How do I know what to upgrade for a specific software?

[Luis_V]

Hi everyone,

 

I use a program to analyze some data, but the run time is around 1 or 2 days to finish. I wanted to know what is my current bottleneck and maybe upgrade it to shorten the run time.

The software is MaxQuant. In summary, it is used to identify proteins by matching an experimental feature to a database, this way it gives the proteins in that sample. I heard that a SSD is very important as it needs to write a lot of stuff on the drive.

My computer is a Lenovo E590. Intel 8th i5, 8 GB of RAM, and M.2 NVMe storage.

I opened Task Manager and the RAM usage is at 99%, but is it that simple? I know that laptops generally cut CPU power to save battery, can it also be the case? Also, how do I assess my SSD to check if it is powerful enough?

[Electronics Wizardy]

If ram is maxed, upgrade thata first. Running out of ram means its probbly hitting swap and that will kill performance.

 

Wht does task manager look like? Screenshot?

[CaptainUseless]

If ram usage is at 99% i would upgrade that first

[Luis_V]

I took pictures of the task manager because the laptop is way too slow for a screenshot. I took one right after opening task manager(1) and another one after it "stabilized"(2).

Also, I copied this from the software's website.

Hardware requirements

• Intel Pentium III/800 MHz or higher (or compatible) although one should probably not go below a dual core processor.
• 2 GB RAM minimum.
• 2 GB RAM per thread that is executed in parallel is required.
• There is no upper limit on the number of cores. Whatever you can fit into a shared memory machine will work as long as the disk performance scales up with it.

Does it mean that if I use 7 cores than I need 14 GB of RAM?

 

Thank you very much for your help!!

 

 

[straight_stewie]

From the MaxQuant website:

Quote

MaxQuant is a quantitative proteomics software package designed for analyzing large mass-spectrometric data sets.

Datasets = RAM

Large = Lots of

 

With only 8GB of RAM, I'm willing to bet dollars to donuts that your performance slowdown is due to memory paging.

Shooting from the hip I'd say that your biggest performance improvement, atleast without building a whole new machine with a better processor, is going to come from adding more RAM.

How big are your datasets? What is the maximum amount of memory supported by your machines motherboard? Those two numbers will help you figure out how much RAM to add to your machine.

[Luis_V]

57 minutes ago, straight_stewie said:

From the MaxQuant website:

Datasets = RAM

Large = Lots of

 

With only 8GB of RAM, I'm willing to bet dollars to donuts that your performance slowdown is due to memory paging.

Shooting from the hip I'd say that your biggest performance improvement, atleast without building a whole new machine with a better processor, is going to come from adding more RAM.

How big are your datasets? What is the maximum amount of memory supported by your machines motherboard? Those two numbers will help you figure out how much RAM to add to your machine.

I see, it makes sense.

 

A data set generally have multiple files with 500 MB to 3 GB each, while the total averages around 8-10 GB. However, the useful information used by the program is only a fraction of the total size.

 

The maximum supported by the motherboard is 32 GB. In my case will 2400 vs 3200 MHz make a difference?

 

Thank you!

[Fasauceome]

32 minutes ago, Luiz.Almeida said:

I see, it makes sense.

 

A data set generally have multiple files with 500 MB to 3 GB each, while the total averages around 8-10 GB. However, the useful information used by the program is only a fraction of the total size.

 

The maximum supported by the motherboard is 32 GB. In my case will 2400 vs 3200 MHz make a difference?

 

Thank you!

make sure you quote us so we can see your responses.

[straight_stewie]

1 hour ago, Luiz.Almeida said:

In my case will 2400 vs 3200 MHz make a difference?

That's a whole conversation in and of itself, and at the end of the conversation you find out that while theoretically it can make a huge difference, in reality it often makes little to no difference, especially without considering lots of specifications and adjusting many settings. Don't just go by speed, latency also makes a big difference.
 

According to the hardware requirements:

Quote

• 2 GB RAM per thread that is executed in parallel is required.
• There is no upper limit on the number of cores. Whatever you can fit into a shared memory machine will work as long as the disk performance scales up with it.

You likely won't saturate your NVME drive with a single i5, so I wouldn't worry about that.

8 threads x 2GB/thread minimum = 16GB minimum.

I can't seem to find any information about whether or not memory usage will increase with the amount of memory available: or in other words, if you have 4GB RAM per thread, will the application use it?

I would suspect that it would as MaxQuant seems like quite a resource greedy application.

Given this, my recommendation would be:

 

Go for the largest amount of not budget RAM that you can afford, with a minimum of 16GB. Prefer quantity over speed for your application, but do get the fastest RAM in your budget and on your motherboards QVC list without sacrificing the amount of RAM you are getting.

[Luis_V]

On 3/29/2020 at 3:02 PM, straight_stewie said:

That's a whole conversation in and of itself, and at the end of the conversation you find out that while theoretically it can make a huge difference, in reality it often makes little to no difference, especially without considering lots of specifications and adjusting many settings. Don't just go by speed, latency also makes a big difference.
 

According to the hardware requirements:

You likely won't saturate your NVME drive with a single i5, so I wouldn't worry about that.

8 threads x 2GB/thread minimum = 16GB minimum.

I can't seem to find any information about whether or not memory usage will increase with the amount of memory available: or in other words, if you have 4GB RAM per thread, will the application use it?

I would suspect that it would as MaxQuant seems like quite a resource greedy application.

Given this, my recommendation would be:

 

Go for the largest amount of not budget RAM that you can afford, with a minimum of 16GB. Prefer quantity over speed for your application, but do get the fastest RAM in your budget and on your motherboards QVC list without sacrificing the amount of RAM you are getting.

Thank you very much for all the info. My plan is to buy two 16 GB from Crucial, but my budget might not allow it for now. I did some little research and found that mixing 8 GB with 16 GB is not ideal but doable. Do you have any personal experience with mixing and matching RAMs?

 

My laptop came with one stick of 8 GB at 2666 Mhz, but I read that the motherboard maximum is 2400 MHz. Since there is already a bottleneck I might not lose much from mixing the RAM.

[straight_stewie]

44 minutes ago, Luiz.Almeida said:

My laptop came with one stick of 8 GB at 2666 Mhz, but I read that the motherboard maximum is 2400 MHz. Since there is already a bottleneck I might not lose much from mixing the RAM.

I would recommend not mixing and matching RAM modules for this application.

It theoretically can be done, and is done often enough without major problem. However, it can cause certain problems that might be detectable, or might not be, and if you're working with important data, you want to minimize the opportunity for data corruption.

I would stick to like sticks, so to speak.

[Luis_V]

11 minutes ago, straight_stewie said:

I would recommend not mixing and matching RAM modules for this application.

It theoretically can be done, and is done often enough without major problem. However, it can cause certain problems that might be detectable, or might not be, and if you're working with important data, you want to minimize the opportunity for data corruption.

I would stick to like sticks, so to speak.

Awesome. Thank you very much, I appreciate it!