Intel Z68 SSD Disk Caching Showdown

[Psi*]

@Optix .... nicely written review & I am still studying. I now have new hardware installed on 3 different machines with SSDs. I like that you did use CrystalDiskMark as it is simple to use & read. I also ran the other bench marks, but I have to admit that they required more commitment to testing than I was willing to invest, for the time being.

@Kougar ... to answer 1 of your questions wrt benchmarks; I started with all that have been mentioned/used here are used at techgage, but I think I'll keep with CrystalDiskMark for the routine sanity check. And my simple mind, I'll drop back to the more through ones given the right motivation.

Also interesting is your thought that the SSD performance differences between the 920 v 990x box is related to the OC. I thought that the PCIe bus speed had it's own bios setting independent of the rest of the tweaks. The machines are working now & won't be finished until this eve so I can't check that out or possibly experiment.

It took all day yesterday to get an Adaptec RAID 6405 controller installed with fresh W7 onto a 1TB HDD & SSD. W7 refused to install once the partitions (1 per drive) were identified. Finally forcing myself to dig into W7 setup error files I saw that setup was trying to install on the USB flash drive I was using to provide the latest Adaptec driver. The solution was to pull the flash drive after the driver was loaded ... then everything went smoothly.

Because of that frustration, I nixed other experiments with the Adaptec. I installed the HDD & the SSD as simple drives, aka independent RAIDs. Initially I set the drives up as a striped RAID. The controller automatically partitioned an identically sized portion of the HDD to match the SSD (60 GB). That would have been a nice experiment as well as some variations, but put to the future for now.

For the way the system is setup now and the intent is to have the SSD as scratch space only for my main number crunchers. In that respect simply, "YAAAAAY", not to be overly scientific or anything.

More later

 

[Kougar]

@Kougar ... to answer 1 of your questions wrt benchmarks; I started with all that have been mentioned/used here are used at techgage, but I think I'll keep with CrystalDiskMark for the routine sanity check. And my simple mind, I'll drop back to the more through ones given the right motivation.

CDM is more than fine, I was just curious to check.

Also interesting is your thought that the SSD performance differences between the 920 v 990x box is related to the OC. I thought that the PCIe bus speed had it's own bios setting independent of the rest of the tweaks. The machines are working now & won't be finished until this eve so I can't check that out or possibly experiment.

Keep in mind how the chipset diagram is layed out. Everything in the southbridge AND the northbridge talks to the CPU through the QPI bus... and the QPI bus is affected when you change the bus multiplier. So it definitely has the potential to affect everything. There should be plenty of bandwidth though, but depending on BIOS settings it's possible it could at least contribute to the difference ya were getting.

Glad ya managed to sort out the other issues ya were having!

 

[Psi*]

.
.
.
Glad ya managed to sort out the other issues ya were having!

Eeeeh .. I didn't say that.

I used a utility that set everything automagically, at least those tweaks are the same on each machine. And, I am really liking the Adaptec controllers. but at $360 each causes one to hesitate quite a bit. At the moment, the box with dual Opteron 290s has the Adaptec controller and substantially faster than the 920 or the 990x boxes. I hope to start a separate thread this weekend posting this.

But the 920 & 990x still have their differences and are still quite consistent. Those CPUs are WC-ed so the idea of swapping the chips doesn't really appeal to me ... just sayin'. Maybe switching things back to stock settings would be as useful? bios does save settings so switching around bios setting is no big deal.