07-25-2015 06:30 PM
I have a Rampage V Extreme motherboard with the latest chipset drivers. I also have installed the Intel 750 NVMe drivers. I cannot enable write caching in Windows 8.1 Pro. I think I disabled it while testing my PC overclocks to ensure I wouldn't get file system errors if my PC crashed but now that i'm stable I cannot enable write caching in Device Manager.
Can anyone help?
KedarWolf
12-25-2015 11:52 PM
I've played a little bit with this. Wrote a small code to test speed. It writes a 3GB binary file.
static void Main(string[] args)
{ if (args.Length != 1) { return; }var bytes = Guid.NewGuid().ToByteArray();
var watch = new Stopwatch(); using (var stream = new FileStream(args[0], FileMode.CreateNew)) { watch.Start(); for (long i = 0; i < 200000000; i++) { stream.Write(bytes, 0, bytes.Length); } watch.Stop();}
Console.WriteLine(watch.Elapsed);
}I intentionally wrote my own too see perf at high level as visible to apps. I also wanted to measure writing before closing stream. Closing stream could cause flushing cache and I wanted to measure without flushing cache.
I used old SATA SSD OSZ Vertex 2 60GB. It allows enabling cache on it. I tested with cache enabled and disabled and there was not difference in performance. Windows caching does not seem to make a difference in case of SSD.
But I also discovered that that extremely old SATA SSD (5+ years old) was 10% faster than Intel 750 PCIe drive. How come? Intel 750 PCIe is advertised to be 4 times faster than the fastest SATA SSD. Any suggestions how to get the advertised performance?
12-26-2015 02:28 AM
Vit. wrote:
I used old SATA SSD OSZ Vertex 2 60GB. It allows enabling cache on it. I tested with cache enabled and disabled and there was not difference in performance. Windows caching does not seem to make a difference in case of SSD.
That is simply not "Windows caching" that you have disabled, but write cache inside the SSD: https://gist.github.com/tomty89/ad455e0d65484973e52e gist:ad455e0d65484973e52e · GitHub
And it's natural that disabling it does not create performance impact on your/my benchmark, coz it's done on top of the filesystem.
12-26-2015 10:42 PM
Tom,
That does seem to be the case.
My question now is how do I get the promised 4x performance improvement out of Intel 750 PCIe drive? Right now it underperforms a 2 generations older SATA SSD.
12-26-2015 11:53 PM
How exactly is the speed anyway? Can you use some PROPER benchmark software (e.g. Crystal Disk Mark) to test it and paste the result?
12-27-2015 12:14 AM
Below are results from CrystalDiskMark. Some observations. Sequential read/write is above advertised 2500 MBps/1200 MBps. But random read/write is well below advertised 460,000 IOPS/290,000 IOPS. I need to collect results for SATA SSD using the same tool to compare.
-----------------------------------------------------------------------
CrystalDiskMark 5.1.0 x64 (C) 2007-2015 hiyohiyo Crystal Dew World : http://crystalmark.info/ http://crystalmark.info/-----------------------------------------------------------------------* MB/s = 1,000,000 bytes/s [SATA/600 = 600,000,000 bytes/s]* KB = 1000 bytes, KiB = 1024 bytesSequential Read (Q= 32,T= 1) : 2650.317 MB/s
Sequential Write (Q= 32,T= 1) : 1300.458 MB/s Random Read 4KiB (Q= 32,T= 1) : 611.138 MB/s [149203.6 IOPS] Random Write 4KiB (Q= 32,T= 1) : 526.938 MB/s [128647.0 IOPS] Sequential Read (T= 1) : 1590.101 MB/s Sequential Write (T= 1) : 1273.622 MB/s Random Read 4KiB (Q= 1,T= 1) : 43.865 MB/s [ 10709.2 IOPS] Random Write 4KiB (Q= 1,T= 1) : 285.223 MB/s [ 69634.5 IOPS]Test : 1024 MiB [C: 23.5% (262.6/1117.3 GiB)] (x5) [Interval=5 sec]
Date : 2015/12/26 23:06:24 OS : Windows 10 Professional [10.0 Build 10586] (x64)