| Life In 19x19 http://prod.lifein19x19.com/ |
|
| Champion Go HD strength on various iPad models http://prod.lifein19x19.com/viewtopic.php?f=18&t=10089 |
Page 1 of 1 |
| Author: | RBerenguel [ Wed Apr 02, 2014 3:54 am ] |
| Post subject: | Re: Champion Go HD strength on various iPad models |
The only meaningful test I can imagine is pitting CGo HD from my iPad Air vs my old iPad 1, playing 10-20 games at maximum level (which should be 10 or 9, because the iPad1 doesn't get the extra strong levels) and comparing results. But this would take a pretty long time, since the iPad1 version at maximum level is quite slow. |
|
| Author: | Boidhre [ Wed Apr 02, 2014 4:22 am ] |
| Post subject: | Re: Champion Go HD strength on various iPad models |
I don't have much time, so I just did a quick 9x9 game between an iPad 2 and an iPad 4, both on maximum level. The iPad 2 won as Black by 0.5 with 6.5 komi for the iPad 4 as White. There was a noticeable difference in speed between the two. If I have time tonight I'll do a 19x19 game. |
|
| Author: | Boidhre [ Wed Apr 02, 2014 5:49 am ] | ||
| Post subject: | Re: Champion Go HD strength on various iPad models | ||
I did a 19x19 game, I made an error and had AI resign turned off (I was wondering) but I didn't notice anything to suggest that the much greater processing power of the iPad4 was being used other than, turns were shorter and the iPad 2 got a *lot* hotter during play. I didn't enjoy this game to watch/input: Both at Level 12.
|
|||
| Author: | RBerenguel [ Wed Apr 02, 2014 9:05 am ] |
| Post subject: | Re: Champion Go HD strength on various iPad models |
Marcel Grünauer wrote: Boidhre wrote: I did a 19x19 game, I made an error and had AI resign turned off (I was wondering) but I didn't notice anything to suggest that the much greater processing power of the iPad4 was being used other than, turns were shorter and the iPad 2 got a *lot* hotter during play. Thank you very much for your time and effort! It's unfortunate that the levels seem to have fixed calculation depth. So that's no reason to buy a more powerful model. Marcel Speed is very relevant. Also, if a powerful go engine is the ultimate goal, a portable tablet is not the best deal. I'd bet that for the same thinking time limits, fuego or pachi in my Macbook Air (granted, slightly heavier than an iPad Air) could beat Crazy Stone on an iPad (Air or otherwise) |
|
| Author: | Boidhre [ Wed Apr 02, 2014 9:54 am ] |
| Post subject: | Re: Champion Go HD strength on various iPad models |
RBerenguel wrote: Speed is very relevant. Also, if a powerful go engine is the ultimate goal, a portable tablet is not the best deal. I'd bet that for the same thinking time limits, fuego or pachi in my Macbook Air (granted, slightly heavier than an iPad Air) could beat Crazy Stone on an iPad (Air or otherwise) Honestly, you wouldn't notice the speed difference if they weren't side by side. The big difference is in the heat generated, the newer A series chips are a lot better that way and Champion Go doesn't seem to tax them much (it really taxes an iPad 1). |
|
| Author: | RBerenguel [ Wed Apr 02, 2014 10:59 am ] |
| Post subject: | Re: Champion Go HD strength on various iPad models |
Boidhre wrote: RBerenguel wrote: Speed is very relevant. Also, if a powerful go engine is the ultimate goal, a portable tablet is not the best deal. I'd bet that for the same thinking time limits, fuego or pachi in my Macbook Air (granted, slightly heavier than an iPad Air) could beat Crazy Stone on an iPad (Air or otherwise) Honestly, you wouldn't notice the speed difference if they weren't side by side. The big difference is in the heat generated, the newer A series chips are a lot better that way and Champion Go doesn't seem to tax them much (it really taxes an iPad 1). That's probably true (can't check, only have iPad 1 for comparison and the difference is... way more than striking.) In any case, even 0.1 second difference matters, when playing. It's just a matter of psychological response: faster responding UI makes for an impression of speed which far outweights the real difference (a personal case in point is the Acme text editor vs Emacs or Macvim: the cursor movement speed/key repeat speed is far, far higher in Acme, it feels way, way faster when editing.) |
|
| Author: | xed_over [ Wed Apr 02, 2014 1:56 pm ] |
| Post subject: | Re: Champion Go HD strength on various iPad models |
RBerenguel wrote: Marcel Grünauer wrote: Boidhre wrote: I did a 19x19 game, I made an error and had AI resign turned off (I was wondering) but I didn't notice anything to suggest that the much greater processing power of the iPad4 was being used other than, turns were shorter and the iPad 2 got a *lot* hotter during play. Thank you very much for your time and effort! It's unfortunate that the levels seem to have fixed calculation depth. So that's no reason to buy a more powerful model. Marcel Speed is very relevant. Not if its not being used. |
|
| Author: | RBerenguel [ Wed Apr 02, 2014 2:40 pm ] |
| Post subject: | Re: Champion Go HD strength on various iPad models |
Huh? Speed is indeed being used. Time to compute is reduced. |
|
| Author: | Yertle [ Wed Apr 02, 2014 6:53 pm ] |
| Post subject: | Re: Champion Go HD strength on various iPad models |
Marcel Grünauer wrote: I have an iPad 3 and Champion Go HD plays fairly well (for my level). Now I am wondering how strong it is on newer iPad models. According to this benchmark site, the iPad Air has over 5x the performance of the iPad 3. So I'm wondering how the playing levels of Champion Go HD are defined. For example, on the highest level, will it play a lot stronger on the iPad Air, or will it play more or less the same kinds of moves but a lot faster? Even if it was stronger on an iPad air it wouldn't make sense to purchase. I have this program and have been purchasing tickets for the server engine and that is really strong, basically the same machine that is a 5D on KGS. I am a 4k KGS and I am struggling against it taking 8 stones. |
|
| Author: | xed_over [ Wed Apr 02, 2014 11:52 pm ] |
| Post subject: | Re: Champion Go HD strength on various iPad models |
RBerenguel wrote: Huh? Speed is indeed being used. Time to compute is reduced. what's the point of computing faster, if it doesn't compute any deeper? so I don't have to wait as long for the same weak move... big deal. |
|
| Author: | badukJr [ Thu Apr 03, 2014 5:15 am ] |
| Post subject: | Re: Champion Go HD strength on various iPad models |
Other than server mode, the issue is this program is not much stronger than gnugo on a PC. On galaxy the wait for level 10 is still shorter than 10 second blitz so I hope the calculation depth is redone in the near future. |
|
| Author: | RBerenguel [ Thu Apr 03, 2014 5:52 am ] |
| Post subject: | Re: Champion Go HD strength on various iPad models |
xed_over wrote: RBerenguel wrote: Huh? Speed is indeed being used. Time to compute is reduced. what's the point of computing faster, if it doesn't compute any deeper? so I don't have to wait as long for the same weak move... big deal. This is not an issue with the device, but with the program. It has probably fixed depths for the levels, so level 10 in all iDevices is as equally strong. Actually, it makes sense, although they should then add more levels as hardware improves. |
|
| Author: | Mike Novack [ Thu Apr 03, 2014 6:48 am ] |
| Post subject: | Re: Champion Go HD strength on various iPad models |
RBerenguel wrote: xed_over wrote: RBerenguel wrote: Huh? Speed is indeed being used. Time to compute is reduced. what's the point of computing faster, if it doesn't compute any deeper? so I don't have to wait as long for the same weak move... big deal. This is not an issue with the device, but with the program. It has probably fixed depths for the levels, so level 10 in all iDevices is as equally strong. Actually, it makes sense, although they should then add more levels as hardware improves. In some of these cases we have versions of the programs intended to be run on communication devices (that are weak computers) and versions intended to run on computers of normal power on up to powerful workstations, etc. Not really reasonable to have a single program (in effect) for those very different environments. Though of course could be two combined in one (that at the start, examines how much crunch power is available and selects the appropriate sub program). Remember, the crunch power of an iDevice is minimal compared to a serious computer though this is deceptive because when used to support services that are I/O limited rather than crunch limited the lack of computing power isn't obvious. |
|
| Author: | Mike Novack [ Thu Apr 03, 2014 7:01 am ] |
| Post subject: | Re: Champion Go HD strength on various iPad models |
We also should note << to avoid apples vs oranges >> "in server mode" has nothing to do with go playing programs running on the idevice. The idevice is intended (primary purpose) as a communications device, has adequate power for that, and in this case being used for that primary purpose, a client communicating with a program running on a remote server. The strength of the game now depends on the hardware where that server is running, not that of the idevice. We should also note that "on a PC" depends very much on the PC. Yes of course, some of the go playing programs out there are running on servers* on very powerful machines. But you would not have a awful lot less power if your home PC had an i7-3770 or i7-4770 in it, neither of which are out of the question. I'm am not sure any of the servers out there are running on twice the crunch power of these. * A "server" is really a computer program, but we usually also call the sorts of machines these programs run on "servers". A bit confusing, but a server (machine) might be running more than one server (program). For example, the same email handling machine might be running both POP and SMTP (server to deliver mail incoming to your computer and accept outgoing mail from your computer) |
|
| Author: | Yertle [ Thu Apr 03, 2014 9:38 am ] |
| Post subject: | Re: Champion Go HD strength on various iPad models |
Marcel Grünauer wrote: In the server engine mode, can you also pause the game (i.e., close the iPad etc.) and resume at a later time? Can you load a kifu and have the server continue to play from a certain point on? You definitely can leave the app, turn off you device, and it will pick up right where you left off. I do this all the time. I'm not sure about loading a kifu and play from that point, unless I missed something I don't think so. You also need an internet connection so the program can communicate with the server. |
|
| Page 1 of 1 | All times are UTC - 8 hours [ DST ] |
| Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group http://www.phpbb.com/ |
|
