7-men Syzygy attempt

[noobpwnftw]

tablebase.sesse.net will also mirror, as soon as I can get everything downloaded and verified somehow (are there checksums anywhere?).

The syzygy file format embeds a checksum in each file, which can be checked with tbcheck program.

[Sesse]

All checked, now mirrored. 1.2 TB in all, I suppose more to come (although I don't have 35 TB of disk…). 10 Gbit/sec, northern Europe.

[syzygy]

All checked, now mirrored. 1.2 TB in all, I suppose more to come (although I don't have 35 TB of disk...). 10 Gbit/sec, northern Europe.

I expect it to be more like 20 TB (and approx 10+10 for WDL+DTZ). But I could be wrong - I don't even remember how I arrived at the estimate.

I didn't expect that all this data was going to be mirrored so generously

[Sesse]

My estimation was just wild extrapolation. People talked about 10 TB for WDL, and for the current generated 7-man TBs, DTZ files are about 2.6x the size of WDL.

[Nordlandia]

Steinar H. Gunderson:

Imagine sesse probing 7-men syzygy, it will definitely be a big jump in analysis power. Do you plan/consider purchase SSDs large enough for probing 7-men WDL or multiple in raid 0?

I guess 7-men DTZ can be stored on HDD.

[syzygy]

My estimation was just wild extrapolation. People talked about 10 TB for WDL, and for the current generated 7-man TBs, DTZ files are about 2.6x the size of WDL.

The pawn tables will probably take the ratio down.

For 5-piece tables, the ratios DTZ/WDL are:

Code: Select all

             WDL        DTZ       DTZ/WDL
pawnless   68132288   287321600    4.217
pawnful   326710624   297447328    0.910
total     394842912   584768928    1.481

For 6-piece tables:

Code: Select all

              WDL            DTZ      DTZ/WDL
pawnless   11896768096   30359010784   2.552
pawnful    60937299376   57032537776   0.936
total      72834067472   87391548560   1.200

Going from 5 to 6, WDL size multiplied by 184.5. To keep 7-piece WDL below 10TB, the factor needs to be just below 151 (I am counting 1kb = 1024 ). But Lomonosov DTM size went up by a factor of about 150, I think, and WDL and DTZ should do better than that.

[IQ]

Lets just hope 7p WDL stay below 10TB - so they fit on one HD.

[Sesse]

Imagine sesse probing 7-men syzygy, it will definitely be a big jump in analysis power. Do you plan/consider purchase SSDs large enough for probing 7-men WDL or multiple in raid 0?

I don't plan to buy 10 TB worth of SSD, no. I currently have two SSDs as cache in front of my RAID-6 array, which cache tablebases or whatever else the OS deems necessary. But of course, if the probing pattern is way spread out, they won't be helping any.

I don't know if anyone ever tried to quantify how many Elo each level of tablebases help? I seem to recall Vasik Rajlich said probing Nalimov bases in Rybka helped almost nothing (like, 1 Elo) over just using them at the root, but that's a long time ago.

[Sesse]

I didn't expect that all this data was going to be mirrored so generously :D

Google evidently thought it should be mirrored some more, and Googlebot appeared at my doorstep to download tablebases… :-P

Now added robots.txt.

[Jesse Gersenson]

I don't know if anyone ever tried to quantify how many Elo each level of tablebases help? I seem to recall Vasik Rajlich said probing Nalimov bases in Rybka helped almost nothing (like, 1 Elo) over just using them at the root, but that's a long time ago.

A few years ago I was trying to get the 7-man tables built and Mark Lefler thought it'd be worth 'maybe 5 elo' but of course it needs to be tested, and there are multiple (hardware) factors which affect the performance.

LC zero recently had good success getting sponsorship for hardware.

How much would disks cost? I'd gladly donate money to help get you disks.