New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
★ VirMach ★ RYZEN ★ NVMe ★★ $8.88/YR- 384MB ★★ $21.85/YR- 2.5GB ★ Instant ★ Japan Pre-order ★ & More
This discussion has been closed.
Comments
It's been ready since about 20 hours ago, I just haven't got to it immediately as I wanted. I actually got the Geekbench done and then ended up wiping one of the servers, I'm redoing it now. All IPv4 functions, and so on. Networking looks the best out of all locations so far but I don't want to hype it up too much.
Happy to hear that. Now I'm excitedly waiting for my VirMach VPSs in Tokyo to be ready.
thanks
Geekbench scores or it didnt happen.
Oh cool, it has the timestamp too. Looks like I did this in the afternoon.
https://browser.geekbench.com/v5/cpu/13862626
Keep in mind some processor settings will be changed after this preliminary Geekbench and scores will come down a little bit to save power and keep the system more stable. Ryzen 5000 series has a tendency to run hotter and then randomly thermal throttle itself which provides inconsistent performance, and we have settings to make that a little bit more consistent. So while one may look better on a benchmark for a few minutes, the other performs better over a long period of time.
All initial activations in Tokyo will be 5900X. These are the ones that ended up being shipped out first, just a coincidence (they were closer up to the front of the office.)
oh
yeah.
its a pity not all are 5950X though. nevertheless still class leading Ryzens.
I only spent maybe 10 hours testing this specifically, but 5900X is better than 5950X in many ways outside of a host cramming in a bunch of people. In every real world situation I could throw at it, it ended up performing better over a long period of time due to these thermal throttles.
5000 series don't play well in 1U rackmount servers, they throttle a lot. And get hot a lot.
I even crammed in a bunch of fans in testing, same thing. I even tried out several different heatsinks, same thing.
It even throttles in my own computer with really good cooling. It just seems like they make it run extreme, so it looks good on all the benchmarks, and then it underclocks after some time, cools off, then goes extreme again. Unless you change the settings and do what I'm mentioning basically. If anyone wants to chime in, feel free. But to me it looks exactly like what a company would do if they wanted to look good on benchmarks.
(edit) They were actually originally going to be 5950's. I have so many of those just laying out there wondering what to do with them. Some servers will still get 5950, but I'm being cautious with the rollout.
(edit2) By the way when I put a bunch of fans, it did go down in temp, but it still throttled. And in some specific overclock modes it did do better for longer but it absolutely killed the power usage efficiency. It really does end up being 5% more performance for 30% more power.
I hope everything goes well because we don't want you to receive any meaningless work orders tickets, and of course, tickets to report technical issues.
interesting observations.
For bare metals! Nobody would say no to a 5950X in mint condition.
Fantastic
Technical issues for these Ryzen pre-orders should still be reported here first unless a ticket with more information is requested. It's a lot more streamlined right now and I keep a list, plus if something gets repeated I can quickly scroll down here whereas tickets require individual responses.
At least at the beginning and for issues originally described as potentially happening:
I do know some people in Phoenix have been waiting, and their services are 100% broken and never created properly, tickets for those are fine. And all tickets are fine really, but obviously it helps us out if there's less.
I'm speechless. I've been given a refund. If I wanted to return it, I would have applied for a refund a long time ago. I still need you to return it. The service is fucking rubbish.
I fully understand, Tokyo is HOOOOOOOT dude!
Well it looks like it's for the best you were given a refund as you do not have to endure the rubbish service.
You are a lifesaver. I was just going to spin up a Vultr server to replace a Asia server that is down. Thanks for saving me the time of setting up the server twice.
Refund all MJJs’ order and everyone will be happy.
You don’t have to deal with tons of work orders. They don’t have to put up with the “rubbish service” and we don’t have to share a same node with them.
I thought that publishing a temporary FAQ page (a static HTML file) and putting the link in the signature would reduce your tasks, but then I realized that making and updating that page is a tasksin itself...
It's fair game because everyone does it. You don't know what cooling solutions, the voltage and the mobos they use during testing. If your cpu temps look good and it's throttling, then maybe the issue is the mobo vrms. Or maybe you need to raise voltage limits. If you are using a top of the line mobo, blowing air onto the vrms and the cpu temps are low and there is still throttling, then you can blame AMD. Ryzens are "consumer" cpus anyway, so they aren't supposed to run at max clock 24/7.
There's not many server motherboards for Ryzen. Actually I think it's safe to say that basically all of them are made by ASRock and only a few SKUs.
Yeah, I mean I suppose it works out for consumers who will only burst. It makes sense, there's probably a reason the equivalent AMD Epyc seems to always use more power and have lower clock rate. I was mostly just explaining why in our use case so far it seems like 5900X would be better and how it still needs to have less power running to it to keep it cool. I do know that these chips are "designed" to handle a higher temperature, but temperature does definitely play a role in the throttling and as much as they say it's "designed" to run at 90C, I'm sure it would absolutely kill the process running 24x7 at 90C or at the very least make sure everything around it runs hotter as well.
And to be fair, my testing is usually very brutal. It's possible under normal conditions, it will be fine but in any case, the way I have them set up, it should still be fine now even under the worst conditions.
Open in several batches, can you agree on a time or date?
I got no server, but a refund today, have never complained about anything and did not opened tickets of getting refund, please check this, order id is #553742
No, sorry. If you need exact timing I'd recommend creating a ticket for a refund.
If you opened a ticket of anything, related to the pre-order, it'd be refunded.
what the fuck?
I opened ticket because the coin payment gateway was not accepting ETH
I have 100+ tickets of people wanting individual updates, individual changes, ETA, questions that have been answered here, asking to cancel, asking to refund but only under certain conditions, etc.
We can't accommodate all that, or check all of them to see which ones did or didn't want refund.
Then moving forward if these people by majority didn't ready any of the info or updates, it can be assumed they will also create tickets about every other step of this pre-order process. It's not going to be 100% smooth, so it's not the right match for customers that want continued updates.
Then in this case give me your ticket ID, I'll have a second look.
(Edit) If you paid via Coinpayments then it would not be able to be refunded anyway and it'd glitch out so you need to create a billing ticket or give me the ticket ID.
it was not even opened by me, but from the system due automated rejection and than automatic approval cause of the successful payment with Crypto, #827708
Just curious, what qualifies as server mobos? As long as there's IPMI?
My point is that if your temps are well under 90C but you experience throttling, the problem is likely with the vrms or voltage limits. You could raise the voltage limits but that's probably not a good idea.
Anyway, epycs are too expensive for SMEs that are willing to build their own servers. AMD should probably have disabled ECC memory or hard capped memory to 64 GB or something along those lines if they wanted epycs to sell better. But maybe that's not a problem when the biggest players custom build hundreds of thousands of servers at once and they even create custom chips for them.
we
want servers!
not
work orders!