<davidlt[m]>
I expect JH7110 to dominate the field for some time.
<davidlt[m]>
PINE64 Star64 and VisionFive V2 will both be popular.
<davidlt[m]>
It has been a long time in making (JH7110). We were in discussions with StarFive Tech for years now probably. There is a good understanding what is needed from SW side.
<davidlt[m]>
They are so far delivering on kernlel/OpenSBI/U-Boot, but that's basically their 1st attempt thus there is a learning curve for them.
<davidlt[m]>
As long as they continue on this path JH7110 boards will be popular.
<davidlt[m]>
Not to mention that PINE64 is involved too.
<davidlt[m]>
Sipeed TH1520 based boards would be a better solution for us. They claim significantly higher performance levels, 4G/core (that's what we have now). 7 compute modules per cluster board which seems to be mini-ITX size with BMC on it. Quite compact solution. Slow IO. eMMC, but there seems to be USB per compute module, so that should be better if that's 5Gbps stuff.
<dtometzki>
davidlt, yes totally agree with you. starfive is doing a good job. How are you involved in the project ?
<davidlt[m]>
We are constantly chatting their software folks since early BealgeV days.
<davidlt[m]>
Me and at least Fu Wei from Red Hat are on their Slack.
<davidlt[m]>
Yeah, that's the Sipeed TH1520 stuff.
<davidlt[m]>
Realistically we would want Ventana Veyron V1, but that doesn't seem to be realistic. SBCs might continue to be the only path in 2023.
<davidlt[m]>
TH1520 is interesting, especially that Sipeed cluster board. VRULL seems to be involved in upstreaming T-HEAD stuff. That's promising.
<davidlt[m]>
Fu Wei will be/is working with SOPHGO SG2042, 64-core T-HEAD C910 system.
<dtometzki>
great very intresting
<davidlt[m]>
So far I don't have a lot of trust there regarding software support (TBD).
<davidlt[m]>
Hardware alone is not enough, we need software support (ideally from the vendor and upstream stuff).
<dtometzki>
if you still need supporters with pleasure
<dtometzki>
but iam not from the vendor
<davidlt[m]>
We need everything :) Hardware, software support (upstream work once the boards land), engineering hours on infrastructure, package fixing, sending out patches to dist-git/upstream, etc.
<davidlt[m]>
So far we have 20+ SiFive HiFive Unmatched boards in the farm.
<dtometzki>
so when i can do anything it would be great
<davidlt[m]>
I plan to add more folks into fedora.riscv.rocks once I am happy with the server.
<davidlt[m]>
There are only a few RH folks that currently have access (we don't have enough HW for everyone), but somehow we need to improve.
<dtometzki>
cool
<davidlt[m]>
I could provide some guidance late next week or so.
<dtometzki>
perfect
<davidlt[m]>
I plan to be traveling in the 1st half of the week so might not be that active.
davidlt has quit [Ping timeout: 268 seconds]
<dtometzki>
No problem
kalev has quit [Ping timeout: 268 seconds]
kalev has joined #fedora-riscv
davidlt has joined #fedora-riscv
leah2 has quit [Remote host closed the connection]