Jump to content

Budget (including currency): ~1000 Euro without GPU

Country: germany

Games, programs or workloads that it will be used for: Gaming (Overwatch rn, but might also be stuff like forza horizon in the future) & Machine Learning

Other details (existing parts lists, whether any peripherals are needed, what you're upgrading from, when you're going to buy, what resolution and refresh rate you want to play at, etc):

So I am already more or less set on a few components, I am just wondering if there are any inputs on mainly the motherboard and power supply. I looked it up and it seems like right now the 13600K is the best value (integrated graphics because those 30 bucks extra are worth the hours of debugging it can save...) and id like to upgrade to DDR5 to be able to still game modern games in a few years time without losing much speed.

 

The Motherboard has a few requirements tho.

FIrstly, it needs two full on GPU slots. I wont fill them right now, but I want the option to do so once I train bigger models.

 

Secondly, I probably need at least 2 NVME slots - more wouldnt hurt. ML needs a lot of space. I could also use SSDs, but honestly I dont really see much of a price advantage at this point (do correct me if I am wrong tho lol). For now I only need a new 2TB drive.

 

Ram... idk? Probably 64GB? Which ones tho?

 

Power supply: I was thinking a 1000W titanium power supply. Mainly because with ML stuff runs quite long and once you run stuff over night, you do not want to lose 20% of the energy. I was thinking Seasonic, but maybe there are more bang for the buck ones? I dont want low quality ones tho.

 

The case should be the Corsair 4000D Airflow. However, I was wondering if that one also exists without glass panels but with some metal stuff? and are fans always included there? If yes, id use those. If not, id use the old ones from my pc rn.

 

As for monitors, any recommendations for 1440p monitors that can be used in a 3 monitor configuration? They are not included in the budget, I will probably get them over time, but they should still not be too expensive and have at least 144Hz...

 

 

So, I hope this was not too much of a tall order, if needed idm being over budget by a little, do argue as to why tho. I only saw one other Machine learning build and that was purely for ML and on an older technology, id like to at least have the option for DDR5.

 

If you have any youtube videos that target this combination, feel free to share them! I have really not seen many videos that target such a build (or say why its a bad idea).

 

Thanks in advance!

 

Link to comment
https://linustechtips.com/topic/1503237-machine-learning-x-gaming-build/
Share on other sites

Link to post
Share on other sites

3 minutes ago, Potted Rose Petal said:

Budget (including currency): ~1000 Euro without GPU

Country: germany

Games, programs or workloads that it will be used for: Gaming (Overwatch rn, but might also be stuff like forza horizon in the future) & Machine Learning

Other details (existing parts lists, whether any peripherals are needed, what you're upgrading from, when you're going to buy, what resolution and refresh rate you want to play at, etc):

So I am already more or less set on a few components, I am just wondering if there are any inputs on mainly the motherboard and power supply. I looked it up and it seems like right now the 13600K is the best value (integrated graphics because those 30 bucks extra are worth the hours of debugging it can save...) and id like to upgrade to DDR5 to be able to still game modern games in a few years time without losing much speed.

 

The Motherboard has a few requirements tho.

FIrstly, it needs two full on GPU slots. I wont fill them right now, but I want the option to do so once I train bigger models.

 

Secondly, I probably need at least 2 NVME slots - more wouldnt hurt. ML needs a lot of space. I could also use SSDs, but honestly I dont really see much of a price advantage at this point (do correct me if I am wrong tho lol). For now I only need a new 2TB drive.

 

Ram... idk? Probably 64GB? Which ones tho?

 

Power supply: I was thinking a 1000W titanium power supply. Mainly because with ML stuff runs quite long and once you run stuff over night, you do not want to lose 20% of the energy. I was thinking Seasonic, but maybe there are more bang for the buck ones? I dont want low quality ones tho.

 

The case should be the Corsair 4000D Airflow. However, I was wondering if that one also exists without glass panels but with some metal stuff? and are fans always included there? If yes, id use those. If not, id use the old ones from my pc rn.

 

As for monitors, any recommendations for 1440p monitors that can be used in a 3 monitor configuration? They are not included in the budget, I will probably get them over time, but they should still not be too expensive and have at least 144Hz...

 

 

So, I hope this was not too much of a tall order, if needed idm being over budget by a little, do argue as to why tho. I only saw one other Machine learning build and that was purely for ML and on an older technology, id like to at least have the option for DDR5.

 

If you have any youtube videos that target this combination, feel free to share them! I have really not seen many videos that target such a build (or say why its a bad idea).

 

Thanks in advance!

 

What GPU are you going to use?

 

Depending on which kind of model/dataset you're going to train on, maybe you could go for 2x48GB kits and then upgrade later if needed, but 96gb should give you some really nice headroom to begin with.

 

Your second GPU slot is likely going to be x4, unless you go for a really fancy MB, but that shouldn't be a problem for ML. You'll likely need risers tho and have a way to mount those GPUs vertically since having them that close to one another would make the top one overheat.

FX6300 @ 4.2GHz | Gigabyte GA-78LMT-USB3 R2 | Hyper 212x | 3x 8GB + 1x 4GB @ 1600MHz | Gigabyte 2060 Super | Corsair CX650M | LG 43UK6520PSA
ASUS X550LN | i5 4210u | 12GB
Lenovo N23 Yoga

Link to post
Share on other sites

18 minutes ago, igormp said:

What GPU are you going to use?

 

Depending on which kind of model/dataset you're going to train on, maybe you could go for 2x48GB kits and then upgrade later if needed, but 96gb should give you some really nice headroom to begin with.

 

Your second GPU slot is likely going to be x4, unless you go for a really fancy MB, but that shouldn't be a problem for ML. You'll likely need risers tho and have a way to mount those GPUs vertically since having them that close to one another would make the top one overheat.

I have yet to fully decide. I think the price constraint on the memory will be hard as is tho, so idk if I need to go more. In the beginning I will probably go 4070 tho, but I will probably upgrade at some point and leave the rest of the build as is. Or just add a second one. Thats why I would prefer to not talk about GPU for now, as it should only be important to the PSU for now.

Link to post
Share on other sites

46 minutes ago, Potted Rose Petal said:

I have yet to fully decide. I think the price constraint on the memory will be hard as is tho, so idk if I need to go more. In the beginning I will probably go 4070 tho, but I will probably upgrade at some point and leave the rest of the build as is. Or just add a second one. Thats why I would prefer to not talk about GPU for now, as it should only be important to the PSU for now.

Oh, if you have further constraints then yeah, maybe go for 2x32gb and that'd be it.

 

If you are looking at 4070-level GPUs, then there's no need to go for 1000W, 850W would be more than enough, even for multiple GPUs. You'd be power limiting them for maximum efficiency anyway.

 

What are your priorities for this build anyway? Gaming or ML?

FX6300 @ 4.2GHz | Gigabyte GA-78LMT-USB3 R2 | Hyper 212x | 3x 8GB + 1x 4GB @ 1600MHz | Gigabyte 2060 Super | Corsair CX650M | LG 43UK6520PSA
ASUS X550LN | i5 4210u | 12GB
Lenovo N23 Yoga

Link to post
Share on other sites

10 hours ago, igormp said:

Oh, if you have further constraints then yeah, maybe go for 2x32gb and that'd be it.

 

If you are looking at 4070-level GPUs, then there's no need to go for 1000W, 850W would be more than enough, even for multiple GPUs. You'd be power limiting them for maximum efficiency anyway.

 

What are your priorities for this build anyway? Gaming or ML?

Hmm the reason I want to go 1000W is because rn I have a 550 PSU and cant use it anymore, which is annoying. I want to be able to use the PSU for the next 10-20 years for sure, which is simply not given if I ever decide to upgrade to the 90-series card in dual configuration.

 

as to what I want, its a sweetspot between? I want to game sometimes and do ML sometimes. It shouldnt be unable to do one thing of those two to the fullest extent of the hardware, meaning, I probably wont powerlimit anything (besides maybe the cpu lol)

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×