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.
Comments
I've already sent network-speed.xyz and lspci results via ticket.
Thanks. All the ones with issues has been Realtek chip so far.
Probably need to figure out guidelines also of what is acceptable level of performance and what is not.
I remember that there were some (quite recent) issues with the Realtek driver some kernels and an update fixed it. Perhaps you have been unlucky with a kernel before the specific fix(es)?
I don't recall the specifics but I'm sure some Google-fu will help if it is the same issue.
Old but nothing new it appears; https://bugzilla.kernel.org/show_bug.cgi?id=202945
Let's see i'll spend sometime googling
Dist upgrade Deb11 -> Deb12, Kernel 5.10 -> 6.10 did not fix this
Lots of issues overtime with realtek nics, but none of them solved this case.
I think we will move on to testing additional cooling // find free faulty unit and make sure it's very cold when testing.
So far seems it's realtek rev 15 NICs with this issue.
The replacement NICs we could use also use a Realtek chip, not sure of the revision tho.
I guess it's time to start concentrating on new year celebration rather than this
See: https://bugzilla.kernel.org/show_bug.cgi?id=217596
Yea saw that and tried the ASPM, AER, MSI modes too.
On deb11 also tried non-free drivers, made only worse.
Updated offers;
All but MD3 has upward price pressure right now.
There's also pending orders not finished yet, once delivery done more upward pressure.
This realtek nic issue throws a wrench in plans, we were planning to bring online 48 nodes hopefully by end of January; but we have no idea what % of those nodes have realtek or intel network chips.
Since we don't yet have price charts; here's manually listed the prices compared to the thread start; (hope i got them all right)
Have you tried upgrading to a recent kernel (like 6.5+) (that should work irrespective of the Debian release). Once you move to 6.4.8+ it looks like it's fixed (though with increased power consumption, which also can be overridden via sysfs settings if BIOS supports it).
YMMV.
Tested Deb12 -> Kernel 6.10, like mentioned above.
I guess you mean 6.1.x (on Debian 12).
See if you can try 6.5.x from bookworm-backports (stable-backports) and if that helps.
No, i am pretty sure it said 6.10 -- don't got access to that system anymore.
Latest in Debian 12.
https://www.debian.org/News/2023/20231210
Ah yeah 6.1.0
Guess need to test that too, once i got my hands on another realtek chip unit.
Current pricing pressure;
Does anyone but me even find this data interesting?
The key fixes are in 6.4.8+ (or thereafter), so 6.5.x should be a safe test case.
Interest is inversely proportional to how far those numbers are from 7
!!! Upgraded My Datahoarding! [1701457074, 8774164527]
I chose this for Quicksync and beacuse of a very good price.
It was 6.1.0, I upgraded it to 6.5.0 as per suggestion from @nullnothere
upgrading to 6.5.0 also did not help
Awesome!
Realized the bonus offer was ambiguous, was meant to be +1 month on each total 2 months. But i guess we are now on hook to add 2 months to both :O
Congrats ^_^
thanks for trying; We'll move towards to HW solution(s). First step, just in case, let's try added cooling on limited number of nodes.
Meanwhile, i'll get the orders in for replacement NICs ... Someone at china will be WTF?! when they see order for 200 XD
That's Awesome. Thanks a lot.
Sounds Good.
Haha yeah, they are gonna be surprised over there !!!
Is there a yabs for the i7 ones? Wasn't able to find one in this thread
here you go my friend
`curl -sL bench.monster | bash -s -- -eu
Region: Europe https://bench.monster v1.7.4 2023-12-15
Usage : curl -sL bench.monster | bash -s -- -Europe
OS : Debian GNU/Linux 11 (64 Bit)
Virt/Kernel : Dedicated / 5.10.0-25-amd64
CPU Model : Intel(R) Core(TM) i7-7700T CPU @ 2.90GHz
CPU Cores : 8 @ 900.429 MHz x86_64 8192 KB Cache
CPU Flags : AES-NI Enabled & VM-x/AMD-V Enabled
Load Average : 0.05, 0.04, 0.01
Total Space : 1.8T (231G ~14% used)
Total RAM : 15845 MB (157 MB + 15527 MB Buff in use)
Total SWAP : 7628 MB (8 MB in use)
IPv4/IPv6 : ✔ Online / ❌ Offline
Uptime : 1 days 1:49
Location : Finland, Helsinki (Uusimaa)
ASN & ISP : AS203003, Magna Capax Finland Oy / Pulsed-Media
## Geekbench v6 CPU Benchmark:
Single Core : 1452 (EXCELLENT)
Multi Core : 4716
## IO Test
CPU Speed:
bzip2 : 151 MB/s
sha256 : 228 MB/s
md5sum : 647 MB/s
RAM Speed:
Avg. write : 3993.6 MB/s
Avg. read : 7884.8 MB/s
Disk Speed:
1st run : 909 MB/s
2nd run : 903 MB/s
3rd run : 900 MB/s
Average : 904.0 MB/s
## Europe Speedtest.net
Location Upload Download Ping
Nearby 836.41 Mbit/s 925.76 Mbit/s 6.312 ms
UK, London (toob Ltd) 322.71 Mbit/s 667.50 Mbit/s 29.904 ms
Netherlands, Amsterdam (XS News) 343.83 Mbit/s 649.59 Mbit/s 32.018 ms
Germany, Berlin (DNS:NET) 288.89 Mbit/s 436.12 Mbit/s 40.988 ms
Germany, Munich (InterNetX) 362.40 Mbit/s 516.72 Mbit/s 35.594 ms
Sweden, Stockholm (SUNET) 762.31 Mbit/s 923.16 Mbit/s 5.413 ms
Norway, Oslo (NextGenTel) 289.46 Mbit/s 458.25 Mbit/s 35.890 ms
France, Paris (Orange) 270.15 Mbit/s 732.30 Mbit/s 42.746 ms
Spain, Madrid (MasMovil) 338.34 Mbit/s 428.72 Mbit/s 48.653 ms
Italy, Rome (Unidata) 292.12 Mbit/s 447.52 Mbit/s 59.425 ms
Czechia, Prague (Nordic Telecom) 318.40 Mbit/s 682.58 Mbit/s 36.883 ms
Austria, Vienna (A1) 275.75 Mbit/s 720.72 Mbit/s 41.746 ms
Poland, Warsaw (ISP Emitel) 356.87 Mbit/s 463.42 Mbit/s 45.524 ms
Ukraine, Kyiv (KyivStar) 256.77 Mbit/s 693.86 Mbit/s 44.080 ms
Latvia, Riga (Bite) 398.85 Mbit/s 744.81 Mbit/s 36.482 ms
Romania, Bucharest (iNES) 329.66 Mbit/s 516.02 Mbit/s 45.260 ms
Greece, Athens (GRNET) 139.31 Mbit/s 421.09 Mbit/s 60.520 ms
Turkey, Urfa (Firatnet) 107.57 Mbit/s 190.09 Mbit/s 134.688 ms
Finished in : 11 min 6 sec
Timestamp : 2023-12-31 17:35:08 GMT
Saved in : /root/speedtest.log
Share results:
- https://www.speedtest.net/result/15695406066.png
- https://browser.geekbench.com/v6/cpu/4212113
- https://sprunge.us/qJyptB`
i've got one i5 and one i7 and they are superb. No downtime and both have Intel chips
is this for real?
also is it in finland?
I want to buy it. i5-7500t 4c/4t 2.7Ghz，32GB DDR4，4000GB NVMe，1Gbps
Yes it is real.
All we got is in Finland, in our own DC and Network.
Those are out of stock, when they come available they go fast. We got 4TB NVMe units 4 of them at least coming, but with i5-6500t/16G i believe ... But someone (aka me) forgot to config them fully to be deployed yet. So it'll be a week or two probs before they are available for order.
you should post yabs.sh and use code formatting; Just so it's easy for people to compare
Looks like there's one free, i'll post a yabs.sh from it shortly for @gbzret4d
@PulsedMedia Where in finland is your DC?
There you go
Helsinki
sorry boss hope this time i got it right for everyone
`# ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## #
Yet-Another-Bench-Script
v2023-11-30
https://github.com/masonr/yet-another-bench-script
## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ##
Sun 31 Dec 2023 08:23:55 PM EET
Basic System Information:
Uptime : 1 days, 2 hours, 49 minutes
Processor : Intel(R) Core(TM) i7-7700T CPU @ 2.90GHz
CPU cores : 8 @ 1946.875 MHz
AES-NI : ✔ Enabled
VM-x/AMD-V : ✔ Enabled
RAM : 15.5 GiB
Swap : 7.5 GiB
Disk : 1.8 TiB
Distro : Debian GNU/Linux 11 (bullseye)
Kernel : 5.10.0-25-amd64
VM Type : NONE
IPv4/IPv6 : ✔ Online / ❌ Offline
IPv4 Network Information:
ISP : Magna Capax Finland Oy
ASN : AS203003 Magna Capax Finland Oy
Host : Pulsed-Media
Location : Helsinki, Uusimaa (18)
Country : Finland
fio Disk Speed Tests (Mixed R/W 50/50) (Partition /dev/nvme0n1p3):
iperf3 Network Speed Tests (IPv4):
Geekbench 6 Benchmark Test:
Test | Value
|
Single Core | 1478
Multi Core | 4717
Full Test | https://browser.geekbench.com/v6/cpu/4212752
YABS completed in 10 min 45 sec`
@zyborg For Code, you will have use ``` (3 backticks) at the beginning and the end instead of just one ` (backtick)
or paste the code, then select it all, and from visual thing flipped P menu choose "Code", it will prepend with several empty spaces which makes it a codeblock.
That thing never worked for me, It just adds 1 ` before and after, and the code block somehow never shows. Ideally it should, but for some reason it doesn't using just 1 `. Weird.