Custom RAM Timings for GPU's with GDDR5 - DOWNLOAD LINKS - UPDATED

[ray_saeed]

I am interested in purchasing custom Elpida and Samsung roms for XMR mining with  Nitro+ RX 470 OC cards (4GB and 8GB). If you have stable custom roms available please message me with a quote. Thanks.

dude,for free i will help you out.im mining with xfx rx470 4gb black's with elpida. all 24 at doing 28.3mhs on eth and 110ish watts each.

Thanks for the offer buddy .  I am getting around 28 mhs in ETH and getting around 700 h/s in XMR. These cards are undervolted, with clocks @ 1100/1950 and 1500 straps. Was reading about some folks getting upto 900 h/s while mining monero and was interested in getting similar results.

[1715588856]

1715588856

[1715588856]

1715588856

[avirdoz]

I am interested in purchasing custom Elpida and Samsung roms for XMR mining with  Nitro+ RX 470 OC cards (4GB and 8GB). If you have stable custom roms available please message me with a quote. Thanks.

dude,for free i will help you out.im mining with xfx rx470 4gb black's with elpida. all 24 at doing 28.3mhs on eth and 110ish watts each.

Thanks for the offer buddy .  I am getting around 28 mhs in ETH and getting around 700 h/s in XMR. These cards are undervolted, with clocks @ 1100/1950 and 1500 straps. Was reading about some folks getting upto 900 h/s while mining monero and was interested in getting similar results.

900H/s is a vanity hashrate - it's absolute max and not efficient. 825 - 875 is probably real.

What about eth h/r with this custom timings?

[LoneRangir]

Do you visualize or simulate your timing parameter changes, or make them and test them live?

[ray_saeed]

I am interested in purchasing custom Elpida and Samsung roms for XMR mining with  Nitro+ RX 470 OC cards (4GB and 8GB). If you have stable custom roms available please message me with a quote. Thanks.

dude,for free i will help you out.im mining with xfx rx470 4gb black's with elpida. all 24 at doing 28.3mhs on eth and 110ish watts each.

Thanks for the offer buddy .  I am getting around 28 mhs in ETH and getting around 700 h/s in XMR. These cards are undervolted, with clocks @ 1100/1950 and 1500 straps. Was reading about some folks getting upto 900 h/s while mining monero and was interested in getting similar results.

900H/s is a vanity hashrate - it's absolute max and not efficient. 825 - 875 is probably real.

I would be willing to buy in that range as long as it is stable.

[xneoenx]

Do you visualize or simulate your timing parameter changes, or make them and test them live?

I usually make them and test them live, but I know some depend on the values of others, so I often adjust for that.

God damn it. I have been flashing at least 15+ times trying all values, where the hell is RAS2RAS Jesus god damn it. Each flash takes like 6 minutes, Flash => Restart => Test @ 2000 => Restart => Test @ 2100. On and on for the past 5 days now. Best part is, each HEX string I do it on paper. FUN! Best I have is 800, at 1160, 2120. But then again, based only on if it runs without crashes. Trying to get those ones first and then work backwards to see how it wokrs as I dont think I know which of these timings even matter, or if all do.

Trying to go about it theoretically without knowing how the algo works, is like finding a needle in a haystack,  thats somewhere in the ocean.

I think ETH is more dependant on speed vs timings no? Im sure we can optimize it, but the pay off wont be much I dont think. 1 - 4% gains? XMR is where the money is 20% gains for a good strap.

[Metroid]

900H/s is a vanity hashrate - it's absolute max and not efficient. 825 - 875 is probably real.

I have been playing with Elpida (EDW4032BAB) 0.85v memory and clock for a day or so, so far I reached 820 h/s on XMR, It needs a lot of work yet, low share difficult rejection rate is around 10% which is bad, memory error is low, need time to improve, it has been fun, after this will be time to play with Hynix.

[Eliovp]

900H/s is a vanity hashrate - it's absolute max and not efficient. 825 - 875 is probably real.

I have been playing with Elpida (EDW4032BAB) 0.85v memory and clock for a day or so, so far I reached 820 h/s on XMR, It needs a lot of work yet, low share difficult rejection rate is around 10% which is bad, memory error is low, need time to improve, it has been fun, after this will be time to play with Hynix.

Got an Elpida 470 4G, best card i've got, does 850h xmr, 30eth, 280zcash without a hickup.

I always thought Hynix would be the better performer but i was wrong. (i'm talking about 470 Hynix, 480's do a shitload better, 850 without a problem..)
And there's a serious difference in Hynix quality.

If you have Elpida's, be glad, they are in my opinion the best ones out there!

Samsung is very decent as well, got my 480 8G's at 800 (can easily do 840-850 as well) core clock 1100, mem 2050 and an undervolt of -120mV vddc.

So in the 80 watt range

[Metroid]

900H/s is a vanity hashrate - it's absolute max and not efficient. 825 - 875 is probably real.

I have been playing with Elpida (EDW4032BAB) 0.85v memory and clock for a day or so, so far I reached 820 h/s on XMR, It needs a lot of work yet, low share difficult rejection rate is around 10% which is bad, memory error is low, need time to improve, it has been fun, after this will be time to play with Hynix.

Got an Elpida 470 4G, best card i've got, does 850h xmr, 30eth, 280zcash without a hickup.

I always thought Hynix would be the better performer but i was wrong. (i'm talking about 470 Hynix, 480's do a shitload better, 850 without a problem..)
And there's a serious difference in Hynix quality.

If you have Elpida's, be glad, they are in my opinion the best ones out there!

Samsung is very decent as well, got my 480 8G's at 800 (can easily do 840-850 as well) core clock 1100, mem 2050 and an undervolt of -120mV vddc.

So in the 80 watt range

Good to hear, there a lot things to test, I have been busy today and probably it will take me few days to get back testing. Yeah on XMR at 0.85v 90 watts, on eth 110 watts. Electricity is a big deal, always keen on performance per watt more than anything and custom timings can help with my quest.

[niko2004x]

Some experimental results.
It seems ~900h/s is a practical limit for Ubuntu 15.04 + fglrx + claymore xmr miner + r9 390 (with H5GC4H24AJR)
at core clock 1135 and any mem clock begining from 1100.
900h/s is relatively easy to obtain with timings tweaking with almost any mem clock starting from 1100,
above 900h/s i had no success.

[dallase]

900H/s is a vanity hashrate - it's absolute max and not efficient. 825 - 875 is probably real.

I have been playing with Elpida (EDW4032BAB) 0.85v memory and clock for a day or so, so far I reached 820 h/s on XMR, It needs a lot of work yet, low share difficult rejection rate is around 10% which is bad, memory error is low, need time to improve, it has been fun, after this will be time to play with Hynix.

Got an Elpida 470 4G, best card i've got, does 850h xmr, 30eth, 280zcash without a hickup.

I always thought Hynix would be the better performer but i was wrong. (i'm talking about 470 Hynix, 480's do a shitload better, 850 without a problem..)
And there's a serious difference in Hynix quality.

If you have Elpida's, be glad, they are in my opinion the best ones out there!

Samsung is very decent as well, got my 480 8G's at 800 (can easily do 840-850 as well) core clock 1100, mem 2050 and an undervolt of -120mV vddc.

So in the 80 watt range

Is this claymore 9.7 windows?   I have a elpida running 1100/2050 and its barely over 700 on sgminer.   

[Eliovp]

900H/s is a vanity hashrate - it's absolute max and not efficient. 825 - 875 is probably real.

I have been playing with Elpida (EDW4032BAB) 0.85v memory and clock for a day or so, so far I reached 820 h/s on XMR, It needs a lot of work yet, low share difficult rejection rate is around 10% which is bad, memory error is low, need time to improve, it has been fun, after this will be time to play with Hynix.

Got an Elpida 470 4G, best card i've got, does 850h xmr, 30eth, 280zcash without a hickup.

I always thought Hynix would be the better performer but i was wrong. (i'm talking about 470 Hynix, 480's do a shitload better, 850 without a problem..)
And there's a serious difference in Hynix quality.

If you have Elpida's, be glad, they are in my opinion the best ones out there!

Samsung is very decent as well, got my 480 8G's at 800 (can easily do 840-850 as well) core clock 1100, mem 2050 and an undervolt of -120mV vddc.

So in the 80 watt range

Is this claymore 9.7 windows?   I have a elpida running 1100/2050 and its barely over 700 on sgminer.   

Nope - he doesn't run Win anymore, I don't think, and I think it's on sgminer.

That's correct,

In other words, run it on windows and you'll most likely get better results..

Greetings!

[bughatti]

anyone in this thread working on a good rom for the sapphire 480 nitro + 4gb elpida that is steady?

[Eliovp]

anyone in this thread working on a good rom for the sapphire 480 nitro + 4gb elpida that is steady?

Got one for that card yeah.

Performance is as you expect it to be.

Awesome 

:p

[niko2004x]

When trying to test new timings on rx 480 (with Samsung K4G80325FB) encountered following unexpected (at least for me)
behavior using Ubuntu 16.04 with amdgpu drivers from stock 4.9.0-rc3 kernel.
If timings in 2000 strap is too tight on rx 480 then system hard crash at boot with 'Fatal error during GPU init mining' error.
Happens at driver loading stage before major os components startup so hardware reset required.

But too tight timings in 2000 strap on rx 470 (with Hynix H5GC4H24AJR) do not lead to crash at boot.
Card is visible by lspci but do not work at 2000 mclk as expected.

WFT?

[doktor83]

My custom timings for 470 8GB Nitro+ OC (11256-02-20G) - Samsung mem

For the powersaver guy :
(1126/2100 @ -120mV)




For normal usage guy :
(1300/2100 @ -100mV)

[Eliovp]

My custom timings for 470 8GB Nitro+ OC (11256-02-20G) - Samsung mem

For the powersaver guy :
(1126/2100 @ -120mV)

https://i.imgur.com/ILSqPVT.jpg


For normal usage guy :
(1300/2100 @ -100mV)

https://i.imgur.com/g4AZZqz.jpg

Congrats, you have a card that won the sillicon lottery...

You do realize that 98 out of 100 cards like that can't handle -100mV while running 1300 core??

Try this on a 6xGPU rig.. and while NOT using claymore.. if you can do that, i will say: Congratulations to you sir.

[doktor83]

My custom timings for 470 8GB Nitro+ OC (11256-02-20G) - Samsung mem

For the powersaver guy :
(1126/2100 @ -120mV)

https://i.imgur.com/ILSqPVT.jpg


For normal usage guy :
(1300/2100 @ -100mV)

https://i.imgur.com/g4AZZqz.jpg

Congrats, you have a card that won the sillicon lottery...

You do realize that 98 out of 100 cards like that can't handle -100mV while running 1300 core??

Try this on a 6xGPU rig.. and while NOT using claymore.. if you can do that, i will say: Congratulations to you sir.

Why do you think that 1300 is much when the def clock is 1266 ?  

Also why not claymore ? Not enough power hungry? Better sgminer ? Why?

Also im running a 5xGPU rig with -120mV for weeks now on ETH (145mhs)...

[avirdoz]

My custom timings for 470 8GB Nitro+ OC (11256-02-20G) - Samsung mem

For the powersaver guy :
(1126/2100 @ -120mV)




For normal usage guy :
(1300/2100 @ -100mV)

Did you check what power usage from the wall with "powersaver guy"?

[doktor83]

I checked it now for you:

Idle : 105w from wall
Working : Hash ~ 807 H/s, VDDC : 0.8938 V, 197w from wall

[avirdoz]

I checked it now for you:

Idle : 105w from wall
Working : Hash ~ 807 H/s, VDDC : 0.8938 V, 197w from wall

Thanks and nice results.