avalon Chip Frequency could run over 300Mh

[dylan186]

somebody recompilation avalon drivers and run avalon Frequency 332Mh,it's hash ability about 80G.

[TheSwede75]

somebody recompilation avalon drivers and run avalon Frequency 332Mh,it's hash ability about 80G.

Is this a question, a statement or what?

[Garr255]

It is entirely doable

I will be selling boards that run the Avalon chips at 350mh/s. Check my signature link!

[GenTarkin]

It is entirely doable

I will be selling boards that run the Avalon chips at 350mh/s. Check my signature link!

Would it be software to use someone like yours firmware & flash them to our actual avy units?

[Garr255]

It is entirely doable

I will be selling boards that run the Avalon chips at 350mh/s. Check my signature link!

Would it be software to use someone like yours firmware & flash them to our actual avy units?

Technically this is possible, but certainly not advisable because the Avalon spec boards are not designed to handle the boost in power draw.

[goxed]

somebody recompilation avalon drivers and run avalon Frequency 332Mh,it's hash ability about 80G.

Question: what should be buf[6] and buf[7] for 325MHz, 350MHz, and 375MHz?



   if (frequency == 256) {
      buf[6] = 0x03;
      buf[7] = 0x08;
   } else if (frequency == 270) {
      buf[6] = 0x73;
      buf[7] = 0x08;
   } else if (frequency == 282) {
      buf[6] = 0xd3;
      buf[7] = 0x08;
   } else if (frequency == 300) {
      buf[6] = 0x63;
      buf[7] = 0x09;
   }

[OnkelPaul]

If the frequency division logic does not do some weird things, this code should work for any frequency (as far as the hardware supports it):
buf[6] = (int)(frequency * 8.01 + 0.5) % 256;
buf[7] = (int)(frequency * 8.01 + 0.5) / 256;

The formula yields exactly the same values for your given frequencies, so you can just drop the cascade of if statements and replace it by those two lines.
Caveat: I did not try it, don't have an ASIC, if you blow up your hardware it's your fault and nobody else's.

(see below, this formula might generate wrong values)

Onkel Paul

[-ck]

somebody recompilation avalon drivers and run avalon Frequency 332Mh,it's hash ability about 80G.

Question: what should be buf[6] and buf[7] for 325MHz, 350MHz, and 375MHz?



   if (frequency == 256) {
      buf[6] = 0x03;
      buf[7] = 0x08;
   } else if (frequency == 270) {
      buf[6] = 0x73;
      buf[7] = 0x08;
   } else if (frequency == 282) {
      buf[6] = 0xd3;
      buf[7] = 0x08;
   } else if (frequency == 300) {
      buf[6] = 0x63;
      buf[7] = 0x09;
   }

It was explained to me that only certain values are actually accepted by the hardware. However the values are simply little endian values with a fairly precise linear relationship with the corresponding frequency so it seems it's a simple multiplier.

0x0803 = 2051 = 256 * 8
0x0873 = 2163 = 270 * 8
and so on
so if you want to try 325 etc it would be

325 * 8 = 0x0A28 so it would be buf[6]=0x28 buf[7]=0x0a
350 buf6=0xf0 buf7=0x0a
375 buf6=0xb8 buf7=0x0b

but bear in mind the power supply probably won't take that, and I have no idea what the other components would do, nor if it will actually accept those values.

[OnkelPaul]

ckolivas is probably right about the multiplier being 8, then the lower 2 (or 3) bits of buf[6] could be flag values and should probably be kept at binary 011.
If that's the case, these expressions might be better:

Code:

buf[6] = frequency * 8 % 256 + 3;
buf[7] = frequency * 8 / 256;

Onkel Paul

[-ck]

For grins, I have tried this and it worked at 325, 350 and 375. The error rate at 375 meant the useful hashrate was lower than at 300, but it worked at 325 and 350 quite well and I got 81GH. The power usage went up by 5W (at the wall) at 325 and 10W at 350.

[dylan186]

For grins, I have tried this and it worked at 325, 350 and 375. The error rate at 375 meant the useful hashrate was lower than at 300, but it worked at 325 and 350 quite well and I got 81GH. The power usage went up by 5W (at the wall) at 325 and 10W at 350.

ckolivas:Could you share your firmware?

[-ck]

I don't create firmware, I write the software and build binaries which I load directly. However, I have committed the code supporting the higher speeds to the git master tree for cgminer and xiangfu will create new test firmware shortly supporting the new speeds. Mine has been stable at 350 now for a few hours, averaging ~82GH.

[dylan186]

I don't create firmware, I write the software and build binaries which I load directly. However, I have committed the code supporting the higher speeds to the git master tree for cgminer and xiangfu will create new test firmware shortly supporting the new speeds. Mine has been stable at 350 now for a few hours, averaging ~82GH.

ckolivas：Thanks for everything

[J35st3r]

I don't create firmware, I write the software and build binaries which I load directly. However, I have committed the code supporting the higher speeds to the git master tree for cgminer and xiangfu will create new test firmware shortly supporting the new speeds. Mine has been stable at 350 now for a few hours, averaging ~82GH.

I love it how you guys keep coming up with tweaks to push the performance further and further. Oh for a time machine to go back to the early days of bitcoin with a recent copy of cgminer and fire it up on a GPU. A look at Satoshi's face when (s)he sees the nethash rate would be priceless  

Then again there are more fun things you can do with a time machine, eg Primer ... http://www.imdb.com/title/tt0390384/
They started with ASIC (well JTAG actually but its sort of relevant) and ended up with a time machine. Didn't end well though.
Sort of ... be careful what you wish for, and geeks and ethics don't mix. Anyway I liked the film.
[/OT]

[dogie]

For grins, I have tried this and it worked at 325, 350 and 375. The error rate at 375 meant the useful hashrate was lower than at 300, but it worked at 325 and 350 quite well and I got 81GH. The power usage went up by 5W (at the wall) at 325 and 10W at 350.

That power increase seems strange. Its as if the chips aren't drawing any more power for the same clock rate - which is understandable [fixed voltage]. If they're stable at that speed, then it suggests:

1) Likely is voltage headroom at stock speeds [for undervolting]
2) Temps shouldn't be an issue
3) Longevity shouldn't be affected.

EDIT: UNLESS, we're maxing something out somewhere which is limiting power draw. Although the chips don't need it, they want it. And something somewhere in the chain is bouncing off its max, which could be dangerous or reduce life.

tldr: we don't know

[candoo]

Would anyone compile bfgminer with the settings for 325 350  and 375? please!

[Dalkore]

I don't create firmware, I write the software and build binaries which I load directly. However, I have committed the code supporting the higher speeds to the git master tree for cgminer and xiangfu will create new test firmware shortly supporting the new speeds. Mine has been stable at 350 now for a few hours, averaging ~82GH.

What temps are you keeping your room at to achieve these speeds?

[goxed]

I don't create firmware, I write the software and build binaries which I load directly. However, I have committed the code supporting the higher speeds to the git master tree for cgminer and xiangfu will create new test firmware shortly supporting the new speeds. Mine has been stable at 350 now for a few hours, averaging ~82GH.

+1

[goxed]

Actually

This is working on Avalon 2


 case 356:
                        buf[6] = 0xf3;
                        buf[7] = 0x0a;
                        break;


cgminer version 3.2.0 - Started: [2013-06-21 16:12:28]
--------------------------------------------------------------------------------
 (5s):79.63G (avg):81.75Gh/s | A:1026  R:227  HW:119  U:169.3/m  WU:1159.7/m
 ST: 2  SS: 0  NB: 1  LW: 5983  GF: 0  RF: 0
 Connected to us3.eclipsemc.com diff 62 with stratum as user 
 Block: 00142ffb0e8b1b37...  Diff:19.3M  Started: [16:12:28]  Best share: 20.6K
--------------------------------------------------------------------------------
 [P]ool management ettings [D]isplay options [Q]uit
 AVA 0: 25/ 48C 2640R | 79.27G/82.26Gh/s | A:1027 R:227 HW:120 U: 169.47/m
--------------------------------------------------------------------------------

[k9quaint]

The overclocking opportunities on custom boards are very interesting.
Finally, I have a use for all this liquid helium I have laying around.