Converting to uefi

[Tomizz55x]

So I want to covert bios to uefi and I can't because I have a 2 recovery partitions

Is this safe?

https://andreagx.blogspot.com/2017/11/mbr2gpt-validatelayout-too-many-mbr.html

[rcmaehl]

Looks about right. Ignoring the second recovery partition, here's how it'd be done normally: 
 

https://www.thewindowsclub.com/how-to-change-legacy-to-uefi-without-reinstalling-windows-10

[Tomizz55x]

4 minutes ago, rcmaehl said:

Looks about right. Ignoring the second recovery partition, here's how it'd be done normally: 
 

https://www.thewindowsclub.com/how-to-change-legacy-to-uefi-without-reinstalling-windows-10

When I try to do that it says:
MBR2GPT: Attempting to convert disk 0
MBR2GPT: Retrieving layout of disk
MBR2GPT: Validating layout, disk sector size is: 512 bytes
Disk layout validation failed for disk 0
MBR2GPT: Conversion failed

[Kisai]

4 hours ago, Tomizz55x said:

When I try to do that it says:
MBR2GPT: Attempting to convert disk 0
MBR2GPT: Retrieving layout of disk
MBR2GPT: Validating layout, disk sector size is: 512 bytes
Disk layout validation failed for disk 0
MBR2GPT: Conversion failed

You might be boned then. If it's a system manufacturer recovery partition (eg HP) you can just delete it and then reinstall it afterwords. Otherwise I'd suggest reinstalling the OS.

 

The MBR2GPT is only for converting from non-GPT to GPT (UEFI) on the hard drive, and I've done it a dozen times or so with machines updated from Win7 at the office. However there are tradeoffs that happen that doesn't happen with a reinstall, one of them being "extra drive partitions" that show up in explorer that shouldn't.

 

In general, it's better to reinstall the OS, and I'd probably take the opportunity to reinstall the OS to a larger NVMe SSD if you have that option over reinstalling to the same drive. If you're not going to do this, then there's not really much point in switching to UEFI anyway. 

 

There's only a few minor reasons to switch to UEFI, one of them is "UEFI-only video card", but you also gain some boot speed, which only becomes apparent on a SSD (eg a "5 second boot") and otherwise once the OS is loaded, it doesn't matter what mode the BIOS is in.

 

There IS however performance to be gained from switching from AHCI to RAID (intel RST) mode if you are using a NVMe drive.

[StDragon]

3 hours ago, Kisai said:

 

There IS however performance to be gained from switching from AHCI to RAID (intel RST) mode if you are using a NVMe drive.

NVMe supersedes AHCI. In fact I don't think they have anything to do with each other other then that AHCI enhances SATA via NCQ. That said however, it's always advisable to use AHCI over RST when using a SATA based SSD drive, regardless if it's plugged in to an actual SATA port or M.2 slot (because M.2 typically supports SATA and NVMe).

Basically, if you've got an NVMe drive, I don't think having AHCI or RST enabled will make a difference either way. If I'm wrong, please point me to the proper documentation including whitepapers if available.

[Kisai]

21 hours ago, StDragon said:

NVMe supersedes AHCI. In fact I don't think they have anything to do with each other other then that AHCI enhances SATA via NCQ. That said however, it's always advisable to use AHCI over RST when using a SATA based SSD drive, regardless if it's plugged in to an actual SATA port or M.2 slot (because M.2 typically supports SATA and NVMe).

Basically, if you've got an NVMe drive, I don't think having AHCI or RST enabled will make a difference either way. If I'm wrong, please point me to the proper documentation including whitepapers if available.

The performance is capped to SATA speeds when set to AHCI due to AHCI overhead. That is why the RST driver requires the BIOS being set to RAID. On Win7 machines, the WinSAT will show a significant jump in RST mode, owing to the fact that NVMe drives operate at PCIe speeds, not SATA speeds. We're not talking about 100% jumps, just increases in IOPS.

 

On Win10, Microsoft has hidden the WinSAT UI stuff, but it's still there:

 

open a command line:

Quote

 

winsat formal

 

go to C:\Windows\Performance\WinSAT\DataStore

 

You'll see stuff like this:

Quote

<WinSPR>
        <SystemScore>8.15</SystemScore>
        <MemoryScore>9.1</MemoryScore>
        <CpuScore>9.1</CpuScore>
        <CPUSubAggScore>8.3</CPUSubAggScore>
        <VideoEncodeScore>9.9</VideoEncodeScore>
        <GraphicsScore>9.1</GraphicsScore>
        <Dx9SubScore>9.9</Dx9SubScore>
        <Dx10SubScore>9.9</Dx10SubScore>
        <GamingScore>9.9</GamingScore>
        <DiskScore>8.15</DiskScore>

and

Quote

        <DiskMetrics>
            <AvgThroughput kind="Sequential Read" units="MB/s" ioSize="65536" score="8.1">508.24687</AvgThroughput>
            <AvgThroughput kind="Random Read" units="MB/s" ioSize="16384" score="8.2">394.52000</AvgThroughput>
        </DiskMetrics>

 

Anyway, People Have Done This, and RST always comes out ahead.

 

[StDragon]

Yeah, that's somewhat interesting. Perhaps RST is doing additional read/write caching or somehow is handling DMA and IRQs more efficiently? Because there's nothing inherent with RST other than being software driven "fake raid". And that's OK, for only a few drives in RAID the AND/OR bits to calculate parity is such low overhead on a modern CPU as to not need an ASIC anyways (dedicated RAID controller hardware). I'll have to do some research as to why that person was getting an large discrepancy in improvement.

 

Speaking of... He should have been 3.2GB/s reads and 2.4GB/s on writes per the specifications of a 1TB Samsung PM981 drive. Not even approaching that in his bench-marking.

 

https://www.samsung.com/semiconductor/ssd/client-ssd/MZVLB1T0HALR/

 

 

[Kisai]

4 hours ago, StDragon said:

Yeah, that's somewhat interesting. Perhaps RST is doing additional read/write caching or somehow is handling DMA and IRQs more efficiently? Because there's nothing inherent with RST other than being software driven "fake raid". And that's OK, for only a few drives in RAID the AND/OR bits to calculate parity is such low overhead on a modern CPU as to not need an ASIC anyways (dedicated RAID controller hardware). I'll have to do some research as to why that person was getting an large discrepancy in improvement.

 

Speaking of... He should have been 3.2GB/s reads and 2.4GB/s on writes per the specifications of a 1TB Samsung PM981 drive. Not even approaching that in his bench-marking.

 

https://www.samsung.com/semiconductor/ssd/client-ssd/MZVLB1T0HALR/

 

 

It's ultimately capped to PCIe speeds, So a 4 lane PCIe SSD is 985MB/sec x 4 lanes on PCIe3 maximum (3940MB/sec.) Where as SATA will always be capped to 550MB, if it's connected as SATA. AHCI over PCIe NVMe, is emulating AHCI. I don't know how controlled of a setting that user had.

 

https://www.atpinc.com/blog/nvme-vs-sata-ssd-pcie-interface

 

Which suggests there's a 25% performance penalty in AHCI. But it's also a YMMV.

 

At my office, we have those PM981 drives as well as Toshiba THNSN5512GPUK, KXG50ZNV512G, KXG60ZNV512G, and some others (I'm not there right now to check my drawer of scavenged parts.) The're all from Dell laptops. 

 

When switching some legacy Win7 machines to Win10, I, as a matter of making sure they're operating at maximum performance, switch them to UEFI, and put the drive into RAID mode if they're not already. So you do notice some speed bumps, but actual file copy performance? Usually, you won't see 3GB/sec, ever. SSD's really really hate small files. Copy multi-GB Outlook PST's around though? Pretty fast.

 

If you look in the reddit thread, there is some suggestion that the RST driver might be adding more caching to it, which would make sense if it was used as a RAID with mechanical drives, but with SSD's, you usually burn all the OS disk cache memory , and then suddenly see the SSD tank it's performance as it has to write it all out. I've seen this personally a few times on Windows and Linux, and it just kinda tells you the big weakness with benchmarking tools.

 

Unless you're going to write twice the system memory's worth of data to the drive, you'll never overcome the drive cache.

[StDragon]

1 hour ago, Kisai said:

I AHCI over PCIe NVMe, is emulating AHCI. I don't know how controlled of a setting that user had.

 

See, that's what I'm not understanding here. PCIe NVMe drives don't talk AHCI. If they do, there must be some emulation going on that's eating into CPU cycles. Again, to my knowledge no such animal exists in the world of SSDs.

Here, read pages 6 - 12 on "A Comparison of NVMe and AHCI" direct from sata-io.org

https://sata-io.org/system/files/member-downloads/NVMe and AHCI_ _long_.pdf

 

Those benchmarks don't make any sense; yet they point out a factual night/day difference. Mind boggling! That shouldn't be possible!

I'm going to guess that it's not that RST is enhancing performance insomuch as not being the bottleneck that AHCI emulation it causing. Effectively, enabling RST is really tantamount to "getting out of the way" and let NVMe do its thing unobstructed.

[Kisai]

3 hours ago, StDragon said:

 

See, that's what I'm not understanding here. PCIe NVMe drives don't talk AHCI. If they do, there must be some emulation going on that's eating into CPU cycles. Again, to my knowledge no such animal exists in the world of SSDs.

Here, read pages 6 - 12 on "A Comparison of NVMe and AHCI" direct from sata-io.org

https://sata-io.org/system/files/member-downloads/NVMe and AHCI_ _long_.pdf

 

Those benchmarks don't make any sense; yet they point out a factual night/day difference. Mind boggling! That shouldn't be possible!

I'm going to guess that it's not that RST is enhancing performance insomuch as not being the bottleneck that AHCI emulation it causing. Effectively, enabling RST is really tantamount to "getting out of the way" and let NVMe do its thing unobstructed.

That document is from 2012, when no consumer laptops or desktops supported NVMe. Like before people had SSD's large enough to matter (256GB and larger) and when nobody had M2 NVMe drives.

 

If you read it though, it pretty much spells out why AHCI is inefficient for SSD's. AHCI is literately in the way. 

 

When you use the RST driver with the BIOS set to RAID, it's bypassing the AHCI layer in the BIOS entirely. This is why you can't selectively turn it off per drive either. AHCI is usually implemented in PCH hardware on Intel chipsets for all I know, where as PCIe RAID mode might reallocate those PCIe lanes for the SATA port to the NVMe drive instead.

 

Which could explain the performance numbers if the user's motherboard had all the PCH devices enabled.

[StDragon]

5 hours ago, Kisai said:

where as PCIe RAID mode might reallocate those PCIe lanes for the SATA port to the NVMe drive instead

That's possible. Though I've always known that either you can use a M.2 slot or a certain SATA port, but not both. But I suppose that depends on the MB.