Loss of random performance for Samsung 830 SSD moving from desktop to mobile?

[gordo]

I was under the impression that my 256GB Samsung 830 would have little perceivable performance loss moving it from a Sandy Bridge Z68 desktop motherboard to the Sandy Bridge mobile chipset found in my new Lenovo X220 laptop.

I immediately noticed the 830 didn't seem as zippy as it did on my desktop, so I ran AS SSD to see what was up. Sure enough, AS SSD shows single threaded 4k reads and writes (two important factors for real-world performance) significantly slower than the identical drive running in a desktop system just a few hours prior.

Before anyone gets on me regarding the obvious stuff, let me get that out of the way here:

* Win7 64-bit was re-installed from scratch on the laptop.
* The SSD is properly aligned as shown below in AS SSD screens.
* OS Trim is confirmed enabled, but i ran a manual optimization with Samsung SSD Magician anyhow.
* Both desktop and mobile were running Intel RST drivers. I've tried 11.2, 11.5, 11.7 on the laptop as well as generic MSAHCI with little improvement on 4k performance.
* AHCI mode is enabled in mobile BIOS.
* System restore, Superfetch, Defrag, Readyboost all confirmed off on mobile system.
* Device Manager reports chipset is Intel QM67 Express Chipset and IDE controller is Intel Mobile Express Chipset SATA AHCI controller, showing the correct 11.x driver and driver date when navigating to properties.

Any thoughts on why the discrepancy or is Intel 6-series mobile chipset really *that* much worse?

Desktop
http://i.imgur.com/m7c7F.jpg

Lenovo Laptop
http://i.imgur.com/ZgG59.png

[canthearu]

The laptop:

Slower CPU
More agressive power saving.

Which will both reduce AS-SSD random performance, as well as make the system feel a bit sluggish.

Disable power saving on the laptop (temporarily) then rerun the benchmark, you will get a much faster result. for the 4K readings.

[gordo]

Good suggestion. I plugged it in and set the power profile to "High Performance" and then made sure minimum CPU state was set to 100%.

Only a very slight improvement to single threaded 4k perf.:
Any other ideas?

http://i.imgur.com/cI8qi.png

[canthearu]

Run linx on 1 thread on the background. This will forcibly lock the CPU at it's highest performance state, then run the AS-SSD benchmark.

[gordo]

You were right on the money. It pushed 4k reads/writes past where I thought they'd go on this machine -- not quite to where I had them on my desktop, but further than I was able to push them otherwise (see below). Two questions:

1. I'm curious what's actually going on here, since I watched the CPU in CPUz throughout . As far as i can tell Linx is not independently increasing the CPU/bus frequency beyond what it's at when I simply set the CPU to 100% in Windows Power Options. In both cases, it floats between 3.0 and 3.2 GHz (30x and 32x 100MHz bus speed). Linx seems to be triggering something else, but I'm not sure what...

2. At least when its plugged in, I'd like to get this kind of performance, so the obvious question is: Is there any way to get the CPU into this performance state without having Linx running in the background? It seems absurd to deliver full performance to an SSD only when pegging a CPU core.

P.S. I actually just realized looking at that CPUz screenshot that my mobile core i5-2520M only has 2 hyper-threaded cores, versus my desktop i5-2500k which has 4 cores non-HT. I never expected the mobile chip to hit 4.5GHz, since it's not unlocked, but I figured a Core i5-25xx Sandy Bridge CPU would be mostly the same thing whether desktop/mobile. Seems rather misleading when you think about it...

http://i.imgur.com/RFun2.png

[gordo]

Well, I played around a bit longer... Downloaded ThrottleStop and noticed that CPU-Z isn't showing CPU frequency correctly after all. ThrottleStop clearly shows the CPU multiplier throttling down when not in use, including during AS-SSD runs (hovers between 12-1600 MHz), so that's probably why my benchmark results suck. Sure enough, when I get a core maxed with a solid single threaded load, TrottleStop shows 3.0-3.2GHz and AS-SSD results improve.

ThrottleStop doesn't seem to be capable of doing anything to actually keep this SandyBridge CPU core @ 3GHz, so I suspect Lenovo has locked out the necessary config options in the BIOS.

Thing is, if I watch my CPU frequency as I load programs and do stuff on my system, the CPU does hang around 3.0GHz for a bit, so maybe, just maybe, I am actually getting decent random performance in day-to-day work and AS-SSD is just not creating enough of a realistic CPU load to model it.

Another thing I noticed is that sequential performance is barely affected even if I force the CPU down to a lowly 800MHz. I guess this means that making a ton of small 4k reads and writes is much more CPU-intensive than one big transfer. Actually makes sense I suppose...

[canthearu]

Thing is, if I watch my CPU frequency as I load programs and do stuff on my system, the CPU does hang around 3.0GHz for a bit, so maybe, just maybe, I am actually getting decent random performance in day-to-day work and AS-SSD is just not creating enough of a realistic CPU load to model it.


Yeah, your CPU won't be throwing thousands of 4K requests at the drive all at once. 4K just shows the extreme extrapolation of how it works, as windows would be requesting a mix of small and large blocks from your SSD as you load programs.

So the benchmark score isn't really that important outside of being a benchmark that you can compare different drives with on the same system.

You laptop will feel slightly slower (only if you are paying real close attention though) then the desktop simply because the CPU is a little bit slower and you have more power conserving design on the laptop. You shouldn't a huge difference though day to day, unless you are really loading the CPU up fully (where the real quad desktop core will flex itself)