rumours 3070ti to have 320 bit gddr6x 8gb (like 3080) and 3080 super to have 384 bit gddr6x (like 3090). Any thoughts?
|
|
|
should be 54.9mhs 1100/1026@787mv @ 112watts ...
Can you please explain how do you go below 800 mV? with linux and atitool Which miner, what memtweak settings? I have 6xVega 56 sapphire REF. 56 bios, and max. is 50mhs. No run above 1070cc and /1027mc/mem 820v. My Vega56 is on : 1090 core 1020 mem 0.812 v Gets me 53.19mhs on mmpos non modded bios. it seems that mmpos uses amdmemtweak timings by default ))
|
|
|
Usually amd gpus need much less core speed to achieve what nvidia achieves, reason you can undervolt a lot more than nvidia gpus.
modern nvidia 3060ti and 3070 need only 800-900 mhz gpu clock to hash 60-64mhs. Difficult to beat ))
|
|
|
should be 54.9mhs 1100/1026@787mv @ 112watts ...
Can you please explain how do you go below 800 mV? with linux and atitool
|
|
|
Your Core and Memory Control are way out of desired values. Try P7 1150/850 and P3 940/825 first, than try to play with timings.
EDIT: Huh, also decrease temp limits or upper fan curve keeping memory temp in 60-65 degrees range.
Wow, the P7 and P3 settings pushed it to 42 MH/s! Thanks a lot! I have no idea what to try in timings, as i dont understand the technical background. Any suggestions? 42... Non modded vega56 on mmpos....53.19mhs core 1090 / mem 1020 / vlt 0.835 130watts. should be 54.9mhs 1100/1026@787mv @ 112watts ...
|
|
|
can i post my msi ventis 2 oc rtx 3070 results while i'm waiting my Post to bring me new 3060ti? 3070 is really good! 63.5 mhs @ 125watt. Win, msi afterburner and its Curve Editor. Rocks! 1155/8250@743mv. Old good claymore. 1 mistaken solution overnight. Number of sent good shares is good enough comparing to vegas and TRM running simultaneously. Well. I gave nbminer a try. And it uses 119watt instead of claymore's 125 with the same speed. cool!
|
|
|
can i post my msi ventis 2 oc rtx 3070 results while i'm waiting my Post to bring me new 3060ti? 3070 is really good! 63.5 mhs @ 125watt. Win, msi afterburner and its Curve Editor. Rocks! 1155/8250@743mv. Old good claymore. 1 mistaken solution overnight. Number of sent good shares is good enough comparing to vegas and TRM running simultaneously.
|
|
|
vega56 or vega64 flashed to 56 can do 51.8mhs with 104watt (driver shown) or 55mhs with 114watt. In fact it will take more juice of course but that makes it rather good even today
|
|
|
i dont understand 1 thing. mining is all about compute power right? 6900 has twice more compute power than 5700 then.it suppose to be at least 100mhs. i dont understand it.
Minins ETH is very dependant to memory bandwith and memory latency... The 6800/6900 have a bandwith just around 10% more than 5700. The remaining question if the impact of the infinity cache: due to its small size relatively to the DAG size of ETH, it should have a minimal impact on ETH mining. ok then why radeon 7 has : 1,024 GB/s bandwith performing worse than 3080 which has less memory bandwith: 760.3 GB/s . Both bandwith and core are used for ETH: for radeon VII, it is probably limited by core computation, for RX 6900 xt, it is mostly limited by bandwith. For RTX 3080: you should downclock the core at around 1200-1300 mhz (to decrease power consumption almost without losing hashrate) and overclock memory at +1000mhz (around equivalent 10000mhz or more) to have the highest hashrate. Therefore you can conclude that this card is more bandwith limited for ETH mining than compute limited. RX 5700 is well balanced for ETH mining between compute power and memory bandwith. hehe could very nice card if there was 3080 chip with radeon 7 memory very nice if there was 6800 chip with radeon 7 memory )) And it exists ... but only as special very expensive compute card
|
|
|
If amd would made navi2 2ghz+ gpu with hbm2 4096 bit wide mem like vii has it would never go to gamers. So, they make pro compute card instead and limited mem bandw for gaming card ...
|
|
|
These can also take advantage of the pipeline improvements by generating multiple DAGs (up to 3 with 16GB GDDR6) that was only possible on the Radeon VII before (see how teamredminer does it). Should offer a couple more %.
Radeon VII still cannot effectively address over 8gb memory with current (and previous) adrenalin (gaming) drivers. Only with some kind of "pro" drivers it can be done (and vega frontier too). But this "pro" drivers suffer from lack of hardware tweaking support that way making them unusable for mining.
|
|
|
I expect them to hash around 63-65 MHs at stock memory frequencies.
no way )) only with oc and timings ))
|
|
|
well, after 2 years nvidia was able to reproduce Radeon VII eth performance with more wattage
|
|
|
from 4100 to 4300h/s on ryzen 5 1600 @ 3600mhz (3200 cl16 ram)
|
|
|
Just to drop a number: With current C31 code Vega 64 does 1.1 g/s at 180 W (settings: 1380 / 1000 @ 0.912V) making it the 2nd most profitable algorithm after CNR for V64 (!!!).
You are totally wrong with your calculation. Vega64 can easily make 50+mhs eth with lower than 180w power consumption. ))
|
|
|
Hi, dok! I was trying to use your new miner for Loki (randomxl) with just my ryzen 1600. gpu mining was disabled. It gives good 3900h/s instead of xmrig 3300-3500 I was using before. But suddenly I discovered some hashrate lose to only 30h/s. Look on pool graph Income become lower too (comparing with xmrig I was using before). Here is my config SRBMiner-MULTI.exe --disable-gpu --cpu-threads 6 --cpu-affinity 0x555 --config-file Config\config-loki.txt --pools-file Pools\pools-loki.txt --log-file Logs\log-loki.txt Al other is std. I can provide log. What can be the cause of the problem?
|
|
|
@claymore Man i can't wait for you to update something on the XT. On the edge of our seats man!
New version with Navi cards support is ready, I will release it when AMD releases DAG-fix drivers, it will be soon. Don’t you know are they going to release driver with Radeon 7 over 8gb slow memory access fix? It will help a lot fir some memory-heavy algos.
|
|
|
nice result but not reasonable ))
|
|
|
Ryzen 5 2600 running at 3,4GHz
randomx-benchmark.exe --mine --jit --largePages --threads 6 --init 6 --nonces 300000
Performance: 3630 h/s
test bench: Asus TUF B450 Memory 2 x 4Gb @ 2133 MHz
Sounds about fight! If possible would be nice IF you could redo benchmark with just one stick of The ram. Actual/real hashrate R5 2600 (no o/c) on the algorithm of RandomX (rx/Loki) is 3000 +/-10 h/s. Mine started right after the Loki fork and change the algorithm on RandomX. I ask the others to confirm the hashrate. P.S. Miner - xmrig 2.99.0 ryzen 5 1600 @ 3.6ghz with 3200ddr4 16-17-17-1T is doing 3800h/s rx/Loki. (asus x370-f gaming)
|
|
|
|