 run DDR400 flawlessly. notice a P4 2.4B was used to OC in the P4S8X review...how much higher would this CPU be able to go considering it already defaults on a 133MHz fsb? there's always more headroom to OC when using a CPU that runs default on a 100MHz fsb (as opposed to 133MHz fsb of the CPU used to OC in the review). perhaps the fsb shouldn't be the sole factor being studied in the OC, but rather the actual CPU freq. the P4 2.4B running on a 150MHz fsb will be clocked @ 2.707GHz. so the fsb only sees a mximum increase of 17MHz. but lets say the test was done with a P4 1.6A (default 100MHz fsb). @ 150MHz fsb, the CPU will be @ 2.4GHz. assuming the CPU is capable of OCing to 166Mz fsb (the max allowed by the P4S8X used in the test), the 1.6A would be @ 2.656GHz. not only does the 1.6A undoubtedly OC better than the 2.4B, but the fsb OC is greater. the fact that Ocworkbench only got up to 150MHz fsb is due to the CPU they used in the test. the CPU was the limitation. not the chipset or memory (by the way i'm sure they used the more stable DDR333 for the Oc test).

anyways, i'm giving this chipset some time to mature...its the most promising one out so far in my eyes...

 

[lookin4dlz]

Yeah, I completely agree that this chipset will get better with the next stepping (or whatever they call revisions for chipsets) and maybe a revised bios will help in the interim. SIS's success with the 645 series had me hoping that the 648 would be more mature from the start. The thing is, dual channel DDR chipsets are coming out very soon (I think I read that Intel had moved their release to Sept??) and if boards with DC DDR work well right off the bat, it may not be worth waiting for the 648 to get better...

 

[DevilsAdvocate]

Where did you read about a possible Sept. release for the Intel dual-channel chipsets?

 

[Sunny129]

what exactly is dual channel DDR? is that kind of like RAMBUS's 32bit vs 16bit RDRAM? and will such boards/chipsets have the illustrious feature set provided by the SiS 648?

 

[Mamoose]

Originally posted by: lookin4dlz
Yeah, I completely agree that this chipset will get better with the next stepping (or whatever they call revisions for chipsets) and maybe a revised bios will help in the interim. SIS's success with the 645 series had me hoping that the 648 would be more mature from the start. The thing is, dual channel DDR chipsets are coming out very soon (I think I read that Intel had moved their release to Sept??) and if boards with DC DDR work well right off the bat, it may not be worth waiting for the 648 to get better...

"Build it and they will"............B obsolete tomorrow

 

[lookin4dlz]

Originally posted by: DevilsAdvocate
Where did you read about a possible Sept. release for the Intel dual-channel chipsets?

Devils, lol I read 2-3 hours of technology online/magazine news each night, so it's hard to remember where I read that. I do remember that it was a recent (last couple of days) read. If I remember correctly, Intel was only going to certify DCDDR at PC2100 because that provides 4.2GB/s of bandwidth, which is what the 533MHz bus of the latest P4 chips use. If I see the reference again, I'll post!

Sunny non-technically speaking (and despite my reading habits, I don't consider myself to be a techie) it's a lot faster type of ram, maxing out I think at 6.4GB/s. Since P4's thrive on memory bandwidth it will benefit that chip a lot more than other chips.

 

[mike9390]

Dual Channel DDR means you have to use two sticks of ram. With a single stick of single channel PC2100(DDR266) you get a maximum bandwidth of 2,100 MB/s. I'm not sure if it's 2,100 megabytes or 2,100 megabits per second. Whatever it is with dual channel DDR when you use both sticks you get a maximum of 4,200 MB/s. With PC2700(DDR333), one stick of single channel is a max of 2,700 MB/s and dual channel is a max of 5,400 MB/s.

The reason Rambus is faster than DDR in a Pentium 4 system is because of the bandwidth that dual channel RDRam offers. With dual channel DDR you get the benefit of the high bandwidth and low latancy that DDR offers. So in theory dual cahnnel DDR should be at least as good as PC1066 RDRam and should in fact should be better.

Mike