[ANN] TeamRedMiner - Old Thread

[sharmanov]

WOW you guys, 1.35GPS on Radeon VII running C31... I really wish you guys would implement BEAM with uplift as you did grin for AMD cards. AMAZING JOB!!!

[todxx]

WOW you guys, 1.35GPS on Radeon VII running C31... I really wish you guys would implement BEAM with uplift as you did grin for AMD cards. AMAZING JOB!!!

Thanks sharmanov

We're working on adding some other cuckoo cycle algos first, but we would eventually like to hit the equihash algos too.  It will probably be a while until we get there though.

[joseph32]

For C31: Anyone managed to get more than 1.2+ on Vega64? If yes, what clocks and timings? Thanks!

[BogdanCo]

For C31: Anyone managed to get more than 1.2+ on Vega64? If yes, what clocks and timings? Thanks!

Can you share the settings for Vega64 on C31? Thanks!

[lebuawu2]

Any info hashrate c31 for rx 580?

[todxx]

Any info hashrate c31 for rx 580?

Depends on the card and tuning (also the OS you're running on), but ballpark range is 0.5 to 0.6 g/s.

[soothaa]

Don't see it mentioned checking a few pages back, will you guys be adding support for the new AMD 5700 soon? I've been reading that the AMD tools for you devs are extremely lacking

[kerney666]

Don't see it mentioned checking a few pages back, will you guys be adding support for the new AMD 5700 soon? I've been reading that the AMD tools for you devs are extremely lacking

We have the cards, but like you write, the toolchain we usually work with isn't fully available yet and the OpenCL stack feels very broken. Compute stuff seems to run fine, but anything requiring mem is a joke. We're working on other things right now, but will of course return in the near future (hopefully) when there is a chance of producing anything of value.

[soothaa]

Don't see it mentioned checking a few pages back, will you guys be adding support for the new AMD 5700 soon? I've been reading that the AMD tools for you devs are extremely lacking

We have the cards, but like you write, the toolchain we usually work with isn't fully available yet and the OpenCL stack feels very broken. Compute stuff seems to run fine, but anything requiring mem is a joke. We're working on other things right now, but will of course return in the near future (hopefully) when there is a chance of producing anything of value.

Guess I'll just sit and drool over the new cards.. might have to invest in the new nVidia "supers" though..

[dragonmike]

Don't see it mentioned checking a few pages back, will you guys be adding support for the new AMD 5700 soon? I've been reading that the AMD tools for you devs are extremely lacking

We have the cards, but like you write, the toolchain we usually work with isn't fully available yet and the OpenCL stack feels very broken. Compute stuff seems to run fine, but anything requiring mem is a joke. We're working on other things right now, but will of course return in the near future (hopefully) when there is a chance of producing anything of value.

Guess I'll just sit and drool over the new cards.. might have to invest in the new nVidia "supers" though..

From what I hear these are not optimised yet either. Early hashrates are a little disappointing. But that should hopefully be fixed/optimised soon, considering it's hardly an architectural revolution we're talking about here.

[lebuawu2]

Any info hashrate c31 for rx 580?

Depends on the card and tuning (also the OS you're running on), but ballpark range is 0.5 to 0.6 g/s.

hi todxx,

Are you going to support algo cuckaroo29 and cuckaroo29s ? if yes, do you think the hashrate for RX 580 8GB will be comparable to GTX 1070 ?

Thanks.

[kokodesu]

WOW you guys, 1.35GPS on Radeon VII running C31... I really wish you guys would implement BEAM with uplift as you did grin for AMD cards. AMAZING JOB!!!

Thanks sharmanov

We're working on adding some other cuckoo cycle algos first, but we would eventually like to hit the equihash algos too.  It will probably be a while until we get there though.

If you are considering addition of equihash, it would be nice if you could also consider adding VDS (equihash + scrypt).

[hbouma]

Team Red Miner v0.5.3 released

Now supporting the MTP algorithm for Zcoin

As mentioned at mintpond, the miner always exits when creating the scratchpad while being ran from awesome miner. I'll get the spinning cursor for a moment and then it will exit. If I try the diagnostic mode, it runs. If I run it using the included batch file, it works there too. So I'm not sure what is up. This is on Windows 10 with A6-3500 APU and RX 470 and RX 580 cards with 18.6.1. drivers. I've been mining with sgminer-fancyl and djm34's miner for weeks before this without problems I would think the settings/hardware should be fine to at least get the miner to start. lol.

Thanks for taking the time to add MTP support!

Hal

Already made a post in Patrik's thread about that issue: https://bitcointalk.org/index.php?topic=676942.msg51710602#msg51710602

Edit @ Kerney: You can use just the standard setup. Its the same command line. In AM it crashs, without AM it runs. Would be great if we can sort out this bug.

Ok, the AM issues should be fixed in v0.5.4, todxx will update with the new release shortly.

Lol, I'm not sure what your fix did, but I still can't run TeamRedMiner on the benchmark page. It still quits right away. Setting an algo to mine with it works though.

Hal

[kerney666]

Is this miner enough to compete with the NVIDIA card in x16r algorithm?
I don't have any AMD card. so i wonder.

It can definitely compete from a hash/USD perspective, but when you factor in power draw, AMD has a tough time competing with Nvidia. It will be interesting to see how the Navi cards perform once AMD gets their shit together and releases a non-broken opencl stack.

[kerney666]

Team Red Miner v0.5.3 released

Now supporting the MTP algorithm for Zcoin

As mentioned at mintpond, the miner always exits when creating the scratchpad while being ran from awesome miner. I'll get the spinning cursor for a moment and then it will exit. If I try the diagnostic mode, it runs. If I run it using the included batch file, it works there too. So I'm not sure what is up. This is on Windows 10 with A6-3500 APU and RX 470 and RX 580 cards with 18.6.1. drivers. I've been mining with sgminer-fancyl and djm34's miner for weeks before this without problems I would think the settings/hardware should be fine to at least get the miner to start. lol.

Thanks for taking the time to add MTP support!

Hal

Already made a post in Patrik's thread about that issue: https://bitcointalk.org/index.php?topic=676942.msg51710602#msg51710602

Edit @ Kerney: You can use just the standard setup. Its the same command line. In AM it crashs, without AM it runs. Would be great if we can sort out this bug.

Ok, the AM issues should be fixed in v0.5.4, todxx will update with the new release shortly.

Lol, I'm not sure what your fix did, but I still can't run TeamRedMiner on the benchmark page. It still quits right away. Setting an algo to mine with it works though.

Hal

The fix addressed a MTP-specific seg fault in an exotic API call that Awesome Miner uses. After the fix I (and other users) could run MTP as expected.

I haven’t heard anything about a benchmark page bug though. If you can run the miner fine when specifying an algo it seems more like an issue with Awesome Miner. There is no benchmark mode in the miner, Awesome Miner would need to provide either a real or simulated pool that provides a job. Have you asked patrike in the Awesome Miner thread?

[kerney666]

MTP improvements and benchmarks

To people interested in MTP: we just released an improved version in v0.5.6, todxx will update the ann title shortly. In short, you should see a little more hash (1-3%) for a overall lower power draw than before. On most Polaris cards, the power draw has been reduced significantly.

Since the Zcoin official sgminer was released with benchmark nrs vs TRM, I took the time to produce our own benchmarks, hopefully with enough data so users easily can reproduce the tests.


TRM vs Zcoin official sgminer
-----------------------------
amdgpu-pro 18.50, listed clocks are powerplay table clocks, not effective clocks.

Note that high clocks have been chosen for benchmark purposes, the one exception
being the second test set with lower clocks for the Vega 64 LC. We do recommend
tuning down clocks rather than using the benchmark configurations below.

Negative power draw difference means lower power draw for TRM.

Hashrates are short term max speed hash rates, not long term averages affected by
pad rebuild pauses.

Note: for 580s, the nrs below might differ significantly between Windows and Linux.
This is related to driver implementations and has nothing to do with the miners.

GPU           CoreClk   MemClk   Sgminer     TRM         HR chg   Power draw diff
Vega 64 LC    1500      1107     3.33 MH/s   3.52 MH/s   + 5.7%   - 5W
Vega 64 LC    1150       500     2.32 MH/s   3.00 MH/s   +29.3%   +20W
Vega 56@64    1407      1107     2.94 MH/s   3.12 MH/s   + 6.1%   - 5W
Vega 56 Hynix 1407       945     2.97 MH/s   3.15 MH/s   + 6.1%   - 2W
570 Nitro+    1215      2050     1.59 MH/s   1.71 MH/s   + 7.5%   - 4W
580 Nitro+    1215      2050     1.77 MH/s   1.92 MH/s   + 8.5%   - 3W

[todxx]

Team Red Miner v0.5.6 released

https://github.com/todxx/teamredminer/releases

Changes in v0.5.6

• MTP improvements for Vega and Polaris (+1-3% hashrate, improved efficiency, esp Polaris)

As kerney666 already mentioned this release includes some improvements for MTP, getting both higher hashrate and consuming less power

MTP improvements and benchmarks

To people interested in MTP: we just released an improved version in v0.5.6, todxx will update the ann title shortly. In short, you should see a little more hash (1-3%) for a overall lower power draw than before. On most Polaris cards, the power draw has been reduced significantly.

Since the Zcoin official sgminer was released with benchmark nrs vs TRM, I took the time to produce our own benchmarks, hopefully with enough data so users easily can reproduce the tests.


TRM vs Zcoin official sgminer
-----------------------------
amdgpu-pro 18.50, listed clocks are powerplay table clocks, not effective clocks.

Note that high clocks have been chosen for benchmark purposes, the one exception
being the second test set with lower clocks for the Vega 64 LC. We do recommend
tuning down clocks rather than using the benchmark configurations below.

Negative power draw difference means lower power draw for TRM.

Hashrates are short term max speed hash rates, not long term averages affected by
pad rebuild pauses.

Note: for 580s, the nrs below might differ significantly between Windows and Linux.
This is related to driver implementations and has nothing to do with the miners.

GPU           CoreClk   MemClk   Sgminer     TRM         HR chg   Power draw diff
Vega 64 LC    1500      1107     3.33 MH/s   3.52 MH/s   + 5.7%   - 5W
Vega 64 LC    1150       500     2.32 MH/s   3.00 MH/s   +29.3%   +20W
Vega 56@64    1407      1107     2.94 MH/s   3.12 MH/s   + 6.1%   - 5W
Vega 56 Hynix 1407       945     2.97 MH/s   3.15 MH/s   + 6.1%   - 2W
570 Nitro+    1215      2050     1.59 MH/s   1.71 MH/s   + 7.5%   - 4W
580 Nitro+    1215      2050     1.77 MH/s   1.92 MH/s   + 8.5%   - 3W

[GKumaran]

MTP improvements and benchmarks

To people interested in MTP: we just released an improved version in v0.5.6, todxx will update the ann title shortly. In short, you should see a little more hash (1-3%) for a overall lower power draw than before. On most Polaris cards, the power draw has been reduced significantly.

Since the Zcoin official sgminer was released with benchmark nrs vs TRM, I took the time to produce our own benchmarks, hopefully with enough data so users easily can reproduce the tests.


TRM vs Zcoin official sgminer
-----------------------------
amdgpu-pro 18.50, listed clocks are powerplay table clocks, not effective clocks.

Note that high clocks have been chosen for benchmark purposes, the one exception
being the second test set with lower clocks for the Vega 64 LC. We do recommend
tuning down clocks rather than using the benchmark configurations below.

Negative power draw difference means lower power draw for TRM.

Hashrates are short term max speed hash rates, not long term averages affected by
pad rebuild pauses.

Note: for 580s, the nrs below might differ significantly between Windows and Linux.
This is related to driver implementations and has nothing to do with the miners.

GPU           CoreClk   MemClk   Sgminer     TRM         HR chg   Power draw diff
Vega 64 LC    1500      1107     3.33 MH/s   3.52 MH/s   + 5.7%   - 5W
Vega 64 LC    1150       500     2.32 MH/s   3.00 MH/s   +29.3%   +20W
Vega 56@64    1407      1107     2.94 MH/s   3.12 MH/s   + 6.1%   - 5W
Vega 56 Hynix 1407       945     2.97 MH/s   3.15 MH/s   + 6.1%   - 2W
570 Nitro+    1215      2050     1.59 MH/s   1.71 MH/s   + 7.5%   - 4W
580 Nitro+    1215      2050     1.77 MH/s   1.92 MH/s   + 8.5%   - 3W

Running on windows for days:
Vega 64 aircooled : 3.03mh/s @ 210 watts

Thx guys!!

[nordmann666]

Grin29 (Cuckarood29) would be awesome cause my little GTX1060 make 3,2sol and vega "only" 3,7sol (nbminer vs lolminer)

i´m sure you can tweak Vega over 4sol

[kerney666]

Grin29 (Cuckarood29) would be awesome cause my little GTX1060 make 3,2sol and vega "only" 3,7sol (nbminer vs lolminer)

i´m sure you can tweak Vega over 4sol

Should we release c29d (pretty good chance we'll do so), I can personally guarantee you it will be > 4 g/s for any Vega at conservative clocks or I'll both eat my shorts and give you a free version.