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

[rednoW]

Yes.  Here's the stock Hynix H5GC4 1625 strap:
999000000000000022559D0010DE7B4480551312B78C450A004C0601750414206A8900A00200312 018112D34A42A3816

And here's what I'm running at 2025 (stable after >24H):
999000000000000022559D0010DE5B4480551312B74C450A00400600750414206A8900A00200312 010112D34A42A3816

What speed do you get on rx470 with Hynix with 1625_modded strap on 2025?

Now i'm using 1375 strap on my Sapphire ref-like rx470 4gb with hynix, 1140/1950 28.4mhs (rock-stable)
(Claymore under windows)

[madLyfe]

your stock 1625 strap ran through the CGI gets you over 2000 mem?

Yes.  Here's the stock Hynix H5GC4 1625 strap:
999000000000000022559D0010DE7B4480551312B78C450A004C0601750414206A8900A00200312 018112D34A42A3816

And here's what I'm running at 2025 (stable after >24H):
999000000000000022559D0010DE5B4480551312B74C450A00400600750414206A8900A00200312 010112D34A42A3816

total newb question. when i first started playing with straps the common standard mod was to take the 1500 strap and apply it to 1625, 1750, and 2000. when you take and edit the 1625 strap, do you apply it to only 1625, to 1625, 1750, and 2000, or do you apply it to all of them, 1500, 1625, 1750, and 2000?

[bardacuda]

total newb question. when i first started playing with straps the common standard mod was to take the 1500 strap and apply it to 1625, 1750, and 2000. when you take and edit the 1625 strap, do you apply it to only 1625, to 1625, 1750, and 2000, or do you apply it to all of them, 1500, 1625, 1750, and 2000?

That should depend on what mem clock you can achieve. You basically want the tightest timings that will still work at the highest mem clock. If your mem can do 2100 and the 1500 strap will only let you get to 1900...well you might want to use the 1625 strap instead. As always trial and error and performance/stability testing will find out what works best for your hardware. Every card is different.

You only really need to apply a strap to the one frequency you will actually be using. Applying it to multiple ones will just give you some lower clocks to fall back on if a higher clock fails.

[madLyfe]

total newb question. when i first started playing with straps the common standard mod was to take the 1500 strap and apply it to 1625, 1750, and 2000. when you take and edit the 1625 strap, do you apply it to only 1625, to 1625, 1750, and 2000, or do you apply it to all of them, 1500, 1625, 1750, and 2000?

That should depend on what mem clock you can achieve. You basically want the tightest timings that will still work at the highest mem clock. If your mem can do 2100 and the 1500 strap will only let you get to 1900...well you might want to use the 1625 strap instead. As always trial and error and performance/stability testing will find out what works best for your hardware. Every card is different.

You only really need to apply a strap to the one frequency you will actually be using. Applying it to multiple ones will just give you some lower clocks to fall back on if a higher clock fails.

so i should be applying the custom 1625 strap to the 2000 field because im shooting for above 2000 mem clock?

[bardacuda]

If it's the tightest one that will run stable at that speed, yes. For Polaris BIOS modding 101 try this forum thread:

http://www.overclock.net/t/1604567/polaris-bios-editing-rx480-rx470-rx460

It's probably better to leave this thread for nerdralph, Wolf0, and the others for their technical discussions. I'm going to take my own advice too and revert to lurking now

EDIT: Ahhh I'm not well-versed in Polaris BIOS modding but if it's anything like Hawaii, the 2000 field applies to mem clocks between 1875 and 2000. You would want to apply to the 2125 field (if there even is one) for clocks between 2000 and 2125. yes the 2000 field (40 0D 03) applies to any clocks above 1751. Anyhow all that info can be found in the link.

[nerdralph]

Yes.  Here's the stock Hynix H5GC4 1625 strap:
999000000000000022559D0010DE7B4480551312B78C450A004C0601750414206A8900A00200312 018112D34A42A3816

And here's what I'm running at 2025 (stable after >24H):
999000000000000022559D0010DE5B4480551312B74C450A00400600750414206A8900A00200312 010112D34A42A3816

What speed do you get on rx470 with Hynix with 1625_modded strap on 2025?

Now i'm using 1375 strap on my Sapphire ref-like rx470 4gb with hynix, 1140/1950 28.4mhs (rock-stable)
(Claymore under windows)

I get 28.2-28.3 with sgminer (kernel 4.10.5/AMDGPU-Pro 16.60).  With Claymore ~28.9Mh.  Sgminer should be just as fast once I finish tuning it (or worst case wait for Wolf's Polaris-tuned ethash kernel).
With the ref-style Sapphire should should be able to get over 2000, maybe even 2100 if you mod the 1500 strap.

[lexele]

Yes.  Here's the stock Hynix H5GC4 1625 strap:
999000000000000022559D0010DE7B4480551312B78C450A004C0601750414206A8900A00200312 018112D34A42A3816

And here's what I'm running at 2025 (stable after >24H):
999000000000000022559D0010DE5B4480551312B74C450A00400600750414206A8900A00200312 010112D34A42A3816

What speed do you get on rx470 with Hynix with 1625_modded strap on 2025?

Now i'm using 1375 strap on my Sapphire ref-like rx470 4gb with hynix, 1140/1950 28.4mhs (rock-stable)
(Claymore under windows)

I get 28.2-28.3 with sgminer (kernel 4.10.5/AMDGPU-Pro 16.60).  With Claymore ~28.9Mh.  Sgminer should be just as fast once I finish tuning it (or worst case wait for Wolf's Polaris-tuned ethash kernel).
With the ref-style Sapphire should should be able to get over 2000, maybe even 2100 if you mod the 1500 strap.

With the first 1625 strap you posted a few days ago, I get 29.7 core1150/mem1950 on rx480 4g hynix.
the strange thing, is this number don't change if I change memclock up to 2100... It's like the memclock is bloqued on 2050 and what's showed on screen is a lie!
(I failed to post screenshot)

[doktor83]

OhGodADecode.exe 555000000000000022CC1C00CE616C47D0570F15B48C250B006AE7000B031420FA8900A00300000 0190F2F39B22D3517

Its for Polaris cards

[laik2]

Yes.  Here's the stock Hynix H5GC4 1625 strap:
999000000000000022559D0010DE7B4480551312B78C450A004C0601750414206A8900A00200312 018112D34A42A3816

And here's what I'm running at 2025 (stable after >24H):
999000000000000022559D0010DE5B4480551312B74C450A00400600750414206A8900A00200312 010112D34A42A3816

What speed do you get on rx470 with Hynix with 1625_modded strap on 2025?

Now i'm using 1375 strap on my Sapphire ref-like rx470 4gb with hynix, 1140/1950 28.4mhs (rock-stable)
(Claymore under windows)

I get 28.2-28.3 with sgminer (kernel 4.10.5/AMDGPU-Pro 16.60).  With Claymore ~28.9Mh.  Sgminer should be just as fast once I finish tuning it (or worst case wait for Wolf's Polaris-tuned ethash kernel).
With the ref-style Sapphire should should be able to get over 2000, maybe even 2100 if you mod the 1500 strap.

You will have to wait a lot, it wont get public

[rednoW]

I get 28.2-28.3 with sgminer (kernel 4.10.5/AMDGPU-Pro 16.60).  With Claymore ~28.9Mh.  Sgminer should be just as fast once I finish tuning it (or worst case wait for Wolf's Polaris-tuned ethash kernel).
With the ref-style Sapphire should should be able to get over 2000, maybe even 2100 if you mod the 1500 strap.

Thanx again! It seems to work 2140/2025 (0.9v/0.95v), gives 29mhs on claymore and some mem errors in hwinfo64 ((
1500 strap modded with your online cgi tool ))
Wonder if someone tried Uber-mix extreme v2 timing from overclock.net ?
It is for samsung 8gb a guess
~236GB/s (~214GB/s Stock) +10% (OclMemBench)
1375 - 1500 - 1625 Mix
No EDC Errors Mem@2100Mhz (1000mv) / Core @1350Mhz
777000000000000022CC1C00AD615B41C0570E152DCB74090068C7000B031420FA8900A00300000 01E123A46DB354019

[doktor83]

Can someone explain me something about usVddcOffsets ?
Reading hex from .rom must be done in little endian, so i find a vddc offset, lets say ffe6, that is 65510 in decimal. Well ok, that is not -26 that i taught it should be.
But ffe6 in big endian INT16 is -26.
If that is so, it means data in .rom is mixed, so how can i know which data is in little endian, which one is in big ?
Or am i even getting this right ?

[cryptominer420]

For the lazy fools out there, I whipped up a cgi version of my strap mod utility.

Code:

http://doreymills.ca/cgi-bin/strapmod?555000000000000022CC1C00AD515A3EC0570E142D4A64080048C700030114207A8900A003000000150E2A3186272E16

Well it works, got my hynix cards up-to 28.3MH ETH and my samsung cards up-to 31.5MH Eth all 4GB powercolor cards.

Also to all the people making tools for us here it would be nice to have donation addresses so that we can send you some tips.

Are you using the 1500 strap for hynix and 1750 for samsung? and what mem clocks did you use to get those (specifically the samsung)?

For me I use the above straps and clock hynix at 1900-1925 (1950 if its stable) and 2050-2100 for samsung (although Im wondering if i can go higher than 2100 on some samsung chips)

MY samsungs are running at 2120-2160 mem set per card for stability.

[Truthchanter]

I used this tool:

0150E2A3186272E16">http://doreymills.ca/cgi-bin/strapmod?555000000000000022CC1C00AD515A3EC0570E142D4A64080049C700030114207A8900A00300000 0150E2A3186272E16

(not that strap) to modify a 480 with samsung mem, 1750, strap, and it seems to max mhs at 2100 mem clock. I can clock it to 2125 but the mhs stays about the same and if I clock it higher than 2125, the mhs drops.. any idea why? I was hoping to get my clock closer to 2200

[laik2]

I used this tool:

0150E2A3186272E16">http://doreymills.ca/cgi-bin/strapmod?555000000000000022CC1C00AD515A3EC0570E142D4A64080049C700030114207A8900A00300000 0150E2A3186272E16

(not that strap) to modify a 480 with samsung mem, 1750, strap, and it seems to max mhs at 2100 mem clock. I can clock it to 2125 but the mhs stays about the same and if I clock it higher than 2125, the mhs drops.. any idea why? I was hoping to get my clock closer to 2200

It's because the tool is meant to change only "some" values that can give you what you see, to do more others must be adjusted properly, and that my friend is not that easy and is related to many tests and on particular GPU, because not all are the same.

[nerdralph]

I used this tool:

0150E2A3186272E16">http://doreymills.ca/cgi-bin/strapmod?555000000000000022CC1C00AD515A3EC0570E142D4A64080049C700030114207A8900A00300000 0150E2A3186272E16

(not that strap) to modify a 480 with samsung mem, 1750, strap, and it seems to max mhs at 2100 mem clock. I can clock it to 2125 but the mhs stays about the same and if I clock it higher than 2125, the mhs drops.. any idea why? I was hoping to get my clock closer to 2200

It's because the tool is meant to change only "some" values that can give you what you see, to do more others must be adjusted properly, and that my friend is not that easy and is related to many tests and on particular GPU, because not all are the same.

And if your RAM isn't cooled, you can start getting errors due to overheating rather than overclocking.

[Truthchanter]

That is fair enough.. 2100 is low enough to potentially work on most gpus and be stable. and with the custom mod tool it gives great mhs anyway so all is well. Thanks for the answers!

On another note I used to use 1500 strap with hynix (and now using the 1500 strap custom modded) but am now testing hynix mem with the
custom modded 1625 mem strap as it seems it may be slightly better mhs (figuring in 1625 mem strap can work with mem clocked a bit higher than 1500) What do you all think?

[bensam1231]

Curiously, how well do custom timings play with other algos? Is this just a Ethereum only thing (or dual mining)?

What happens if you try to run cards with custom timings on something like Equihash?

[Truthchanter]

Curiously, how well do custom timings play with other algos? Is this just a Ethereum only thing (or dual mining)?

What happens if you try to run cards with custom timings on something like Equihash?

Using my eth timings for zec worked fine. However there may be more optimum timings for zec or xmr compared to the eth ones frequently discussed.

[bensam1231]

Curiously, how well do custom timings play with other algos? Is this just a Ethereum only thing (or dual mining)?

What happens if you try to run cards with custom timings on something like Equihash?

Using my eth timings for zec worked fine. However there may be more optimum timings for zec or xmr compared to the eth ones frequently discussed.

Yeah, just discussing from a stability and performance standpoint (such as you end up with worse performance then normal with the new straps) while mining altcoins. I'm not a big fan of getting shoehorned into a corner.

I assume if it's a stability issue you just back off the memory clock speed to the point that the timings are actually meant for or is it even a issue?

[Walrusbonzo]

Thanks everyone for the shared info, tools and shared straps. I've settled on using one of nerdralph's and I've gained nearly 1mh/s per card, 28.6~ to 29.6~ using 1125/2050 clocks.