My first encounter with Corsair's RAM came via a posting in the forums here, whereby a reader had gotten ahold of a stick of CAS2 PC133 ram and was showing off the speed he could get on his Asus P3V4X. At the time, I had been working on my Abit KA7 review and had been exploring the phenomenal memory speed of that board - I only had CAS3 ram, though. The combination of this motherboard and some Corsair CAS2 PC133 ram promised to be a spectacular one, so Jim Baker from Realtime kindly sent over a stick for me to check out.

CAS?
Firstly though, you may be wondering what all the CAS2 fuss is really about. You may have some PC133 memory already, or you might even be running some PC100 memory overclocked to 133MHz. Is this Corsair RAM any different? Well, yes.. allow me to digress a little into the world of SDRAM timing. For the most part, when talking about SDRAM timing, you generally speak of two settings. The first is the memory bus speed, expressed in MHz. This is often the same as the CPU's Front-Side-Bus (FSB) speed, but not always - an asynchronous memory bus allows you to run the memory higher, lower, or the same as the FSB. The second setting is the CAS rating, being 2 or 3. There are other settings, such as pre-charge and RAS, but generally people only talk about ram in terms of MHz and CAS. CAS, or Column Access Strobe, refers to a part of the memory access cycle. In simple terms, you can consider it a delay as the SDRAM subsystem addresses the correct part of the memory array. Therefore, a delay of 3 (as in CAS3) makes for slower overall memory access than a delay of 2 (CAS2). So, smaller numbers are better. I guess the next step is CAS1 but that's unlikely to happen anytime soon.

Now, one of the classic tradeoffs when overclocking, especially since the advent of VIA's various chipsets with asynchronous memory busses - but even back in the BX days, is taking a stick of PC100 CAS2 ram and dropping it to CAS3 to allow you to reach higher bus speeds - running your PC100 ram at 133MHz, for example. There is a performance penalty there, due to the higher CAS delay, but if you can raise the MHz speed up high enough you will find an overall gain. In my experience, the break-even point is about 124MHz, and to really make it worthwhile you have to get it to about 133MHz. Not all CAS2 PC100 memory will do 133MHz, even at CAS3, though..

However, a lot of PC133 memory these days is only CAS3 rated. This means the manufacturer guarantees it to run at 133MHz with a CAS delay of 3. This in reality is no better than many PC100 CAS2 sticks, which will run at the same settings. Even at 133MHz, the performance gain is not that fantastic compared to 100MHz CAS2. The reason why the Corsair is so special is that it's guaranteed by the manufacturer to run at 133MHz at CAS2. This is great for overclockers, as it means we still have the option of dropping back to CAS3 if we're pushing the RAM to higher frequencies.

The "222" is what makes it interesting..

Benchmarking
But enough theory, let's see some benchmarks. How much difference does CAS2 really make compared to CAS3? I used the Abit KA7 as the tesbed for this review, with the KA_42 bios and an Athlon 750 overclocked to 900MHz, L2 cache @ 2/5. The FSB was left at 100MHz, but I used the "FSB+PCI" BIOS option to run the ram at 133MHz for those tests. The memory was set to "Turbo" timing with 4-way interleaving on.

To be honest, I was surprised by how little difference going from CAS3 to CAS2 makes. There's definitely a difference, but it a real-world application like Q3 it doesn't translate to much. Of course, the more memory-intensive the task you perform, the more noticeable the speed increase. Something like SETI@HOME, where FSB and memory speed are absolutely vital, would produce a more pronounced difference. However, from that graph the cost difference seems unjustified.

Furthermore, I did some speed comparisons with a 128MB stick of KingMax PC133 CAS3 ram I have here. The speeds at each setting were so close to the Corsair that it wasn't worth putting a graph together to demonstrate them. Of course, the KingMax couldn't handle CAS2 @ 133 as the Corsair can, but we can see that's not such a big difference in speed anyway. So why the big deal..?