Topic: RX 470 = 30Mhs ++ GPUz 80W (Read 2431 times)

I'm happy with mine doing 27 MH  from time to time i like to mess around at first but usually after i find the Sweet spot i stay with it. btw the 28 Mh test lasted two days then it crashed so i went back to 27 MH which is what the sweet spot seems to be .



thanks toptek

So I dove into Hex editing for the BIOS and its not as complicated as its made out to be. For the layman, I still feel what watttool offers is pleanty. With HEX what you can additionally do is create intermediate memory timing straps and can include an offset in the bios to reduce core volts consumption.
With all the tweking I could do, the BEST I can get my RX 470's is to hash at 25.6 Mh/s while consuming 90W from the wall (GPUZ is like 54W), with a 5 card system at 510W with a lot of bios editing in hex (not polaris). This was a part of my pet 250Mh/s with 1KW at the wall single frame rig/project that I finally achieved yesterday. Its got two 5 card systems in one compact open frame. I will be setting up my 4th such rig to touch 1Gh/s soon.
I have a power ceiling at my place that limits how many rigs I can have running, so most of my research and efforts have been focused into efficiency and not peak performance. Also, the cards run cool, have 0 memory errors, are significantly undervolted and most importantly the risers run cool so i'm insuring a good long run for my hardware.

I think we all know by now that these cards can be pushed pretty hard. But unless you have a proper cooling setup with ducts and air conditioners, and unless you are sure that you are going to dump these cards when ETH goes POS, I would not recommend pushing them to the edge. Find a suitable balance and stick to it, even for perf enthusiasts 27 Mh/s is what you should be happy at with a RX 470 4GB running 24/7. I've made my peace with less.

For me if a crash happens, then it's usually a hangup and the PC does not reboot automatically, and what is more it's continously drawing 350W while hung.

This can be cured by having USB watchdogs. I already ordered the arduinos to make some... not that my rigs are not stable currently.

This is unclear.
Let's suppose you have 150MH on your rig and by running "unstable" you get additional 3MH total. If you have ssd and miner on autostart it would take say 3 minutes until it's up and hashing.
Five restarts a day make for 15min*150MH = 2250 MHm of lost hashes
so even in this case you still get positive result: 60*24*3 -15*153 = 2025MHm which is still almost 1% increase of total output

What is your total number of shares for 30mh?

(incorrect shares divided by total shares) * 100
gives you the waste rate in %.

for 29 MH i had 0 incorrect shares in 24 hours but 30 mh i had about 10 incorrect shares a day .. i wonder how much would that effect the end result.

There back to default clocks with the mode i redid @ 27 mh and Ming a way.
I need those cards to much right now to burn one out .

http://imgur.com/1sGfiPU  or clocked it to 28 MH gonna see if that holds ok for a while with one card if it does I''ll put my other 3 at 28 MH , 29 MH was a nite mare for me .



Great, if the OP does get it to run stable @ 30 mh.


Thanks top.

heh, 3 mins ... don't kill your card, let it run where it's stable & MORE importantly don't suggest or show the public something which is not useful or cause harm or damages to others.

just my 2 cents, flame if you want.

although if it's "stable" for 24 or 48 or even a week, did you ever look at how MUCH memory errors it's generating ? prolly higher than bitcoin current diff now.

post a screenshot of the memory errors heh...

peace out.

i'd suggest to fine tune the gpu to the lowest possible memory errors with lowest power & you WILL most likely end up around 27-28 mhs STABLE

I had mine doing 29 MH by following this video but it wasn't for long mine got really unstable at that  speed or mine were . mine are 4 GB  Cards .

I'm sure i Probably could have tweaked it enough to hold at 29 MH or higher if i tried .

https://www.youtube.com/watch?v=uy0roLG7ARc  got bored real fast tiring to many crashes .

Yeah ... will let you guys know .. btw i am running these two cards directly off the mobo and no raiser.. this is my pet project.. my farms are on a safe 28.5mhs.

its stack on 1050.

What voltage is your Memory Voltage set too? 1000 ?

24-48 hours are not enough some times, i remember a case were the crash occured after 3 days, you need to run for one week to be 100% sure

and btw gpuz is incorrect, it's more about 100w not 83

9th hour and yet no reboot nor crashes.. Incorrect of shares is at 10. Should be acceptable running at 30mhs constantly..

5 hours in and has 2 incorrect shares not reboot nor gc crash ... if this works all my cards are changing

Sometimes you can get lucky with the memory chips, 2100 mhz is a really high OC.

You might be able to even push it to 2200 mhz, but Im not so sure what effects long term that will be.

Push "S" and see if you have any incorrect shares

Agree with you ... will run it for 24 hours and will see if it crashes.. so far 2 hours in ... still standing great.

Nope... its a 4GB card

8gb correct?