I have two identical drives but one is much slower?

[Dc469]

Backstory: I have two identical 1tb xpg gammix s70 blade drives. One has an Ubuntu install and the other windows 10. I have a very slow windows 10 boot. But once in the OS everything is as fast as I would expect. I disabled all my startup processes but that did not solve the issue. I thought, ok let's run spinrite, maybe the boot area is having some issues. It didn't detect anything but I did notice something strange. These drives all show the same sector/head/etc values in spinrite, but the sole difference is the EXTD speed, the Ubuntu drive is 500MB/s and the windows drive is 80MB/s. See photos.

I don't think I have a hardware issue? I'm on an x570 dark hero motherboard. The first m.2. slot is run on the cpu bus and the second is on the chip set, I only have one x16 graphics card and the manual didn't say anything about sharing bandwidth.

Also, I don't know what this EXTD speed in spinrite is, but the difference suggests something about my long windows boot time. Could the drive have been formatted poorly? I don't really want to do a fresh install but I'm not sure what else to try. Thanks for any ideas!

 

[DanR]

@Dc469 I think the issue here is that, for whatever reason, your system is neither loading nor therefore using the proper drivers for the chip set slot during boot-up. Hence the boot process is confined to S-L-O-W BIOS I/O speed until Windows is up and running. Then Windows uses it's own native drivers for accessing this drive at the speed the hardware is capable of (i.e. fast).

Unfortunately I do not have the knowledge for fixing this issue.

I note that BIOS 86 has a transfer of 127, as expected for a BIOS drive. What do you see for BIOS 85?

You might consider swapping the two drives. I would expect the Windows drive would now boot fast in the CPU slot while the Ubuntu drive now boots slow in the chip set slot.

 

[Dc469]

Oh I thought I had uploaded two screenshots in the original post. Here is the other one. Note that the EXTD speed is different for some reason.

This in in spinrite, so the freedos driver should load the same thing for both drives. I checked in windows anyway and the driver is the same for both.

Edit: apparently the site doesn't like my second screenshot. Here is drive 85

https://imgur.com/a/rsYODuN

 

[DanR]

Oops! I do not see the second screen shot.

 

[Dc469]

Sorry I edited the post, it won't upload for some reason so I put it on imgur.

 

[Dc469]

Basically, I'm wondering what the EXTD speed means in spinrite. It may know something about the drive cause that seems like a clue to my issue.

 

[DanR]

That makes two of us as I do not know either. My searching has come up blank. This would be a question for @Steve

 

[PHolder]

I assume both of the drives are on the same motherboard? You should check to see if there is documentation about how the PCIe lanes are divided. Alternatively, I guess you could try swapping the drives and see if the speed sticks with the NVMe port or the drive.

 

[peterblaise]

@Dc469 wrote "... two identical 1tb xpg gammix s70 blade drives. One has an Ubuntu install and the other windows 10. I have a very slow windows 10 boot. But once in the OS everything is as fast as I would expect ... in spinrite ... the sole difference is the EXTD speed, the Ubuntu drive is 500MB/s and the windows drive is 80MB/s ..."​

Free ReadSpeed https://www.grc.com/readspeed.htm for DOS can be run from the same SpinRite boot drive, and offers to show responsiveness measurements across 5 zones, add the /1 /2 /3 or /4 command-line options to cause it to log a vast array of smaller zones, and then we can read the resulting logs in a text editor to see if there are slow spots but otherwise full performance, or if there is low performance across the drive.

I also look at a responsiveness graph using free Hiren's Boot PE Windows 10 or 11 https://www.hirensbootcd.org/download/ or https://www.hirensbootcd.org/old-versions/ and make graphs using the free HDD Scan and or HD Tune programs from that boot drive.

If the slow drive shows that it's only slow at the front, then it's 'worn' and may respond to a rewrite.

If the slow drive shows that it's slow across the whole drive, then it may be the socket or drive itself, try swapping the positions of the two drives.

A SpinRite 6.1 Level 3 often returns an SSD to full performance ( SpinRite has adjustable variables depending on how the drive responds to default settings ).

Another terrific way to 'rewrite' is to clone to a new drive, use the new drive, then secure erase the old drive and repurpose it - measure to see if it's performance is OK after secure erase, and then clone back if you'd like.

.

 

[AlanD]

Which is your Primary operating system? I don't mean the one that you use most, but the one that is loaded initially? In other words, does Ubuntu/Grub load first and then chainload Windows or the other way round? It could be something in the boot order which is causing the slow boot time.

 

[Dc469]

@AlanD I don't think that's the issue as spinrite shows a physical speed difference and it's not using any of that boot chain.

For everyone else, I think I figured it out, or something anyway. I did have some other sata drives for storage installed. It looks like there are a limited number of lanes from the cpu and then some lanes from the chipset. I unplugged all the drives and got my EXTD speed on the slow drive from 80mb/s to 136. So progress. I still don't know what the EXTD figure is though.

It does seem however that I may be limited to which USB ports I can use, if they are all on the same controller or something.

 

[peterblaise]

@Dc469 wrote "... I may be limited to which USB ports I can use, if they are all on the same controller or something ..."​

Are you testing a USB-connected drive?

.

 

[Dc469]

No, but the usb hub shares the lanes from the cpu right?

 

[peterblaise]

@Dc469 wrote "... No, but the usb hub shares the lanes from the cpu right? ..."​

​

- - - - -

SpinRite runs from RAM, from memory, testing one drive at a time, one BIOS/IDE/ATA/AHCI drive ( BIOS may include any USB, SCSI, SAS, nVME drive if they are presented to BIOS during boot ).

The only 'conflict' might arise when the SpinRite program also writes to the boot drive to update the configuration file or write a LOG file.

We can turn off writing to the config file with the command line option of ROCFG:

SPINRITE ROCFG​

... and we can also turn off logging inside the program when it is loaded ( there is no SpinRite command line option to control logging ! ).

That might reduce the SpinRite program from any potential 'conflicts' during testing when booting from a USB drive.

Some USB sockets 'compete' with attached USB keyboards or USB mice - we can unplug mice, try a PS2 keyboard if there is such a socket, and move our boot device to any USB socket that is recognized as bootable to reduce conflicts between USB devices.

We can also format and install SpinRite onto an HDD or SSD in a USB socket, then move that HDD or SSD internally to boot the computer, bypassing any USB boot challenge, and load the SpinRite program into RAM from the internally booted HDD or SSD.

We 'load the SpinRite program into RAM' by simply running the SpinRite program, and it loads itself into RAM and runs itself from there.

The SpinRite program can also load into RAM from a CD, which is unwritable by definition.

Others have made a RAM disk, copied the SpinRite program there, and then loaded the SpinRite program into RAM from the RAM disk - very fast.

- - - - -

What are your concerns, what do you think might be happening here?

Can you not swap the drives that you want to test in whatever sockets they are mounted in, or remove both drives, then reinstall only one drive at a time, and test only one drive at a time?

.