Topic: [XMR] JCE Miner Cryptonight/forks, now with GPU! - page 82. (Read 90841 times)

dont know what you all do...my Ryzen 1700@3800MHz and 3200MHz Mem "only" get 1500hs on cnlite
tried the "best config" from JCE and also yours

Im getting quite consistent over 2KH/s with my Ryzen 7 2700X for mostly any cnlite variant with the following config:

"cpu_threads_conf" :  
[  
     { "cpu_architecture" : "ryzen", "affine_to_cpu" : 0, "use_cache" : true, },
     { "cpu_architecture" : "ryzen", "affine_to_cpu" : 2, "use_cache" : true, "multi_hash" : 2, },
     { "cpu_architecture" : "ryzen", "affine_to_cpu" : 4, "use_cache" : true, "multi_hash" : 2, },
     { "cpu_architecture" : "ryzen", "affine_to_cpu" : 6, "use_cache" : true, "multi_hash" : 2, },
     { "cpu_architecture" : "ryzen", "affine_to_cpu" : 8, "use_cache" : true, },
     { "cpu_architecture" : "ryzen", "affine_to_cpu" : 10, "use_cache" : true, "multi_hash" : 2, },
     { "cpu_architecture" : "ryzen", "affine_to_cpu" : 12, "use_cache" : true, "multi_hash" : 2, },
     { "cpu_architecture" : "ryzen", "affine_to_cpu" : 14, "use_cache" : true, "multi_hash" : 2, },
]


What i basically did was to look how large the scratchpad is for cnlite (2MB) and tried to use as much of my CPUs cache for the scratchpad, thus the 6 cores doing "multi_hash": 2
I think that a ryzen is quite good for cryptonight, bad for cnheavy but exceptionally good for cnlite.

Hope this helps you to maybe squeeze a few H/s more out of your ryzen.

This is a very interresting point.

Let's imagine something: Since I've dedicated netcode for Nicehash, it would be very simple for me just not to report the bad shares. I would display a correct physical 490 H/s and no reject.
Question: in such case, would you come back to JCE?

The morale to understand is about the Effective hashrate, do you get a better real-life hashrate on JCE than on other miners? A lot of user say so. Don't get too focused on error messages, and especialy not on those from Nicehash which has a very strange behavior, look at the real end results. If JCE displays more errors, maybe that's just because it never hides any one.

About nicehash, it has a zero tolerance about Stale shares. If a share gets stale a the very end of its computation, JCE doesn't drop it and still send it to Nicehash, in case it could be accepted, betting on the ping delay. If it fails, it logs the failure. Normal.

Btw, i get better results with multi_hash 432 or 464, and reaches ~518, which is crazy considering the rx550 is such a small low power card. But the RX550 and 560 depend a lot on the memory quality, my worse rx560 never went above 355h/s, which is stupid low.


I still have room for CN-Heavy optims, but i'm getting stuck with my old cards, they're close to hardware max and do not report well the speed increases. My bonaire for example is still at 268 h/s whatever i try, and claymore was also maxing at that exact same number.
On my dual HD 7950 i'm now at 1096, still behind claymore 9.7 (1146) but above other miners.

Back to CPU, JCE is supposed to be the fastest, since i'm the only dev to use dedicated assemblies for every possible combination of CPU+algo+multihash+cache+..... This is why the binary is a trillion times bigger than others. Also i've found a OpenCL optim i could port back to CPU, even the 64-bits. The hashrate may be something like +0.2% higher. Since i'm playing with the hardware max, that's a welcome improvement, yet negligeible.

The best CN-Light/Turtle/Aeon config for R 1600 is in the github doc:

IPBC is no longer Light, now it's Heavy, i need to update my doc.

BTW, at some point I got 2,1KH/s, which is more than the best I've ever had with cast XMR.

JCE-Miner, I'm running now the miner in my 2nd Ryzen. Is a Ryzen 5 1600 at 3.75GHz, RAM at 2933.

Is incredible! 2KH/s (just 10% less than R7 1700) mining CN-Lite V1

It's the --auto tuning the best optimisation so far to R5 1600?

The new GPU miner 0.31c is indeed a bit faster in BitTube v4 (CN Heavy variant) than its predecessor. My Radeon RX 580 4GB @stock hashes with 672 H/s now (vs. 655 H/s with the old 0.30c). Settings are:

Higher intensity is not possible because a FullHD display is connected.

CN bittube v4 with RX 570 4GB
Display GPU intensity 384 on both threads (C 1230Mhz; M 2100MHz)
Ohter GPUs: intensity 464. (GPU 1:1220/2020. GPU2: 1230/2025).
Total: 2053H/s. 340W at the wall

Too much "Share above target."  on pool: cryptonightv7.eu.nicehash.com:3363
config.txt:
GPU:sapphire RX550 2GB(1071/1500@1100/1950) + CPU: XEON E5430, Hash: 490H/s + 82H/s
Win10x64(ver:1709), drv: Win7-Crimson-ReLive-Beta-Blockchain-Workloads-Aug23.
Back on SRBminer(1.6.3) + JCE_cpu_miner.(0.31c), Hash: 482H/s + 82H/s
SRB:

Thanks!

Iamtutut: The cpu and gpu version always support the same set of forks, the missing Bittube in the zip of yesterday was just a package goof i fixed.

Great! And I works! 

Is then bittube still CPU only with your miners ?

Bittube v2 has both the complexity of CN-Heavy and CN-IPBC. This is why it's clearly slower than others.

My opencl in the 0.31c was based on the pre-bittube code, sorry i just got confused with my own files and shipped the bad release. I reuploaded it. Same version, no other change, that's just to fix a bad zip upload.

There IS if you need to stay with Windows 7, for what ever reason, I've got mine.

It's not slower...

I don't know how old your cards, but I use Claymore 11.3 on my R9 270X 4Gb with latest 18.6.1 drivers on Windows 10. Speed is the same as on old drivers. There is no need to stay on so old drivers...

The BitTube code of the new GPU miner 0.31c seems to be broken. I immediately get rejected shares with that version. 0.30c instead works with the same settings.

Win7 ... 18.x drivers don't allow to enable compute mode in Win7 :/ so I use blockchain driver or 15.12 (as its best for Claymore miners on old cards)

It's a variation of Heavy.

newBittube is not heavy algo? From speed looking like it's on heavy...

Hi, JCE-Miner

Thanks for that very good miner  

Testing last 0.31a version. I have, like you, Ryzen 1600, but on 3600mhz

I prefer manual config and decrease number of working core - best when 6, only new Bittube - 4.

So, my results ( not very long test )

CN7 ~ 490 h/s

CNlite ~ 1780 h/s

newBittube ~ 180 h/s

Don`t test heavy - no wallet for it.

JCE-Miner, what hash you have at CN7, light and heavy ?  

Ppl, give me pool settings and wallet to test CNheavy algo