Ryzen 3000

[zombie67]

Don't get me wrong, I am *really* cheering on AMD with this. I want AMD to stomp Intel HARD. I want BOINC dominance!! Also, the only way to force real advancement.

I want one, and I will pay for one. But this is starting to move from frustrating to damaging, IMO.

And I don't believe any retailers promise about availability for now.

[VietOZ]

my local microcenter has 2x 3960X available as of the time of this post. I'm still aiming for another 3970X though. I have all the gear ready, just not the damn cpu.

[zombie67]

Hah! 2 days later, and still "out of stock", the cost has gone up to $2.1k. And the 3960X has also increased from $1.4k to $1.5k.

All without ever shipping a single unit.

Newegg finally added the 3970X back, but still not available. And with a second price increase, now $2,225.

They *do* have the 3960X available now. A second price increase for this too, now $2,000. That's a $600 increase since announce. Yikes!

[John P. Myers]

Please don't overpay. MSRPs are $1400 and $2000 respectively

[zombie67]

Hopefully before xmas. They were announced a few weeks ago already

Lemme see if i can find an official link.

Here: https://www.gskill.com/community/150...ts-Up-to-256GB

This does pose the question though...32GB modules do take a timing hit. I wonder if 8x16 with better timings vs. 4x32 would negate the 2 DPC penalty. I'm thinking it would.

Do whatever you want Z

Okay, I finally have a 3970X on order and need to buy the rest of the parts now. It looks like the new G.Skill 32gb DIMMs are not yet available, right? So I guess I am going down the path of "better timing" 8x16gb. Could I please get a recommendation or two for which kit to buy?

[zombie67]

CPU cooler: You probably need a high end AIO. NZXT Kraken X72 for example.

That thing is 394mm long. What chassis do you recommend for that? This is a dedicated cruncher with m.2 storage, so I don't need any drive bays, and I want plenty of room for loooooong GPUs. I was going to with a second one of these, but it can't fit a cooler that long. I guess I could go with a shorter version (X62). But I am guessing more cooling is better, for these AMD Threadrippers. It's my first, and I am not clear on the cooling requirements.

Edit: After finding no cases that fit 394mm, I notice that the description of the X72 is "Kraken X72 360MM LIQUID COOLER". They are very specific saying the radiator is 394mm. I assume the "360" comes from 3x120mm fans cheek-to-cheek, and that is kind of an industry standard for describing an AOI with that configuration. In this particular case, the radiator is longer than the fans. Fair enough.

But when the chassis manufacturer (using the chassis above as an example) says it will fit "Three 120mm or two 140mm fans and up to a 360mm radiator with a maximum thickness of 50mm without fans", do they mean it will really fit a 394mm radiator with 3x120mm fans?

I am totally confused, and I really do NOT want to have to trial-and-error vial mail order.

[John P. Myers]

That thing is 394mm long. What chassis do you recommend for that? This is a dedicated cruncher with m.2 storage, so I don't need any drive bays, and I want plenty of room for loooooong GPUs. I was going to with a second one of these, but it can't fit a cooler that long. I guess I could go with a shorter version (X62). But I am guessing more cooling is better, for these AMD Threadrippers. It's my first, and I am not clear on the cooling requirements.

Edit: After finding no cases that fit 394mm, I notice that the description of the X72 is "Kraken X72 360MM LIQUID COOLER". They are very specific saying the radiator is 394mm. I assume the "360" comes from 3x120mm fans cheek-to-cheek, and that is kind of an industry standard for describing an AOI with that configuration. In this particular case, the radiator is longer than the fans. Fair enough.

But when the chassis manufacturer (using the chassis above as an example) says it will fit "Three 120mm or two 140mm fans and up to a 360mm radiator with a maximum thickness of 50mm without fans", do they mean it will really fit a 394mm radiator with 3x120mm fans?

I am totally confused, and I really do NOT want to have to trial-and-error vial mail order.

That case will support the X72. All radiators are physically larger than the size given, especially for AIOs since they have an additional area to hold a little extra fluid. If radiators were exactly the same size as the fans that could be mounted to them, there'd be no room for the tubing to connect.

[zombie67]

That case will support the X72. All radiators are physically larger than the size given, especially for AIOs since they have an additional area to hold a little extra fluid. If radiators were exactly the same size as the fans that could be mounted to them, there'd be no room for the tubing to connect.

Got it. Thanks!

[VietOZ]

Sorry for the late reply, Z! For memory I went with this : https://www.amazon.com/gp/product/B0...?ie=UTF8&psc=1
For case: I went with this : https://www.amazon.com/gp/product/B0...0?ie=UTF8&th=1
My configuration for that case is using 5 intake fans ... 3 on the back side, 2 on the bottom. Radiator mount on top with fans configured to push the hot air out to the top. Works nicely so far and also kept my VRM and other stuff cool. I'm crunching WCG full load and cpu temp averaging around 73C at 4ghz.

[pinhodecarlos]

Hi all,

Merry Christmas to all in first place.

Need some guidance on a machine with lots of cores and at least with 128GB of memory. At NFS@Home we have a bunch of outstanding work to be done and the backlog keeps increasing. Usually we submit the NFS@Home datasets to clusters but only Greg, the admin, manages this so we at mersenneforum will soon create a crowdfunding link so we can support him on this. I would say that for a 32 thread machine a post processing task can take between one to two months of CPU. It’s is better to buy a new domestic AMD machine or Intel one? Is it better to go for a server CPU or just pay for Amazon cloud CPU time? The client uses is MPI based and if we connect nodes these must be linked with infiniband since during the post processing operation between nodes lots of data must be transferred.