Intel Z68 SSD Disk Caching Showdown

[Rob Williams]

When we tested out Intel's 'Smart Response Technology' last month, we liked what we saw. But at $110 for a 20GB SLC SSD, we wondered if a larger, more cost-effective option could still make the best use of the technology. With that, we're pitting Kingston's SSDNow V+100 64GB drive, at $150, against Intel's, to see if we retain SRT's effectiveness.

After reading through Ryan's in-depth follow-up look at Intel's Smart Response Technology, discuss it here!

[Relayer]

I'm confused. Is this showing the SSD making a difference in game frame rates? I can't read it any other way, but I've never heard of such a thing.

[Optix]

Your guess is as good as mine, especially since the initial RST testing showed...



There should be no difference between the frame rates whether the HDD is accelerated or not so switching SSDs shouldn't make even the slightest difference.

During testing there were several results that simply did not make any sense so I'm not sure if it comes down to buggy software or poor implementation but I know my testing is sound. In any instance where the results didn't line up, the tests were run an additional 5 times after a cold reboot.

I was hoping for a more clear cut list of results but sadly that didn't happen. If anybody knows why these numbers are so screwy, feel free to post!

***EDIT: 1111!

[Kougar]

That's really unusual... a SSD shouldn't boost FPS. It will let the game start much more quickly (so everyone can stop and wait at all the corporate logos) and will make levels load significantly faster, but not FPS. Unless the game is really paging the disk reading/writing, but games are deliberately designed not to do that so they can keep the FPS rates steady.

[Optix]

Exactly. I'm ready to chalk this up to being a young technology with some kinks to work out because I just don't have any explanation for some of the results.

[Relayer]

I wonder if the manufacturers might be able to explain it? I'm pretty sure Kingston would want to remedy this.

[Optix]

I would think so as well but at least one question was answered - what drive will work best. It all depends on the work being done. Lots of small files = Intel/Larger files = Kingston.

GIGABYTE has launched a series of boards with Intel's 20GB 311 SSD built into it so disk caching comes all in one box.

[Relayer]

Do you have benches with the same system without using the SSD cache function just run from a hard drive and/or SSD? I'd be curious what the FPS was in those situations. Is the storage system only affecting the FPS with the cache function? If so, is it being improved with one or the other, or is it being adversely affected by it?

Sorry for so many questions, but I find it extremely curious.

[Optix]

The graph that I posted earlier is from the initial look at disk caching. It shows the FPS when using only the hard drive and then again with both acceleration modes enabled using the Intel 311 SSD. Based on these results, there should be -no- difference in the numbers using the Kingston drive since there is almost no difference.

You can ask Rob, I worked away on this article for over 2 weeks because of these wonky results and I'm not closer to understanding why the tests came out the way they did. This puppy should have been done in a week tops.

[Relayer]

Thanks a lot for taking the time to fill me in. Sounds like the ball's in Kingston's court to see what the problem is.

Hopefully it's not because it doesn't say "Genuine Intel".

It shows the FPS when using only the hard drive

Can you run it with the Kingston SSD only (without SRT)?

[Optix]

I plan to hopefully do that in the near future if time allows but I have three cases sitting beside me with more on the way and five CPU coolers to beat up on. They have to take priority but if I have time to revisit this, I surely will.

The main problem now is that the Z68 board is with Rob pending a review and the Kingston SSD is being used in another system. If the stars line up, I'll give it a crack because those numbers are still bugging me so I don't consider this to be a closed case.

In hindsight, it would have made sense to include those numbers as a way to isolate the root cause of the decrease in performance but at the time I didn't feel it was necessary since previous testing performed by others showed that RST simply couldn't match a stand alone SSD.

[Psi*]

The video card used is PCIe 16 lane. Could it be reducing the available bandwidth or interacting with, the bandwidth available to the SSDs even tho the SSDs are SATA II?

I was just about to post a new thread wondering about a discrepancy that I am experiencing. I have 2 systems nearly identical; ASUS P6X58D Premium e/w 24 GB RAM, one with i7 920, one with an i7 990X. I just installed a SATA III SSD in each, Corsair force 3 60 GB. Both systems are OC-ed to >4.2 GHz & >4.4 GHz, respectively.

The SSD in the 920 system is ~30% slower than the SSD in the 990X. This is after examining all of the tweaks to be found on the net. I can say that with no programs running that there is 0 activity on the SSDs tho.

The 920 system also has a NVIDIA Tesla C2070 GPGPU, the 990X has a Radeon HD 5770. Both are also PCIe 16 lane, but the Tesla would obviously (I think) put a bigger load on the PCIe bus than the Radeon. This difference is the only thing that I can attribute the performance to. Next would be to start moving parts around so see what follows what if anything ... ugh!

This is new territory for me so I just thought I would put it out there for other thoughts.

[Optix]

I don't think that's the case seeing how 36 lanes are reserved for discrete GPUs even if they are not being used. Plus I believe that the SATA III capability is being handled by a separate add on chip.

Maybe Robert should weight in on this one.

[Rob Williams]

Originally Posted by Psi*

The video card used is PCIe 16 lane. Could it be reducing the available bandwidth or interacting with, the bandwidth available to the SSDs even tho the SSDs are SATA II?
...
The SSD in the 920 system is ~30% slower than the SSD in the 990X. This is after examining all of the tweaks to be found on the net. I can say that with no programs running that there is 0 activity on the SSDs tho.

I'd double-check things like AHCI, the driver version, making sure the exact same port is used on each motherboard, checking the properties of the SSD in both machines to see if any of the information doesn't match up (including TRIM being enabled), et cetera.

On-board SATA solutions don't use the PCIe bus, but rather the internal chipset bus (DMI in Z68's case). If the GPU affects SSD performance, something else has to be at play.

I'll ping Robert, he would be the best person to ask.