GGTharos Posted March 22, 2009 Posted March 22, 2009 Looks like things aren't going to take off terribly quickly for MP games: http://www.infoworld.com/article/09/03/20/Multicore_chips_pose_next_big_challenge_for_industry_1.html [sIGPIC][/sIGPIC] Reminder: SAM = Speed Bump :D I used to play flight sims like you, but then I took a slammer to the knee - Yoda
Distiler Posted March 22, 2009 Posted March 22, 2009 (edited) I'm not an expert, but I think the problem is that calculation power per core has not increased so much in the last years. Intel is already developing 80 core cpu, etc., clearly future is many cores, I bet software developers are struggling their brains to go multithreaded. For example, Rise of Flight and Oleg's Storm of War are going to be multicore enhanced games, so it seems bright for simulation development (we need more cpu powerrrrrrr!!). Source engine (valve) is also mutithreaded. BTW how is DCS developing regarding multicore enhancements? touchy subject if you're not going that way, but I wish to know. Edited March 22, 2009 by Distiler AMD Ryzen 1400 // 16 GB DDR4 2933Mhz // Nvidia 1060 6GB // W10 64bit // Microsoft Sidewinder Precision 2
GGTharos Posted March 22, 2009 Author Posted March 22, 2009 DCS will develop multi-threading at some later time. The problem stated in the article is that OS' aren't ready to come with more than 4 cores, and there's no real tools for programmers to easily code for MP. [sIGPIC][/sIGPIC] Reminder: SAM = Speed Bump :D I used to play flight sims like you, but then I took a slammer to the knee - Yoda
Distiler Posted March 22, 2009 Posted March 22, 2009 And what are the specific problems (besides the tools) for coding multithread? I've read something about "syncronizing" data between cpu's, or something like that, but I couldn't understand it well. I know how to program (fortran haha), but I've never faced that problem. AMD Ryzen 1400 // 16 GB DDR4 2933Mhz // Nvidia 1060 6GB // W10 64bit // Microsoft Sidewinder Precision 2
GGTharos Posted March 22, 2009 Author Posted March 22, 2009 Then I suggest, as a beginning that you look at sharing information between threads, and inter-process communication. You can probably read those up on the net somewhere. You won't get the full depth of the problem however until you start coding. ;) To give you an idea, I recall when we had an i++ somewhere in the code, and it was occasionally interrupted at the i, causing the i to be incremented by the thread, then again by the ++, thus causing some odd object loading issues. Took a while to debug that one, too, despite how simple it seems. [sIGPIC][/sIGPIC] Reminder: SAM = Speed Bump :D I used to play flight sims like you, but then I took a slammer to the knee - Yoda
Made.In.China.00 Posted March 23, 2009 Posted March 23, 2009 I have a question. If we keep adding cores, will the CPU and GPU evetually "merge"? I mean there are a lot of difference between CPU and GPU, but the main difference is CPU has a few very powerful processers, and GPU has a lot of, more then 200 for the high-end ones, not so powerful cores.
EtherealN Posted March 23, 2009 Posted March 23, 2009 I don't think they will merge, they are too specialized for that. To "merge" them you would basically look at integrating specialized graphics architectures into a CPU that needs to be good at more general things. This would make the CPU more expensive and niched to those that actually want both facets. Opposed to that is the current modular approach with distinct CPU and GPU that allows Intel and AMD to basically develop and sell the same product to both hobbyists, researchers, industrial applications, workstations and retail/service market display devices. Spreads the development costs on more customers and helps with standardization, both of which help the margins of both Intel/AMD, OEM's, software developers and so on. Of course, for computer games in particular it would be awesome to have a completely standardized hardware market that you service, which is why so many developers and publishers focus so on the console market theese days. But for the PC market it's just not possible. To make it happen we would probably be looking at the GPUs and CPUs diverging into other directions over time and finding a common ground in a roundabout way, but I have a hard time imagining what that would be. (Though there might well be a couple hundred thousand hardware architects out there that already have a very clear idea of what they want to do in that very direction. :P ) There has already been some interesting developments where the nVidia Tesla architecture is used for some rather creative things in research, using it's gazillion-core architecture and specially coded programs to run some fairly advanced research applications on a comparatively extremely cheap "super"computer. But from what I have understood of the technology, even those are just a poor-man's alternative that is more effective on the money but of course doesn't get quite close to "real" supercomputer performance. Sort of like those awesome Playstation clusters. :P [sIGPIC][/sIGPIC] Daniel "EtherealN" Agorander | Даниэль "эфирныйн" Агорандер Intel i7 2600K @ 4.4GHz, ASUS Sabertooth P67, 8GB Corsair Vengeance @ 1600MHz, ASUS GTX 560Ti DirectCU II 1GB, Samsung 830series 512GB SSD, Corsair AX850w, two BENQ screens and TM HOTAS Warthog DCS: A-10C Warthog FAQ | DCS: P-51D FAQ | Remember to read the Forum Rules | | | Life of a Game Tester
Recommended Posts