Old PC vs New PC (same drive): Sequential same, random very different

[rseiler]

I'm wondering if it's normal for there to be such a difference in random/4K results on an older PC vs a newer one using the same model drive and the same SATA interface--when there's essentially no difference for sequential.

-Does the faster CPU really count this much?
-Has there been some significant improvement in this part of the chipset pertaining to this specific thing?
-Is it the driver difference (Microsoft vs Intel RST)?
-Because one drive is slightly more filled than the other?

None of these seem plausible, so maybe this isn't normal.

Drive (both systems):
Samsung 850 EVO 250GB SATA. Plugged in a native 6.0 Gb/s port. Samsung Rapid Mode off.

2011 PC:
-Intel Cougar Point Z68 (Intel Sandy Bridge)
-Core i5-2500k / 16GB RAM
-Windows 10 (storahci). Aligned.

2017 PC:
-Intel Sunrise Point C236 (Intel Skylake-S WS)
-Intel Xeon E3-1225 v5 / 8GB RAM
-Windows Server 2016 (iastora). Aligned.

Sorry, I don't have AS SSD for this system.

Thanks

 

[Elixer]

Yeah, most of that can be chalked up to driver differences, and if intel's driver is using cache and chipset optimizations that have been done.

 

[UsandThem]

From a previous thread I posted, I found out the Microsoft default driver is faster than the Intel one. I tried everything else over a year, but I never would have thought the Intel one would be slower.

https://forums.anandtech.com/thread...rus-needed-slow-850-evo-issue-solved.2498047/

 

[rseiler]

Oh man, that thread opens up so many possibilities. What Elixer says makes so much sense given that RST does say "Disk Data Cache: Enabled" on the Manage tab (not something that I can set, it just is, and there's nothing at all about the subject in the Performance tab), and it makes sense that Intel would know a little more about its chipsets than Intel does.

I'm tempted to try the chipset-appropriate RST driver on the old system, just to see, but it's 3.5 years old and doesn't even officially support Win10.

 

[rseiler]

For fun, I decided to try both the latest official Intel RST for Sandy Bridge (12.9) as well as the modded Fernando one 14.8 (15.x sounds unstable). Both were very close to each other, but downhill (with the exception of Magician) from the Microsoft driver, as compared to the results above for the old machine.

That is what UsandThem found in the other thread, too, so it's not a surprise. I guess in the case of the newer hardware...it's because it's newer hardware (chipset efficiencies, whatever), and, perhaps, Intel RST is especially suited for it (I'm not going to be trying the MS one there to find out).

 

[UsandThem]

That is what UsandThem found in the other thread, too, so it's not a surprise. I guess in the case of the newer hardware...it's because it's newer hardware (chipset efficiencies, whatever), and, perhaps, Intel RST is especially suited for it (I'm not going to be trying the MS one there to find out).

I always installed the Intel RST drivers too. However, I think they are only needed if you run your drives in RAID. I just happened to run the benchmark test in Samsung Magician one day, and saw they were abnormaly low compared to what I read in reviews. I have 5 systems with 850 EVOs, so I checked them all, and they all were slow. Since I "downgraded" to the standard Microsoft ACHI driver, they all perform as they should now, and I have had zero issues. I would have never thought of going to that driver, but after a year of trying to figure it out on my own, I'm glad I posted and someone had the answer. I know I will never install Intel RST again. Good luck!