Deleted | Bitcointalksearch.org

neoh1

brand new

Activity: 0

Merit: 1

Quote from: heavyarms1912 on September 09, 2020, 01:55:10 AM

Quote from: kingcolex on August 30, 2020, 08:52:43 PM

Quote from: heavyarms1912 on August 28, 2020, 05:07:22 PM

Quote from: kingcolex on August 27, 2020, 08:37:08 AM

Quote from: heavyarms1912 on August 26, 2020, 03:59:29 PM

Quote from: ?? on ??

Quote from: heavyarms1912 on August 18, 2020, 04:29:09 PM

Quote from: ?? on ??

Gpus you're good until a new more efficient model comes around and then you're almost immediately dead. Asics promise large revenues before they launch and always when they hit hands are mining for a fraction of promised revenue and then are worth nothing more than the fans in a few months.

Not true. Vegas were launched in 2017 and they still profit in 2020 that's 3 years. It all depends on how core bound an algo is only then FPGAs can shine. Even if they get on memory bound algo they won't be as many times efficient than a GPU on a core bound algo. In which case the next gen GPU will also impact this old FPGA just like it does the old gen GPU.

That's due to the hbm in vegas, they still are losing the efficiency game which has so many units out there compared to Fpga it's not comparable. Like I was saying the F1 miner by hac is using chips from 2011 and is still very profitable at standard power rates. This is an Fpga using a much newer and more efficient chip and will be profitable for a long long time.

You're offering justification for Vegas?
My comment was only for the GPU not being profitable after an year is not true. There are several variables. Only heavily core bound algos would have that kind of efficiency gains on an FPGA.

This is a core based fpga, so yes it has large efficiency gains against core based mining. When it comes to memory based mining there are now memory based fpga, they are currently hashing things like ethereum. If you're wanting to Target memory based long term I suggest looking into that, while the Core algorithms will continue to have bitstreams built for the VU9 based hardware like the TPS1530.

I think GPUs are much better off on memory based algo. The ones that out after so long on ETH are also long away from being 2x efficient than a GPU.
A 5700 Navi GPU can do 55 MHs at 130-140w at wall and costs $350 whereas the FPGA costs $1.5-2k. ROI on the GPU is 6-7 months as of today whereas on that FPGA it's 2.5-3 yrs. Will ETH be even mineable for next 3 yrs?

I also like the control you've on GPU to do a lot more than just mining whereas on FPGA you're at the mercy of the dev. I made a mistake to get in early on the Acorns. Not again.

My apologies on thread crapping. Embarrassed

I think you're letting your negative experiences with acorns dilute your view on fpga in general. For example the 5700 does 55mh a5 130-140, currently FK33 are doing 50mh at around 100w and currently still being optimized, lower clocks are easily doing 1.1w/mhs. This 1.1w is expected to be possible at the 70mhs once more optimization pushes through, and you're missing the point of an ethereum bitstream, it is to be a backup bitstream. Third party devs are already already are expecting their other algorithm bitstreams to out profit eth. There are also a lot of other possibilities mixing core and memory based mining. But focusing back to core mining and the 1530 there will be core mining for years and years, and the mainstream fpga hardware will get bitstreams for years and years.

FPGA is not the same old mining game.

When was FKs sold and when did bitstream come out?
Point being FKs were sold almost an year ago and bitstreams that makes sense just started coming out. We already have RTX 3000 series announced and Big Navi coming in soon.

RTX 3080 does 85-115 MHs. (not sure on power consumption but it should definitely be efficient that Turing especially at 85 MHs). A $699 card does that. Nope, I don't see the need to jump into FPGAs for memory bound algos.

I just wanted to come here and say that "this aged well" lol

kingcolex

sr. member

Activity: 254

Merit: 1258

Quote from: grendel25 on September 05, 2020, 02:55:50 PM

any links to info about performance benchmarks on various bitstreams/algorithms as well as power usage for the various bits/algos?

I can't justify $3K without knowing what performance I'm getting.

Thanks

The fpga discord is the best place to find out all about that. https://discord.gg/Up2tDy

kingcolex

sr. member

Activity: 254

Merit: 1258

Quote from: heavyarms1912 on August 28, 2020, 05:07:22 PM

Quote from: kingcolex on August 27, 2020, 08:37:08 AM

Quote from: heavyarms1912 on August 26, 2020, 03:59:29 PM

Quote from: ?? on ??

Quote from: heavyarms1912 on August 18, 2020, 04:29:09 PM

Quote from: ?? on ??

Gpus you're good until a new more efficient model comes around and then you're almost immediately dead. Asics promise large revenues before they launch and always when they hit hands are mining for a fraction of promised revenue and then are worth nothing more than the fans in a few months.

Not true. Vegas were launched in 2017 and they still profit in 2020 that's 3 years. It all depends on how core bound an algo is only then FPGAs can shine. Even if they get on memory bound algo they won't be as many times efficient than a GPU on a core bound algo. In which case the next gen GPU will also impact this old FPGA just like it does the old gen GPU.

That's due to the hbm in vegas, they still are losing the efficiency game which has so many units out there compared to Fpga it's not comparable. Like I was saying the F1 miner by hac is using chips from 2011 and is still very profitable at standard power rates. This is an Fpga using a much newer and more efficient chip and will be profitable for a long long time.

You're offering justification for Vegas?
My comment was only for the GPU not being profitable after an year is not true. There are several variables. Only heavily core bound algos would have that kind of efficiency gains on an FPGA.

This is a core based fpga, so yes it has large efficiency gains against core based mining. When it comes to memory based mining there are now memory based fpga, they are currently hashing things like ethereum. If you're wanting to Target memory based long term I suggest looking into that, while the Core algorithms will continue to have bitstreams built for the VU9 based hardware like the TPS1530.

I think GPUs are much better off on memory based algo. The ones that out after so long on ETH are also long away from being 2x efficient than a GPU.
A 5700 Navi GPU can do 55 MHs at 130-140w at wall and costs $350 whereas the FPGA costs $1.5-2k. ROI on the GPU is 6-7 months as of today whereas on that FPGA it's 2.5-3 yrs. Will ETH be even mineable for next 3 yrs?

I also like the control you've on GPU to do a lot more than just mining whereas on FPGA you're at the mercy of the dev. I made a mistake to get in early on the Acorns. Not again.

My apologies on thread crapping. Embarrassed

I think you're letting your negative experiences with acorns dilute your view on fpga in general. For example the 5700 does 55mh a5 130-140, currently FK33 are doing 50mh at around 100w and currently still being optimized, lower clocks are easily doing 1.1w/mhs. This 1.1w is expected to be possible at the 70mhs once more optimization pushes through, and you're missing the point of an ethereum bitstream, it is to be a backup bitstream. Third party devs are already already are expecting their other algorithm bitstreams to out profit eth. There are also a lot of other possibilities mixing core and memory based mining. But focusing back to core mining and the 1530 there will be core mining for years and years, and the mainstream fpga hardware will get bitstreams for years and years.

FPGA is not the same old mining game.

kingcolex

sr. member

Activity: 254

Merit: 1258

Quote from: heavyarms1912 on August 26, 2020, 03:59:29 PM

Quote from: ?? on ??

Quote from: heavyarms1912 on August 18, 2020, 04:29:09 PM

Quote from: ?? on ??

Gpus you're good until a new more efficient model comes around and then you're almost immediately dead. Asics promise large revenues before they launch and always when they hit hands are mining for a fraction of promised revenue and then are worth nothing more than the fans in a few months.

Not true. Vegas were launched in 2017 and they still profit in 2020 that's 3 years. It all depends on how core bound an algo is only then FPGAs can shine. Even if they get on memory bound algo they won't be as many times efficient than a GPU on a core bound algo. In which case the next gen GPU will also impact this old FPGA just like it does the old gen GPU.

That's due to the hbm in vegas, they still are losing the efficiency game which has so many units out there compared to Fpga it's not comparable. Like I was saying the F1 miner by hac is using chips from 2011 and is still very profitable at standard power rates. This is an Fpga using a much newer and more efficient chip and will be profitable for a long long time.

You're offering justification for Vegas?
My comment was only for the GPU not being profitable after an year is not true. There are several variables. Only heavily core bound algos would have that kind of efficiency gains on an FPGA.

This is a core based fpga, so yes it has large efficiency gains against core based mining. When it comes to memory based mining there are now memory based fpga, they are currently hashing things like ethereum. If you're wanting to Target memory based long term I suggest looking into that, while the Core algorithms will continue to have bitstreams built for the VU9 based hardware like the TPS1530.

heavyarms1912

full member

Activity: 729

Merit: 114