Show us your SSD performance 2

paulpicks21 said:

I was just testing this PrimoCache out, it certainly gives some impressive numbers.

With the recent discussion on the PrimoCache, I decided to bench my Ramdisk out of curiosity. I was a bit surprised with the results compared to what I remember from several months ago. I don’t recall the numbers being so disproportionate to each other.

If anyone has any insight as to why the numbers seem strange, please often your opinion. Paul's numbers seem to be more proportional to each other -- what you would expect. So I put Paul’s numbers in there for comparison. See photos below.

P.S. - the AS-SSD people need to change their font size or widen their numeric windows; I noticed both in Paul's and mine that whenever the numbers go over 9,999MB/s, the window truncates the "MB/s" suffix. They likely never expected the numbers to get so high.

Can I ask how much RAM you dedicated to the Cache garuda? And did you alter any other settings apart from the amount of RAM used for Read/Write?

My results vary quite a lot from run to run. Here is a run with 12GB used.

paulpicks21 said:

Can I ask how much RAM you dedicated to the Cache garuda? And did you alter any other settings apart from the amount of RAM used for Read/Write?

I use a RamDisk program (and there are many out there), not PrimoCache. It depends on the project in determining how much RAM I dedicate. For large projects/files in Adobe Premiere or AfterEffects et al, my RamDisk is 50GB; for projects using Prem, AE, Photoshop, etc together, I can go up to 90GB RAM. For small stuff I scratchpad at 10-20GB. It all depends.

When using large files (over 20-40GB total, have cache assigned for preview files, and rendering --- lots of RamDisk is needed. Way more than I initially calculated. I have learned from many people much smarter than me: you can never have too much memory or too many cores. They were right. Enough preaching.

I should mention one other thing here, with this new Primo thing getting attention. A word of caution for those who have installed PrimoCache on their main computer. Romex has reported BSOD bugs with versions prior to v0.9.2 back in November. Version 0.9.2 may have fixed some of those bugs. It may or may not cause data loss when it blue-screens depending on what your level-2 cache settings are. I would not trust this beta version with any critical files, until they get the acknowledged bugs ironed out. Enabling level-2 cache can cause data issues or BSOD in some instances. It still seems to be in Beta and who knows when it will become stable. It may still have known issues that are unresolved.

Some have claimed that the PrimoCache developers have known about past BSOD bugs and yet failed to promptly warn customers about these unresolved issues. I would use caution in what files you address with this software until a stable version arrives. You can check the Romex forum for warnings, news, and updates at this link:
Romex Software Forum ? View forum - PrimoCache Section

garuda said:

paulpicks21 said:

Can I ask how much RAM you dedicated to the Cache garuda? And did you alter any other settings apart from the amount of RAM used for Read/Write?

I use a RamDisk program (and there are many out there), not PrimoCache. It depends on the project in determining how much RAM I dedicate. For large projects/files in Adobe Premiere or AfterEffects et al, my RamDisk is 50GB; for projects using Prem, AE, Photoshop, etc together, I can go up to 90GB RAM. For small stuff I scratchpad at 10-20GB. It all depends.

When using large files (over 20-40GB total, have cache assigned for preview files, and rendering --- lots of RamDisk is needed. Way more than I initially calculated. I have learned from many people much smarter than me: you can never have too much memory or too many cores. They were right. Enough preaching.

I should mention one other thing here, with this new Primo thing getting attention. A word of caution for those who have installed PrimoCache on their main computer. Romex has reported BSOD bugs with versions prior to v0.9.2 back in November. Version 0.9.2 may have fixed some of those bugs. It may or may not cause data loss when it blue-screens depending on what your level-2 cache settings are. I would not trust this beta version with any critical files, until they get the acknowledged bugs ironed out. Enabling level-2 cache can cause data issues or BSOD in some instances. It still seems to be in Beta and who knows when it will become stable. It may still have known issues that are unresolved.

Some have claimed that the PrimoCache developers have known about past BSOD bugs and yet failed to promptly warn customers about these unresolved issues. I would use caution in what files you address with this software until a stable version arrives. You can check the Romex forum for warnings, news, and updates at this link:
Romex Software Forum ? View forum - PrimoCache Section

Thanks for the info.

I should have worded my question better, I was attempting to try and help with your disproportionate numbers.
What I should have asked was how much RAM did you have allocated when you ran the AS SSD test in your screenshot? I understand you are not using Primo, but my findings in Primo indicate the more RAM allocated gives higher scores in AS SSD. I figured if you were using a particularly large amount of RAM then this could explain the disproportionate figures. It's only a theory. Maybe try it with 4GB or 12GB allocated to see if there is any difference.

Paul.

Garuda AS SSD says the alignment id bad for your RAM disk even though 8032K is divisible by 4. The alignment must be divisible evenly by for the structure of the SSD files system and how it addresses data. I don't know exactly how that translates to the scores shown though.

I can't say how it can be changed on a RAM disk. It could also be the way AS SSD reads it since RAM isn't the same as an SSD's NAND memory.

paulpicks21 said:

..... I figured if you were using a particularly large amount of RAM then this could explain the disproportionate figures. It's only a theory. Maybe try it with 4GB or 12GB allocated to see if there is any difference.

Paul, your question was fine, sometimes I don't interpret so good. I ran AS-SSD with both 50GB & 10GB RAM based on your theory earlier posted. But the numbers were very similar both ways: 50 or 10G; it was the disproportionate part that threw me. And thanks for your suggestion in trying different sizes, often that makes a difference as you found in your runs. I'm stumped on this one. I hope Britton has found the answer.

Britton30 said:

Garuda AS SSD says the alignment is bad for your RAM disk even though 8032K is divisible by 4. The alignment must be divisible evenly by for the structure of the SSD files system and how it addresses data. I don't know exactly how that translates to the scores shown though.

I can't say how it can be changed on a RAM disk. It could also be the way AS SSD reads it since RAM isn't the same as an SSD's NAND memory.

Good eye finding that. I have since checked Diskpart and it verifies that the alignment is bad. Since I think alignment relates to handling block sizes, you likely discovered the reason for disproportionate readings -- it handles different block sizes better than others when align is off. And this seems REALLY off: 8032K. Thanks! I'm putting together the documentation to send to RamDisk company to ask for their help on this. And yes, I checked for any settings in the RamDisk setup for alignment adjustment and found nothing. I'll have to see what the RamDisk company says.

I just restored an image to this fresh drive, enabled Rapid Mode. Not as high marks as the 8840EVO but still impressive.

Impressive indeed.

when mine gets to access time it gives me this. anyone know why?

Something has locked access? A Ram drive, some caching software? A Guy