BIT 6.0 - Architectures

[JohnDorian]

My test package includes the x86 and x64 versions of PassMark. I'm trying to decrease the size of my test package as much as possible and was wondering if there was a version of BIT that works on both systems.

With a lot of software, the x86 version will work on both OSes. Is that the same for BIT?

[David (PassMark)]

The 32bit software will for the most part run on 64bit operating systems, but we don't officially support it.
The problem is that 1) You can't collect as much system information as 64bit device drivers must be used with 64bit systems & 2) You can't test all the RAM when running 32bit.

The days of the floppy disk are over however. Wouldn't it be easier to buy a 4GB flash drive for $5 and not worry about the size.

[JohnDorian]

The 32bit software will for the most part run on 64bit operating systems, but we don't officially support it.
The problem is that 1) You can't collect as much system information as 64bit device drivers must be used with 64bit systems & 2) You can't test all the RAM when running 32bit.

The days of the floppy disk are over however. Wouldn't it be easier to buy a 4GB flash drive for $5 and not worry about the size.

We're moving away from testing via flash drives because they're a large cost over time. At our current volume, each tester requires 20 USBs, which we purchase for 5.99 USD a piece. As we get new testers, or USB flash drives break, get lost or get stolen, we need to purchase new ones. We currently have about 20 testers in our US location and we have more in our Canada location.

So I'm compressing our test package so it can be downloaded over the network. At that point, I'm worried about bandwidth. I need to make the package as small as possible.

For the limitations you mentioned, information gathering doesn't apply to us because I programmatically grab that info myself. I only use BIT for stress testing.

As for the RAM, that could be an issue. But... it can be managed depending on how BIT works in that regard. What is it that requires an x64 executable for an x64 environment? If it's just the memtest.exe file, then that's easy enough to move around.

[David (PassMark)]

I still think you are wasting your time and inviting problems in the long term.

1Gbit/sec networking is standard. Which means you are going to save 1/10th of a second per machine. Which you could save just as easily by making a separate 32bit and 64bit package. Is anyone still selling 32bit in any case? It is really a legacy O/S now.

IMHO, if your testers are stealing dozens of $5 parts it is time to get new testers.

I trust you have licenses to cover the use of our software on the 800 machines you claim you are using it on?

[JohnDorian]

I still think you are wasting your time and inviting problems in the long term.

1Gbit/sec networking is standard. Which means you are going to save 1/10th of a second per machine. Which you could save just as easily by making a separate 32bit and 64bit package. Is anyone still selling 32bit in any case? It is really a legacy O/S now.

IMHO, if your testers are stealing dozens of $5 parts it is time to get new testers.

I trust you have licenses to cover the use of our software on the 800 machines you claim you are using it on?

What problems could be brought forth if it's just a few one file?

Our network cards are 1Gbit, as well as the entire infrastructure. As for testers stealing, that was just a tiny example. I included it because it's a possibility. Although it's fairly unlikely as they all have to go through a metal detector. Most of the replacements come from broken USBs. They get used pretty roughly.

And yes, we have the appropriate licenses, but where did you get 800 from? We never test 800 units at once.

My test package is currently 90 MBs. It compresses to 11, which is great. But I feel I would be remiss if I didn't at least try and get it smaller. I'm not sure how much a difference that cutting out ~20 MBs would do to a package that is already pretty small. But I still want to try. If I can't, I can't.

[David (PassMark)]

What problems could be brought forth if it's just a few one file?

Only being able to test 2GB of RAM with 32bit would be 1 problem. Accurate system information in the report would be another. Plus we never really tested the 32bit code on 64bit systems, as we assume that everyone using a 64bit O/S would prefer the 64bit software. So any unknown problems you encounter and come to us about we'll just tell you to switch to 64bit. i.e. you are probably creating more work for yourself in the future and lowering the quality of the testing.

but where did you get 800 from?

You said you have at least 40 testers and each tester has 20 instances of the software on USB. So 20 x 40 = 800.

[JohnDorian]

Only being able to test 2GB of RAM with 32bit would be 1 problem. Accurate system information in the report would be another. Plus we never really tested the 32bit code on 64bit systems, as we assume that everyone using a 64bit O/S would prefer the 64bit software. So any unknown problems you encounter and come to us about we'll just tell you to switch to 64bit. i.e. you are probably creating more work for yourself in the future and lowering the quality of the testing.

I suppose I'll stick with using both versions, but just for clarifications, I would use the x86 version of memtest.exe on an x64 version of the OS. The reason I asked which software was required for the x64 version, was so I could move around those parts, programamtically.

PassMark is just a small part of our testing software. We only use it for stress testing. I've created a bunch of other hardware tests. So I was thinking that if memtest was the only file, I would store both versions (x86 and x64) in separate folders. If on an x64 machine, I'd copy the x64 version to the PassMark folder and do the same for the x86 folder.

You said you have at least 40 testers and each tester has 20 instances of the software on USB. So 20 x 40 = 800.

That a miscommunication on my part. Overall, we only test less than 100 notebooks at one time.