Best for chess: i7 5820K or i7 6700k ?

[Jesse Gersenson]

I really don't see how you got those Elo numbers, but they in any way don't seam right.
...
Regarding dual e5-2670, it has 140% more nps than i7-6700 which considering less efficiency on 16 cores vs. 4 cores gives around 65% more strength which further reflects into around 50Elo at 1m, 40Elo at 5m and 25Elo at 120m.

Using Kn/s from Sedat's 2015 stockfish benchmark.

23374 Kn/s = dual e5-2670 = 16 cores
9726 Kn/s = i7-6700k = 4 cores

The equation I'm using is:
19,452 Kn/s = 9726 Kn/s *(SQRT(16 cores / 4 cores))

Elo calculations based on 19,452 Kn/s vs 23,374 Kn/s, game in 1 minuet:

game in 1 minuet=(LOG((23374/19452))/LOG(2))*100
game in 5 minuet=(LOG((23374/19452))/LOG(2))*75
game in 120 minuet=((LOG((23374/19452))/LOG(2))*35

[APassionForCriminalJustic]

http://pricewatch.com/computer_systems_with_os/

Not a single Xeon on the list .
Currently the best buy is dual CPU E5-2670. For up to 600$ full system half the strength (i.e. same as single E5-2699v4) of TCEC machine (and I'm not talking nps coz that's meaningless, but real strength Elo-wise).
http://www.techspot.com/review/1155-aff ... l-xeon-pc/

Yeah sure, nodes per second are "meaningless". Lol. Give me a break. Quit giving people BS advice. Two 2670s versus one 2699 v4 is a joke. Of course you would go with the 2699 v4. You might as well avoid the dual-socket scenario in this case since the speed of the RAM in a single-socket setup should be faster. Moreover, I would much prefer Broadwell over Sandy Bridge any day of the week. I would take 22 cores - and that 55 megabytes of L3 Cache to.

There are much better CPUs out there that one can buy cheap off ebay that will surely provide better computer-chess performance. Going with some ancient Sandy Bridge setup is simply not smart in my opinion unless you just want something dirt cheap.

[APassionForCriminalJustic]

I really don't see how you got those Elo numbers, but they in any way don't seam right.
...
Regarding dual e5-2670, it has 140% more nps than i7-6700 which considering less efficiency on 16 cores vs. 4 cores gives around 65% more strength which further reflects into around 50Elo at 1m, 40Elo at 5m and 25Elo at 120m.

Using Kn/s from Sedat's 2015 stockfish benchmark.

23374 Kn/s = dual e5-2670 = 16 cores
9726 Kn/s = i7-6700k = 4 cores

The equation I'm using is:
19,452 Kn/s = 9726 Kn/s *(SQRT(16 cores / 4 cores))

Elo calculations based on 19,452 Kn/s vs 23,374 Kn/s, game in 1 minuet:

game in 1 minuet=(LOG((23374/19452))/LOG(2))*100
game in 5 minuet=(LOG((23374/19452))/LOG(2))*75
game in 120 minuet=((LOG((23374/19452))/LOG(2))*35

A 4 core is weak. If you want true performance then get more cores. I don't care how you slice it and dice it, more power is more power. A 4 core is no more powerful than a flat tire. If you can search more nodes and positions, and even get more depth then that will equate to more strength. Anything else is illogical. We don't need mathematical formulas to prove anything.

[leavenfish]

Forgive butting in...but within a few days I am looking to replace my old laptop - Core i7-2700QM (2.2ghz/3ghz) Sandy Bridge

Not interested in overclocking. Only interested really in some engine vs engine comparisions...but mostly letting an engine analyse my giant opening system over, say overnight.


If speed is all important...would not an

AMD Quad-Core A12-9800 APU (3.9 GHz, up to 4.2 GHz, 2 MB cache)

be every bit as good as say an

Intel i7-6700 (3.4 ghz up to 4ghz 8mb SmartCache)

This continues to befuddle me...if someone knows of a url to a really good explanation of what cpu makes for my basic needs, please do let me know.

[Cardoso]

My suggestion for you is to wait for AMD Zen (the 8 core variant) to be launched and then decide.
Maybe there will be price cuts from Intel and maybe AMD Zen itself could be a good performer for chess.

regards,
Alvaro

[leavenfish]

The thing is...all those other components in a computer do not get cut. I don't think there would be much of a $$ savings...and there are some 'black Friday'/'Cyber Monday' deals to be had.

My price range...well, would like to keep it at or under $1k.

[Jesse Gersenson]

For chess engines, given cpus with equal specs, the Intel cpu will outperform the AMD cpu.

[leavenfish]

Well yes, of course. That basically goes without saying.

What I am thinking is that since 'real time' analysis or gameplay is not the important thing - I'm wanting an engine to analyse opening systems I play and maybe go thru my games to look for errors...say over night or over a couple of days - I'm thinking a cheaper, slower cpu just left on a bit longer...or say giving it 5 min a move vs 3 min, or go to a set ply...that an AMD or even the i7-6700 (instead of the 600mhz faster i7-6700K) would be just as good. My old SandyBridge laptop just gets too hot.

I'm just not sure where a good trade-off is. Saving $3-400 is a tangible thing after all.

It's funny, I've searched hundreds of pages of Talkchess.com (using search terms) and just googling and I can't find where anyone really addresses this...most just seem to want to build really fast rigs for engine vs engine matches and that's not so important to me.

Thanks for your input.

[Milos]

Well yes, of course. That basically goes without saying.

What I am thinking is that since 'real time' analysis or gameplay is not the important thing - I'm wanting an engine to analyse opening systems I play and maybe go thru my games to look for errors...say over night or over a couple of days - I'm thinking a cheaper, slower cpu just left on a bit longer...or say giving it 5 min a move vs 3 min, or go to a set ply...that an AMD or even the i7-6700 (instead of the 600mhz faster i7-6700K) would be just as good. My old SandyBridge laptop just gets too hot.

I'm just not sure where a good trade-off is. Saving $3-400 is a tangible thing after all.

It's funny, I've searched hundreds of pages of Talkchess.com (using search terms) and just googling and I can't find where anyone really addresses this...most just seem to want to build really fast rigs for engine vs engine matches and that's not so important to me.

Thanks for your input.

Are you looking for laptop or desktop? What is your budget?

[brianr]

It would also help if you could be more specific about analyzing openings and looking for errors. Also, how many games/positions--a few thousand or millions?

Any modern hardware will be fine if you only want 30 ply searches, but if you want to pound positions with a super engine to 40-50+ ply depths then a higher-end box would be helpful.