Anyone in the SF Bay Area with an X399 setup?

[somebrains]

I just got there.
I have Intel systems available to me.
Everything worked fine on z270, z370, x99, x299 by setting XMP and reboot.

I'm really curious about what would happen if I stuffed a Threadripper setup with one 64gb kit at 3200, then the 2nd one.


Anyone down to experiment?

 

[Aluminum]

Corsair, usually the worst choice for anything ryzen, what timings?

Odds are pretty good this is not samsung ram and all 16GB so far are dual rank which means odds are really high you are not getting 3200, especially not on 8 slots.

Doesn't mean its not possible with other ram but fair warning that there are 0 QVL kits of 8x16GB for TR at 3200 so far, only gskill (all their high end is samsung) has one at 2933.

 

[somebrains]

Corsair, usually the worst choice for anything ryzen, what timings?

Odds are pretty good this is not samsung ram and all 16GB so far are dual rank which means odds are really high you are not getting 3200, especially not on 8 slots.

Doesn't mean its not possible with other ram but fair warning that there are 0 QVL kits of 8x16GB for TR at 3200 so far, only gskill (all their high end is samsung) has one at 2933.

Cmk64gx4m4b3200c16

It's been a long day.
We had people pitch ML, AR, VR, blockchain, etc etc etc enhanced interaction collection.
Most of the time I can get away with running a random experiment off GitHub on my Dev acct.
If any of this spaghetti sticks, I'll want a local box.

I bought a 7820x bc of that eBay 20% off promo in the hot deals section.
One day I may max out whatever CPU x299 ends up topping out at.

I like Intel, the "it just works" thing is comforting when you're drilling down pitches bc you know the understanding of layers of applications are shaky at best.

But I'll always be curious what Threadripper $vs performance will be. My use cases don't need to be tested, I just want to know if I'll have any serious options when I have to throw a box together and go with it.

Even if the box fires up at 2133 I'd call it something.

 

[Nathan_P]

Well, I've just built my 1920x system and the corsair memory I used worked at xmp timings right off the bat, mine was the 3200mhz 32GB rgb kit, but I would check the memory qvl on the mobo makers website just to be sure. But i'm not in the bay area otherwise I'd help.

 

[somebrains]

Well, I've just built my 1920x system and the corsair memory I used worked at xmp timings right off the bat, mine was the 3200mhz 32GB rgb kit, but I would check the memory qvl on the mobo makers website just to be sure. But i'm not in the bay area otherwise I'd help.

No sweat.
I'm not terribly concerned about my kits working.
Just curious about the x399 platform.

We had some issues today a local box would have helped clear up.

None of these workloads need +64gb on a local box bc they'll be shipped to AWS for Prod.

But it's nice to work out recurring deployment issues locally if possible.

 

[somebrains]

These kits are Samsung, came across it somewhere in a couple QVLs.

Sat in a room with 200 "data scientists" recently.
It was the usual deal where the vendors and thought leaders try to break down where those that build and analyze models can involve MBAs after the jobs run.

Infrastructure selection was miserable.
Most of the DCs had no idea how to select infrastructure whether it's functions, containers, VMs, bare metal.
That's not their bag usually, hopefully whoever let's them kickoff jobs knows this.

Those with more grandiose scale need to get away from managed services like Kubernetes and just go build an X299 box bc dataset import will take forever and cost a lot when you are just playing around.

I'm starting to think machine learning is the next layer of BI.
This is a team sport, and it's not just engineering deciding on final output being of any value to your business.

The potential for spend in a distributed 3-tier BI architecture is huge in time+$.

Granted I'm aaproponent of public cloud as I worked back when $1-10mil infrastructure spends would be used for less than a year before a massive pivot happened.

There's a lot of work to be done before you are production ready.

I still want to see if a x399 box will simply behave.