Time to break even for AMD's RX Vega 64 mining XMR Monero: 4.54 months.
Time to break even for 3x RX Vega 64 mining XMR Monero: 4.4 months.
Time to break even for NVIDIA GTX TITAN V mining XMR Monero: 1 year, 10.9 months.
Time to break even for NVIDIA GTX TITAN V mining ETH: 1 year, 3 months.
I wonder why the article fails to ever actually point out that Titan V is actually delivering more performance per watt. Obviously Vega is the better value proposition regardless.
I wonder why the article fails to ever actually point out that Titan V is actually delivering more performance per watt. Obviously Vega is the better value proposition regardless.
Volta 12nm and HBM, so I'd be surprised if it wasn't more efficient than Vega. But if you are buying a brand new $700 card, efficiency is whatever. For second-hand owners it might be more important when cross-shopping last year's high-end with this years midrange.
Volta 12nm and HBM, so I'd be surprised if it wasn't more efficient than Vega. But if you are buying a brand new $700 card, efficiency is whatever. For second-hand owners it might be more important when cross-shopping last year's high-end with this years midrange.
For the consumer I agree that efficiency is only relevant to anyone when a user's brand of choice is the one currently excelling therein. However for mining there are people who run large farms looking to strike gold, and at that scale efficiency is a big deal. Probably not a big enough deal to justify the Titan V specifically, but nonetheless.
I must say, though, and no offense to any Vega owners (I'd buy a 56 if I was in the market for a GPU today), but its efficiency is really Fermi-tier bad, to the point where it has to at least be acknowledged. A PSU sufficient for a 1080 build may not be sufficient for the same build with a V64.
For the consumer I agree that efficiency is only relevant to anyone when a user's brand of choice is the one currently excelling therein. However for mining there are people who run large farms looking to strike gold, and at that scale efficiency is a big deal. Probably not a big enough deal to justify the Titan V specifically, but nonetheless.
I must say, though, and no offense to any Vega owners (I'd buy a 56 if I was in the market for a GPU today), but its efficiency is really Fermi-tier bad, to the point where it has to at least be acknowledged. A PSU sufficient for a 1080 build may not be sufficient for the same build with a V64.
An RX Vega is also slightly more powerful than a GTX 1080. To get V64 power out of a 1080, it would need V64 power. I'd rather have another inch under the gas pedal.
And vega cards require constant babysiting to maintain that speed. On top of a mountain of other issues with mining with vega cards.
The XMR train left last month, you're all a little late.
And vega cards require constant babysiting to maintain that speed. On top of a mountain of other issues with mining with vega cards.
The XMR train left last month, you're all a little late.
I wonder why the article fails to ever actually point out that Titan V is actually delivering more performance per watt. Obviously Vega is the better value proposition regardless.
But then you would need a motherboard, that can hold those Vega, and not melt. Along with a much beefier PSU. or multiple PSUs. Then there is in-house wiring to consider.... I mean, if you live in an igloo in Canada, and get free elec, go for it.
But then you would need a motherboard, that can hold those Vega, and not melt. Along with a much beefier PSU. or multiple PSUs. Then there is in-house wiring to consider.... I mean, if you live in an igloo in Canada, and get free elec, go for it.
I wonder why the article fails to ever actually point out that Titan V is actually delivering more performance per watt. Obviously Vega is the better value proposition regardless.
Isn’t Cryptonight that algorithm that is very memory and communication heavy such that it achieves almost no speed up over CPU (purposely done to make it ASIC resistant)?
$7 a day and plummeting fast on an $800 card is not a good idea, especially considering that the cryptonight algo is the only thing they do well for the price.
Last week my vega card rig was making ~$12 a day each.
$7 a day and plummeting fast on an $800 card is not a good idea, especially considering that the cryptonight algo is the only thing they do well for the price.
Last week my vega card rig was making ~$12 a day each.
Just checked, I get 612 H/s on a single NUMA node on my old Sandy Bridge CPU... Titan V has less than 100% speed up according to that with similar power consumption (~120W)
E:
Would probably be more efficient to mine on a TR or EPYC system than on GPU tbh
Actually, VEGA 56 getting over 2kh/s (average is 2kh/s, max is about 2150h/s for now). Give it few more months for proper miner support and good timings, and it will do over 2.2kh/s average.
Hell, i get 1kh/s on my RX 480/580 cards.
Isn't Cryptonight that algorithm that is very memory and communication heavy such that it achieves almost no speed up over CPU (purposely done to make it ASIC resistant)?
Scrypt was meant to be resistant in that way. Cryptonight is similar with its main feature being that it is very latency sensitive. It fits in modern CPU caches but is too large for GPUs to do efficiently at the moment. As such, it's suitable for CPU mining right now.
Scrypt was meant to be resistant in that way. Cryptonight is similar with its main feature being that it is very latency sensitive. It fits in modern CPU caches but is too large for GPUs to do efficiently at the moment. As such, it's suitable for CPU mining right now.
Yeah, pretty much. Pretty poor algorithms to run power hungry GPUs on. Granted that if @ku4eto's benchmarks are true, that's very slow and inefficient even when compared to 6 year old CPUs.
Yeah, pretty much. Pretty poor algorithms to run power hungry GPUs on. Granted that if @ku4eto
's benchmarks are true, that's very slow and inefficient even when compared to 6 year old CPUs.
Welp, i can screenshot you my sgminer. No Vega tho. I may get a screenshot from Wolf for this.
Also, i can show you screenshot on how it runs on 2 cores on E5-2660 v2 (with and without double L3 usage), and on a FX-8350.
A single CPU is handling better the cryptonight algo, than a GPU.
But you cant scale it properly. For whatever price you want. Thats why GPU's are preferred.
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Related Threads
?
?
?
?
?
Ask a question
Ask a question
Overclock.net
27.8M posts
541.2K members
Since 2004
A forum community dedicated to overclocking enthusiasts and testing the limits of computing. Come join the discussion about computing, builds, collections, displays, models, styles, scales, specifications, reviews, accessories, classifieds, and more!