scryptr
Legendary
Offline
Activity: 1793
Merit: 1028
|
|
October 15, 2018, 09:03:23 PM |
|
GDDR5 MEMORY DOES NOT NEED A PILL--
It performs at full compute capacity without a Pill on nVidia architecture. GDDR5X memory on 1080/1080ti cards does not perform at full compute capacity without the Pill. With the Pill, the larger number of CUDA cores on a 1080/1080ti card yields a proportionally higher hash rate compared to lower 10-series cards. Without the Pill, the mining performance (compute performance) is not proportionally greater. --scryptr
OhGodAGirl has definitely alluded to the existence of a “pill-like” solution for GDDR5 cards. But it was only privately available. It’s not about “compute capacity” (which isn’t even a real metric), it’s about memory timings. The pill works by injecting new memory timings into the GPU in real time. If we could modify and flash the BIOS of nvidia GPUs, the pill wouldn’t even be necessary. You can always improve the performance with memory intensive algorithms by running tighter timings, and that’s exactly what the mining specific GPUs do. YOU HAVE A POINT-- But, the compute capacity of the CUDA cores is near fully realised on GDDR5 cards. The improvement for the GDDR5X memory cards is far greater with the Pill. If there is a pay solution in existence for GDDR5 cards, it is likely not worth any investment for the miner with a handful of rigs. Where did OhGodAGirl make these comments? --scryptr
|
|
|
|
|
|
In order to achieve higher forum ranks, you need both activity points and merit points.
|
|
|
Advertised sites are not endorsed by the Bitcoin Forum. They may be unsafe, untrustworthy, or illegal in your jurisdiction.
|
|
|
gsrcrxsi314
Member
Offline
Activity: 367
Merit: 34
|
|
October 16, 2018, 12:11:35 AM |
|
the cuda cores have almost nothing to do with the memory algorithms. AMD cards dont even have cuda cores, and they see performance benefits, why? from increased memory timings and speed, just like with nvidia cards. you need just enough core speed to feed the data in and out of the memory.
i don't know where her comments are, i just remember reading it along with other peoples comments. there was also an interview she did where i think she mentioned it. look around. you should find it
|
|
|
|
cudapop
Member
Offline
Activity: 93
Merit: 41
|
|
October 16, 2018, 03:45:13 AM Last edit: October 16, 2018, 03:55:48 AM by cudapop |
|
... But, the compute capacity of the CUDA cores is near fully realised on GDDR5 cards. ...
Actually you can easily test this to show that the cores are not fully utilized with GDDR5 memory: download the source code of ethminer from github, modify the kernel so that in the DAG sampling loop instead of reading from global memory it reads from shared* memory, and you'll be able to see what the CUDA cores are capable of if they were accessing very low latency shared memory as opposed to the much higher latency global memory. For actual figures, a GTX 1060 which would normally hash at around 23 MH/s when it uses global memory, is capable of around 53 MH/s when it is using shared memory instead. This shows what it's cores are capable of if they had very low latency memory to use instead of the slower GDDR5 device memory. (* for non-CUDA programmers: shared memory is program-controllable cache memory for data sharing within a thread block and has the same latency as the L1/texture caches)
|
|
|
|
crypto4pizza
Newbie
Offline
Activity: 102
Merit: 0
|
|
October 26, 2018, 05:47:59 AM |
|
I had the executable saved on the desktop of my rigs. They all disappeared. Has this happened to anyone else??
|
|
|
|
peteycamey
Legendary
Offline
Activity: 1184
Merit: 1017
|
|
October 26, 2018, 08:34:07 AM |
|
I had the executable saved on the desktop of my rigs. They all disappeared. Has this happened to anyone else??
your anti-virus probably deleted it, thinking it is a virus.
|
|
|
|
androstan1234
Jr. Member
Offline
Activity: 251
Merit: 6
|
|
November 05, 2018, 04:37:09 PM |
|
Does anyone else notice that the 1080ti on the pill loses about 1-2 MH/sec in the first few minutes of mining, then stabilizes? Mine have consistently done that, and I'm not sure why. I'm still getting 50-51 MH/sec on each, stable, so I'm pretty satisfied, but curious if anyone else sees this effect and has an idea of the cause.
|
|
|
|
scryptr
Legendary
Offline
Activity: 1793
Merit: 1028
|
|
November 15, 2018, 03:50:12 PM |
|
Any news on the 1070ti pill?
W/o pill gtx 1070ti makes like gtx 1080 with pill, which for 1070ti's memory imposible! GDDR5X MEMORY-- The memory used by 1080/1080ti cards is GDDR5X memory. Without the Pill, these cards give less than optimal mining performance. GTX cards in the 10 series below the 1080 level use GDDR5 memory, which performs well as-is. There is no big need for a Pill for GDDR5 memory cards. Better mining code may help. That said, there may be a possible Pill for 980/980ti cards, as they also have GDDR5X memory. A 980ti card will mine ETH only slightly better than a 970 card. The greater number of CUDA cores in a 980ti does not result in a proportionally greater perfromance. A 980ti with a Pill should mine ETH about 28+ MH/s. Performance with GDDR6 memory (2080ti) has yet to be determined publicly. --scryptr 980/980ti use GDDR5 also, not GDDR5X. GDDR6 MEMORY NEEDS A PILL-- The RTX 2080ti cards mine ETH only slightly faster than a GTX 1080ti on the Pill (57-58MH/s vs 52-54MH/s). This is not proportional to the on-chip processing capability. There have been a few recent reports. And, GTX 980-980ti cards have GDDR5X memory, they are not capable of mining ETH much faster than a GTX 970 with GDDR5. This is not proportional to the on-chip processing capability, the CUDA cores. --scryptr
|
|
|
|
|
Dr_Victor
|
|
November 17, 2018, 05:22:11 PM |
|
If you post your result in this thread gtx 1080 ti makes 55+ Mh/s with pill.
|
yobit.net is banned from signatures
|
|
|
peteycamey
Legendary
Offline
Activity: 1184
Merit: 1017
|
|
November 17, 2018, 06:28:12 PM |
|
we need a pill (optimized memory timings) for X22i algo.
|
|
|
|
scryptr
Legendary
Offline
Activity: 1793
Merit: 1028
|
|
November 18, 2018, 02:04:53 AM Last edit: November 18, 2018, 05:48:54 AM by scryptr |
|
we need a pill (optimized memory timings) for X22i algo.
CRYPTO-MINING NEEDS A PILL-- The market is a shade away from being futile. But seriously, I thought the chain-type algos (x16r, x16s, x22i, etc) were mostly processor bound, and high energy. How would a blanket memory timing help over various chained algorithms? --scryptr
|
|
|
|
peteycamey
Legendary
Offline
Activity: 1184
Merit: 1017
|
|
November 18, 2018, 01:37:44 PM |
|
we need a pill (optimized memory timings) for X22i algo.
CRYPTO-MINING NEEDS A PILL-- The market is a shade away from being futile. But seriously, I thought the chain-type algos (x16r, x16s, x22i, etc) were mostly processor bound, and high energy. How would a blanket memory timing help over various chained algorithms? --scryptr some of these algos are memory reliant, it would increase the overall hash-rate.
|
|
|
|
crairezx20
Legendary
Offline
Activity: 1638
Merit: 1046
|
|
November 18, 2018, 02:08:42 PM |
|
we need a pill (optimized memory timings) for X22i algo.
Why not contact ohMyGirl or ohmygod company about your request because the last ETHpill was developing by ohmygod company. I was waiting for this if they can make a pill for gtx 1080ti x22i algos.
|
|
|
|
iSparta
Member
Offline
Activity: 728
Merit: 10
|
|
November 20, 2018, 06:38:35 PM |
|
This is a very good application that significantly increases the hashrate of GDDR5X memory cards during the mining of Ether. It would be nice to create the same programs for other popular cryptocurrencies, for example, Monero.
|
|
|
|
impynick
Jr. Member
Offline
Activity: 77
Merit: 6
|
|
November 20, 2018, 08:10:42 PM |
|
This works with Monero as well on 1080ti.
|
|
|
|
scryptr
Legendary
Offline
Activity: 1793
Merit: 1028
|
|
November 21, 2018, 06:06:56 PM |
|
we need a pill (optimized memory timings) for X22i algo.
CRYPTO-MINING NEEDS A PILL-- The market is a shade away from being futile. But seriously, I thought the chain-type algos (x16r, x16s, x22i, etc) were mostly processor bound, and high energy. How would a blanket memory timing help over various chained algorithms? --scryptr some of these algos are memory reliant, it would increase the overall hash-rate. CORE-INTENSIVE ALGOS MAY PERFORM POORLY-- The pill works for ETH by design, and helps with Monero variants, but some algos like ProgPow perform poorly with the Pill. With the chained algos, the pill would need to detect which algo is being mined and switch ON or OFF accordingly. This may not be practical to code. If the algo's performance is neutral to the pill, it could remain active. I game with the Pill, and mine Monero after gaming. The Pill just runs, with no problems detected. I don't code enough to actually know what is possible. --scryptr
|
|
|
|
joseph32
Member
Offline
Activity: 413
Merit: 21
|
|
December 19, 2018, 02:57:21 AM |
|
Sorry for this question... I have a rig mixed with 1080's and 1070's. How can I exclude the 1070's, so the pill only runs with the 1080's? Tried with the --revA 1,5,... But didn't worked
|
|
|
|
scryptr
Legendary
Offline
Activity: 1793
Merit: 1028
|
|
December 19, 2018, 04:11:58 PM |
|
Sorry for this question... I have a rig mixed with 1080's and 1070's. How can I exclude the 1070's, so the pill only runs with the 1080's? Tried with the --revA 1,5,... But didn't worked JUST RUN IT-- It only affects the 1080 cards. There is no need for any command line flags. The Pill will not affect the 1070 cards. --scryptr
|
|
|
|
VasilyS
|
|
December 19, 2018, 04:56:08 PM |
|
we need a pill (optimized memory timings) for X22i algo.
CRYPTO-MINING NEEDS A PILL-- The market is a shade away from being futile. But seriously, I thought the chain-type algos (x16r, x16s, x22i, etc) were mostly processor bound, and high energy. How would a blanket memory timing help over various chained algorithms? --scryptr some of these algos are memory reliant, it would increase the overall hash-rate. CORE-INTENSIVE ALGOS MAY PERFORM POORLY-- The pill works for ETH by design, and helps with Monero variants, but some algos like ProgPow perform poorly with the Pill. With the chained algos, the pill would need to detect which algo is being mined and switch ON or OFF accordingly. This may not be practical to code. If the algo's performance is neutral to the pill, it could remain active. I game with the Pill, and mine Monero after gaming. The Pill just runs, with no problems detected. I don't code enough to actually know what is possible. --scryptr I tried with ETH Enlargment pill with Phoenix miner - it doesn't work. On x16r it doesn't work as well. But it works perfectly with Claymore miner while Ether mining.
|
|
|
|
RivAngE
Full Member
Offline
Activity: 728
Merit: 169
What doesn't kill you, makes you stronger
|
|
December 19, 2018, 06:19:19 PM |
|
I tried with ETH Enlargment pill with Phoenix miner - it doesn't work. On x16r it doesn't work as well. But it works perfectly with Claymore miner while Ether mining.
And I'm using Ethminer (0.17 rc2) because I don't want to support Claymore, I didn't like his attitude in some of his posts and he has given up on new updates, at least for public usage. Ethminer hashes at 57 MH/s, that's the pool's side average.
|
|
|
|
|