Solidigm NVMe Custom Modded Driver for All NVMe Brands SSDs & Any NVMe SSDs

[IMABEARLOL]

@TheLostSwede FYI, likely you are experiencing the firmware bug that SKHynix/Solidigm refuses to address.

I encountered this issue on multiple drives from them.

I believe the issue comes down to the firmware bugging out and failing to do proper block tracking and garbage collection and so trim is actually never happening. People believe it's controller defect which is why no firmware has been released.

If you initialize the disk again performance will return until it runs out of "fresh" flash blocks again. Obviously, benchmarking speeds up this process.

 

[davidaz]

WD SN750 BLACK 500 gb

kingston kc3000 2tb

stock vs Solidigm NVMe Custom Modded driver

 

[TheLostSwede]

@TheLostSwede FYI, likely you are experiencing the firmware bug that SKHynix/Solidigm refuses to address.

I encountered this issue on multiple drives from them.

I believe the issue comes down to the firmware bugging out and failing to do proper block tracking and garbage collection and so trim is actually never happening. People believe it's controller defect which is why no firmware has been released.

If you initialize the disk again performance will return until it runs out of "fresh" flash blocks again. Obviously, benchmarking speeds up this process.

Sadly, running a manual TRIM made no difference.

 

[IMABEARLOL]

Sadly, running a manual TRIM made no difference.

View attachment 376845

This is destructive to all the data, but 'clean' the disk using DISKPART. Then use windows partition manager to initialize and create a new partition and trim it. You should see performance return. But again, this is only until the drive runs out of fresh blocks and the bug returns.

 

[TheLostSwede]

This is destructive to all the data, but 'clean' the disk using DISKPART. Then use windows partition manager to initialize and create a new partition and trim it. You should see performance return. But again, this is only until the drive runs out of fresh blocks and the bug returns.

That's my OS drive, so I won't be doing that.

 

[THEBOSS619]

That's my OS drive, so I won't be doing that.

when you return to MS-Stock In-Box drivers, you regain all the performance back right?

 

[TheLostSwede]

when you return to MS-Stock In-Box drivers, you regain all the performance back right?

We discussed this, the answer is still no, as you can see above.
You suggested a system restore, but I will re-install the OS at some point, but not right now.

 

[THEBOSS619]

We discussed this, the answer is still no, as you can see above.
You suggested a system restore, but I will re-install the OS at some point, but not right now.

Oh indeed indeed, I totally forgot that we discussed this previously. Forgive me for that, Sometimes my brain gets blank from time to time

 

[FreedomEclipse]

If we are using the old custom driver. Would it be worth using the new one?

 

[DrDre]

I think you have to try it, it's not complicated and it takes minutes.
Both are running good, but the latest has more features.

 

[THEBOSS619]

If we are using the old custom driver. Would it be worth using the new one?

as @DrDre mentioned , test is the only way to see if it's worth it or not also once I get back from the break... I have so many ideas of optimization for Solidigm drivers it just takes time due to stability testing & my very limited spare time

 

[aliceS]

I've been scratching my head lately with this ASPT thing. What does it exactly do? Shouldn't controllers already engage and desingage high peformance or idle states by itselves? Does it makes the states switch between quickly by removing something from those calls? Or, perhaps, controllers don't really sleep like CPUs do?

 

[davidaz]

3DMARK DIRECT STORAGE BENCHMARK TEST

 

[DrDre]

I've been scratching my head lately with this ASPT thing. What does it exactly do? Shouldn't controllers already engage and desingage high peformance or idle states by itselves? Does it makes the states switch between quickly by removing something from those calls? Or, perhaps, controllers don't really sleep like CPUs do?

The Importance of Efficient SSD Power Management - Phison Blog

The SSD has been gradually maturing in the storage world, and users have begun replacing HDDs with SSDs because of their higher performance, lower weight,

phisonblog.com

Technology Power Features - NVM Express

NVMe™ Technology Power Management Features NVM Express (NVMe) architecture contains various features to manage the power of SSDs. In the data center, this can be used to throttle an SSD to a specific TDP (thermal design power) to manage platform thermals and total power draw. Max power is...

nvmexpress.org

NVMe

Outlines best practices for power management for NVMe storage devices

learn.microsoft.com

Maybe this helps

 

[THEBOSS619]

I've been scratching my head lately with this ASPT thing. What does it exactly do? Shouldn't controllers already engage and desingage high peformance or idle states by itselves? Does it makes the states switch between quickly by removing something from those calls? Or, perhaps, controllers don't really sleep like CPUs do?

To add to @DrDre statement and links, I will try to simplify my explanation as much as I can...

True, controllers have the ability to engage and disengage high performance or idle states by itself but in Windows (and maybe in Linux too) provides you the ability to customize it about when does it do that rather than letting the controllers doing it on there own, By default Windows controls those states and it's better that way so that Windows can manage SSD latencies and tries to keep in sync with controllers states.

For example.. whenever controllers tries to come out of idle states... as these process takes time ... it will usually cause high latencies for the operation of concurrent read/write requests happening in real-time and hence Windows will try to throttle some requests giving controllers chance rather than being choked to oblivion (or else it will be forced to reset after 60s timeout of no response) + SSDs aren't effective when they are operating near throttling temperature point so ASPT helps in that as well.

It's a whole complexities happening in the background regarding ASPT

 

[LLLLLL12700]

PM981 works fine, and I also use it for placing 3D render assets

 

[aliceS]

The Importance of Efficient SSD Power Management - Phison Blog

The SSD has been gradually maturing in the storage world, and users have begun replacing HDDs with SSDs because of their higher performance, lower weight,

phisonblog.com

Technology Power Features - NVM Express

NVMe™ Technology Power Management Features NVM Express (NVMe) architecture contains various features to manage the power of SSDs. In the data center, this can be used to throttle an SSD to a specific TDP (thermal design power) to manage platform thermals and total power draw. Max power is...

nvmexpress.org

NVMe

Outlines best practices for power management for NVMe storage devices

learn.microsoft.com

Maybe this helps

To add to @DrDre statement and links, I will try to simplify my explanation as much as I can...

True, controllers have the ability to engage and disengage high performance or idle states by itself but in Windows (and maybe in Linux too) provides you the ability to customize it about when does it do that rather than letting the controllers doing it on there own, By default Windows controls those states and it's better that way so that Windows can manage SSD latencies and tries to keep in sync with controllers states.

For example.. whenever controllers tries to come out of idle states... as these process takes time ... it will usually cause high latencies for the operation of concurrent read/write requests happening in real-time and hence Windows will try to throttle some requests giving controllers chance rather than being choked to oblivion (or else it will be forced to reset after 60s timeout of no response) + SSDs aren't effective when they are operating near throttling temperature point so ASPT helps in that as well.

It's a whole complexities happening in the background regarding ASPT

Aah, I understand it better know. So it is somewhat of a way to reduce latency between states, like power modes in Windows 10/11.

Makes much more sense. Thank you very much for the explanation and the articles!

 

[FreedomEclipse]

Unfortunately no before and after screenshots but I can confirm that the OLD DRIVER works with the Phison PS5021-E21 controller.

Unfortunately this laptop only goes up to Gen 3 as its an old HP 830 G5.

 

[MyKillK]

My Interrupt Policy tool reports that the NVME storage driver has a max MSI limit of 130, although it does show the assigned MSI limit at 2,048 as per the driver settings. But I'm wondering if this is an issue, trying to force more MSI messages than it can actually handle? I think while the PCI Express limit might be 2,048, the devices themselves on the end of PCIe bus can have different limits?

 

[sause]

Intel DC P5800X 400GB on an overclocked 9800X3D system connected to the primary m.2 slot.

Going thru the X870E chipset/slimsas incurs 7-8% penalty in 4KQ1 for anyone curious.

This is not typical result. It was ran in Windows 11 safe mode just to see how high I can get it to go. Normally would see around 410-418MB/s in normal Windows 11 operation.

Your driver consistently scores higher in this test than the default windows driver. Just wanted to say thanks!

 

[THEBOSS619]

My Interrupt Policy tool reports that the NVME storage driver has a max MSI limit of 130, although it does show the assigned MSI limit at 2,048 as per the driver settings. But I'm wondering if this is an issue, trying to force more MSI messages than it can actually handle? I think while the PCI Express limit might be 2,048, the devices themselves on the end of PCIe bus can have different limits?

it's not an issue and it's completely normal. You can't force HW to use more MSI bec. it depends on the controllers design and how the OEM/Vendors decided how much they use MSI. The only control you have is reducing or limiting the MSI usage but not going over the limit.

Some HW like for example "Intel Wi-Fi" can go over the limit of MSI but it won't function and will fail to load the drivers