Slashdot Mirror
Ubisoft Points Finger At AMD For Assassin's Creed Unity Poor Performance
MojoKid (1002251) writes "Life is hard when you're a AAA publisher. Last month, Ubisoft blamed weak console hardware for the troubles it had bringing Assassin's Creed Unity up to speed, claiming that it could've hit 100 FPS but for weak console CPUs. Now, in the wake of the game's disastrous launch, the company has changed tactics — suddenly, all of this is AMD's fault. An official company forum post currently reads: "We are aware that the graphics performance of Assassin's Creed Unity on PC may be adversely affected by certain AMD CPU and GPU configurations. This should not affect the vast majority of PC players, but rest assured that AMD and Ubisoft are continuing to work together closely to resolve the issue, and will provide more information as soon as it is available." There are multiple problems with this assessment. First, there's no equivalent Nvidia-centric post on the main forum, and no mention of the fact that if you own an Nvidia card of any vintage but a GTX 970 or 980, you're going to see less-than ideal performance. According to sources, the problem with Assassin's Creed Unity is that the game is issuing tens of thousands of draw calls — up to 50,000 and beyond, in some cases. This is precisely the kind of operation that Mantle and DirectX 12 are designed to handle, but DirectX 11, even 11.2, isn't capable of efficiently processing that many calls at once. It's a fundamental limit of the API and it kicks in harshly in ways that adding more CPU cores simply can't help with.
http://games.slashdot.org/comm...
This is a side effect of what happens when game franchises become more profitable than movie franchises. Once the flow of money starts on a game with a budget in the tens of millions like the Call of Duty, Assassin's Creed, or Grand Theft Auto franchises to name just a few - there comes a point of no return where you finish what you started, because you've sunk millions upon millions into something that just turned out *wrong*, like this. Just like Warner Brothers couldn't put the Green Lantern movie on hold and rewrite it to not suck, Ubisoft backed themselves into a corner.
Nobody had the balls or the power to say "Wait a minute, we're overreaching. Let's scale this back to something that will actually run." Instead, they launch a buggy, bad game because they're into just the marketing campaign for tens of millions of dollars. It's so much worse for consumers than a flop of a movie, because you're spending $60+ on the cost of entry, and when the reviewers are embargoed there's just no way to tell if you're going to get screwed. Thank the big budget productions and stock market demands for this kind of disaster.
Every time I see something like this, or a botched Call of Duty release, I get a *little* less annoyed with Valve for not saying a word about Half-Life 3/Ep. 3. They're private. They can take the time without investors freaking out.
My own pointless vanity vintage computing page
the linked comment is super cool, but unsourced. However, it has a ring of truth for me. the comment calls out several nvidia technologies like TXAA and SMAA. It was the second time today I heard those terms. The first was when watching a gamestop video where an designer talks about all the cool tech that makes far cry 4 so pretty on the compuper. "Together with NVIDIA, Ubisoft has been working to incorporate GAMEWORKS technologies to add visual enhancements for the PC version of the game." LINK
If we want to see a change of heart from the publishers, people should start actually returning defective games.
Despite what the EULA may say, lemon laws make it illegal to sell something that doesn't work, and even if a store says they don't take returns of software, if you tell them the return is because it's defective they'll take it anyways.
I can guarantee you if all the people who gave cash to Ubisoft turned around and asked for their money back because the game is defective (doesn't even play on a console), Ubi would think twice about pulling similar shenanigans in the future.
Here's something that doesn't need 'conspiracy' to understand. Unity is playing bad on the PC because they're issuing 50k draw calls on DX11.
Ironically, instead of blaming AMD for this, AMD is actually providing a solution. I don't like it personally, but the Mantle API specifically solves this problem today while we wait for DX12/OpenGL Next.
Of course, it's only available on AMD hardware and besides, because Ubi is in a company wide PR deal with nVidia to use GameWorks(TM) THEY CAN'T USE IT!
So instead of blaming AMD, Ubi should either go sit in a corner (because they know what they did wrong), or they need to look into a mirror (because they don't recognize that they're the real problem)
Belief is the currency of delusion.
That must be the reason the game runs like utter crap on PC as well. Ohh wait, it is not!
TotalBiscuit - "Let's not play Assassins Creed: Unity yet "
https://www.youtube.com/watch?...
They knew they had a shit game before the release.
https://www.youtube.com/watch?...
Everyone who buys Wild Hunt will receive 16 specially prepared DLCs absolutely for free, regardless of platform.
>and no mention of the fact that if you own an Nvidia card of any vintage but a GTX 970 or 980, you're going to see less-than ideal performance.
That is because even on (triple) sli'ed GTX 980's you will not get a decent performance. Gaming reviewer/personality TotalBiscuit has gotten terrible performance (dropping below 20 fps at times, full of glitches, crashes) on a new rig with two GTX 980's. In his video he also mentions asking a friend of his with three gtx 980's if he was experiencing the same issues and he was. How do you even manage to mess up that badly?
Their stance is completely bogus. Let's take a look at DICE's experience adding a Mantle renderer for Battlefield 4, presented at the AMD & Microsoft Developer Day conference here in Stockholm this past June: http://amd-dev.wpengine.netdna-cdn.com/wordpress/media/2012/10/Rendering-Battlefield-4-with-Mantle-Yuriy-ODonnell.ppsx
The numbers don't lie. For those who don't want to download a PowerPoint viewer, I'll give the money shot:
Benchmarking machine: Core i7-390x, AMD Radeon R9 290x, running at 1080p with Ultra graphics settings
DX11 renderer: Minimum frame rate 42fps, average frame rate 78fps
Mantle renderer: Minimum frame rate 94fps, average frame rate 120fps
The only thing I get out of NVidia not wanting to make use of the Mantle API is a pathological case of Not Invented Here syndrome, combined with a long-term gamble of DirectX 12 providing a cross-vendor implementation of an API similar to Mantle. For those wanting to learn more about DX12, there's a presentation from the same conference here: http://amd-dev.wpengine.netdna-cdn.com/wordpress/media/2012/10/Introduction-To-DX12-Ivan-Nevraev.ppsx
In general it seems to provide the same sort of benefits as Mantle, in that it removes a lot of legacy cruft from the pipeline and puts the onus of redundant state checking and resource management on the application authors. This shouldn't be a major problem in and of itself, as one of the things mentioned during the talk was that a lot of engines already do these things, so the kernel-mode driver doing the same checks is simply extra work. I can't blame NVidia for holding out for DX12 given that it will provide a similar bare-metal interface as Mantle, while having support across IHVs, but to say that there's no benefit to using Mantle - and by extension, a bare-metal GPU interface in general - is patently ridiculous given the performance improvements that companies using Mantle have seen.
Problem is the game plays like shit no matter what hardware you're using. The claim against AMD is unfounded because a GTX770, a near-flagship card, can't play the game worth a damn, even with lowered settings. I can't beleive I wasted money buying that garbage.
The problem (or solution rather) is that developers don't want to write the same game 7 times. OpenGL(Linux/Mac OS X/PS3), DirectX9(PC baseline), DirectX12(PC high end), Mantle(PS4), Metal(OS X/iOS8),OpenGL ES2(Android, old iOS),OpenGL ES3(iOS)
They will simply design a middleware that can "intelligently" pick a rendering backend, and if the game suffers, it suffers because of the weakest backends (DirectX9, OpenGL ES2) force it to. This is a problem with Unity (the 3D game engine), and is a problem with Unreal engine.
Oddly enough the Crytek engine actually works better on AMD hardware (and Crytek games are often bundled) because the games support higher DirectX levels out of the box.
But no single-player game engine will ever work for a MMO game, due to the need of many objects in motion at once. The same Crytek engine used for a MMO looks a lot like a 6 year old game. This is because they trade off detail for simultaneous objects because of the need to limit draw calls.
Right, but just a couple months ago these guys were talking up their new game and how they were trading high FPS off in favor of high object/poly counts. I believe they were especially pleased with themselves about how many unique actors they could get into the crowds.
So it seems to be a bit of a self-made problem; over-driving your target hardware OR API is a programming problem- target better hardware or use a different API, but stop blaming other people for sucking
Look, gamers(as a whole) have no history of actually doing anything to reign in companies that abuse. The AAA companies' executives are familiar with this, and the sense of entitlement that causes gamers to demand things for petty reasons and then given in without a fight.
I'm in year 6 of a personal EA/Ubissoft/Activision boycott, and I feel like anyone who cares about getting decent treatment as customers should be too. Unfortunately, they clearly don't need my money, and are doing fine. It's just one of those cases where successful marketing clearly beats out delivering a decent product.
This is fundamentally what GamerGate is too, petty, entitled people making unreasonable demands, and then doing nothing self-sacrificing to make any meaningful changes.
I can probably make some educated guesses about what may have transpired, at least from the performance side, since I've done engine-side programming for AAA games in the past.
Unless you're working with an established and already-polished game engine, all the art content for a game of this size has to be built far in advance of when the engine is fully ready to render it at full efficiency. By it's very nature, optimization is something that has to occur near the end of development for a game, since there's no way to optimize game features until they're largely finished and can approach performance issues holistically. The hardest thing about that is you have to make a very early prediction about how much your game will be able to render. It's extremely time-consuming to fix if it turns out your engine simply can't cope with the amount of artwork or game content it's being asked to process, as that artwork and game content has been in the pipeline for years.
The kicker is that you can't really know for sure what the bottlenecks are exactly and how you can improve on them before you begin the investigation and optimization process, nor can you really predict with 100% certainty how effective your efforts will be, or how long it will take. This is why the recommended specs on boxes are often, at best, simply guesses that are made by the engine developers many months in advance of the title's ship date, and are a reflection of how well they *think* they can get the game engine working. Of course, in other cases, it's managerial wishful thinking, trying to sucker people with lower-end systems into purchasing the game. To me, it seems entirely likely that the programmers either overestimated how much they could optimize the engine / game code or the artists went far beyond their established budgets. Maybe both. Management compounded this issue by not giving the development team time enough to fix the problems.
None of this excuses them in the least, of course, especially on consoles with immutable, fixed hardware to test on. They should have owned up a many months ago and let people know the game wouldn't be ready, because there's zero chance they didn't know about all these problems. Unfortunately, there's a great deal of pressure put on programmers to simply try to patch up the game as best they can given the current time left in the schedule, rather than re-assessing realistically how much time they *actually* need to fix the game, because, you know, money. Instead, I'd imagine that those guys were crunching for many months before the game shipped, and they're still crunching away with insane hours, trying to fix all those bugs. It probably ending up being counter-productive too, because, at least in my case, the quality of my code dropped rather dramatically when I was exhausted.
It's pretty difficult to really know what's going on inside a company. For any game we released, I always saw lots of fan speculation about what was going on, and more often than not, it was well off the mark. So definitely take any speculation, including mine, with a grain of salt. What's absolutely inescapable, though, is that Ubisoft management is ultimately responsible for the go/no-go ship decision, and decided that they didn't care enough about their customers or their reputation to bother getting their game polished to an acceptable standard before launch.
I haven't bought an Ubisoft game since they started on this ridiculous anti-consumer DRM campaign, and this makes me really glad I'm still staying the hell away from them. Yeah, I'd probably have enjoyed the Assassin's Creed series, but there are plenty of game companies that don't piss all over their customers, and they'll be getting my dollars instead.
Irony: Agile development has too much intertia to be abandoned now.
Steam has indeed come a long way, about 10 years ago it was loathed and hated by gamers. Many people would not buy a game if it needed Steam, Ubisoft with their crappy launcher are where Steam was 10 years ago.
10 years ago, Steam was a glorified auto-updater that sat there and sucked up system resources... something like 64MB of RAM when 128-512MB was standard.
Steam now takes something like 128MB of RAM in a time when 8192-16384MB of RAM is common. In addition to being an auto-updater, it also has a store, friends list, friends chat, game library, a non-puke green color scheme, and a host of other features. ...and if you ask me how I know this, I'll toss my Steam "11 year" badge at your face.
(Note: I'm guesstimating at these RAM usage numbers, and they're the numbers when you're not actively using it.)
GLaDOS for President 2016! "Well here we are again. It's always such a pleasure." -- GLaDOS, 2011