POPPUR爱换

 找回密码
 注册

QQ登录

只需一步,快速开始

手机号码,快捷登录

搜索
查看: 3731|回复: 28
打印 上一主题 下一主题

AMD Richer Huddy 抱怨石巨人测试程序三角形太多太小导致显卡性能浪费

[复制链接]
跳转到指定楼层
1#
发表于 2010-10-14 20:35 | 只看该作者 回帖奖励 |倒序浏览 |阅读模式
“I’m confused as to why you’d want to upset people. Historically, with nVidia, there’s the constant smell of burning bridges. Over time, they seem to damage relationships everywhere  and, in the long term, if hampers them”, said Huddy when we asked about the differences each company has in approaching developer relations (DevRel). Bold stuff.

It’s worth mentioning here that along with a couple of geniuses called Doug and Servan, Huddy helped invent DirectX in the first place while working for a company called Rendermorphics. Shortly afterwards, Servan and Doug went to a small software company called Microsoft and Huddy joined an organisation called nVidia. Here endeth the history part.

Huddy shows his true colours

While Microsoft lays down the environment for the development of PC games with its DirectX API, there are differences in each company’s hardware and drivers which means each is better at some things and worse at others. Then there are the consoles. nVidia had the original Xbox graphics and the Playstation 3. Intel had the first Xbox CPU. AMD (ATI) has the Wii and Xbox360 graphics. Proper mish-mash of styles and abilities.

Huddy told us how many software development studios are actively trying to balance all of these combinations, “We’re open to working with anyone, but realistically, we spend most of our time with the 100 or so studios who will be able to put a game into the charts”. Why use that distinction? “These games are going to sell in volume, so more people will be impacted if the code can be improved. It’s an intelligent use of our resources”.

How many of these top studios does AMD work with ? “All of them. My team is scattered across the globe and, between us, we’re fluent in more than 20 human languages. Any developer that will impact the market, regardless of where they are based or how large a company they are, will get help from AMD”.

Just one more quick checks on Huddy’s background to confirm if he’s been doing all this DevRel stuff long enough to have an opinion, “I spent 4 years helping to run develop relations with nVidia before moving across to ATI and I’ve been with this team for the best part of a decade”. OK. Experience established. Let’s continue.

“With one exception, every game released through to 2012 will have been developed on AMD’s Radeon”
We asked Huddy about the ways in which differences in hardware design translate into the DevRel push. “AMD being able to deliver DirectX 11 hardware into the hands of developers a full 6 months ahead of nVidia, means that with one single known exception, every game released through to 2012 will have been developed on AMD’s Radeon hardware. that gives us a huge advantage”.

Despite the name, these Radeon things are no good at all in removing stains from your clothing at low temperatures

“Even after nVidia finally began seeding production Fermi cards, many of the DX11 game development projects were already well underway on AMD hardware. The result is that many of those games will, quite naturally, be tuned for the Radeon series”. Logical statement form Huddy.

We wanted a specific case where DevRel would be used to make the most of a company’s hardware. Huddy gave us a pearl.

“When all you have is a hammer, then everything in the world looks like a nail“.

Sounds like a wise saying, but what does he mean?

“We’re speaking with every development studio in the work that’s likely to create a piece of software that makes it into the charts. All of them are telling us the same thing. nVidia is pushing a single message and that’s tessellation“, explained Huddy.

“Tessellation is about enriching detail, and that’s a good thing, but nVidia is pushing to get as much tessellation as possible into everything”. Huddy reminds us of the problem you get when adding too much salt to your pasta. “Tessellation breaks your image components down in order to add more detail at lower levels and, when it’s done right, it can be stunning”.

Huddy then got scientific, “These days, the most typical resolution for serious gaming is 1080p. A resolution where you have 1920 dots left to right and 1080 from top to bottom. That gives you around 2 million pixels’ worth of work onto the final screen. However, you actually end up working with a much larger picture, to allow for things like light sources and shadow generators that are off screen in the final image, but which still need to be accounted for. The same goes for situations where something is in front of something else, but you don’t know that at the start, so you end up doing work on pixels that may or may not make it to the final cut”.

“Overall, the polygon [Triangle - Ed] size should be around 8-10 pixels in a good gaming environment”, said Huddy. “You also have to allow for the fact that everyone’s hardware works in quads. Both nVidia and AMD use a 2×2 grid of pixels, which are always processed as a group. To be intelligent, a triangle needs to be more than 4 pixels big for tessellation to make sense”.

Interesting enough, but why are we being told this? “With artificial tests like Stone Giant, which was paid for by nVidia, tessellation can be done down to the single pixel level. Even though that pixel can’t be broken away from the 3 other pixels in its quad. Doing additional processing for each pixel in a group of 4 and then throwing 75% of that work away is just sad”.

Huddy touched on benchmarks and we know he will be biased in this department. Knowing he has a built-in bias, we asked him what makes a good benchmark. “It needs to be reproducible, so you get the same results time after time. It also needs to test a new feature. What’s the point of replacing a benchmark with a new one that tests exactly the same thing? Ideally the new benchmark should add something, maybe improved visual quality, so that the user experience is better”.





Searching for the truth in graphics benchmarks gets you into deep water

So far, no mad rantings or foaming at the mouth, “If you’re testing a graphics chip, then try to avoid testing the CPU. With the new cards, it should be a Shader Model 5 test so that you can get an idea of how future games will work”, said Huddy. “Lastly, we all know that code changes constantly before launch. For a benchmark to be really good, it should be created on the final launch code”. Maybe even after the first patches have been released ?

Thinking about the $2M that nVidia paid to buy Crysis 2, we asked Huddy if he thought that nVidia now has a huge DevRel team working putting as many tessellation calls into the game as will fit.

Huddy declined to comment directly on that game, but he did speak about the Batman Arkham Asylum AA fiasco which saw 2 versions of the game released in quick succession following a global flame war across some of the most influential graphics forums on the planet.

He explained, “Batman used a deferred rendering engine. That creates a special situation for applying anti aliasing. There are well known ways of doing it and, as luck would have it, there is a vanilla implementation that works equally well on AMD and nVidia graphics hardware. Let me be clear, this method is well known, works well on AMD and nVidia cards and there is a minimum impact on performance for the customers who have bought the game”.

“Instead of releasing unified code that works on every customer’s card the same, nVidia got the first release of the game to detect if the installed card was a GeForce or Radeon and, if it saw a Radeon, it would turn off the AA feature”, said Huddy.

“The only person this harms if the customer who spent their hard earned money on a brand new game, believing that Batman Arkham Asylum had been written to work as well as possible on the customer’s system”, he pointed out.

Smoother these days, apparently

“This highlights perfectly the fundamental difference in thinking between AMD and nVidia. We would never do that”, said Huddy. “nVidia needs to learn that you should always put the gamer’s experience ahead of your own ego. Issues like the deliberate and unnecessary reduction in image quality seen in the Batman Arkham Asylum situation, shows that nVidia is willing to single out half the market and nobble their experience. That’s just not right. You should never harm PC gamers just to make yourself look good”.

Huddy told us that the publisher re-evaluated the situation in light of the flame-war and, a short while later with no huge effort, a different version of Batman Arkham Asylum appeared on the market with perfect AA on Radeon cards. Nice.


http://www.kitguru.net/component ... g-bridges-says-amd/
头像被屏蔽
2#
发表于 2010-10-14 20:46 | 只看该作者
提示: 作者被禁止或删除 内容自动屏蔽
回复 支持 反对

使用道具 举报

3#
发表于 2010-10-14 20:49 | 只看该作者
那么天堂2是什么太多? 哦,是NVIDIA给的钱太多了,导致A卡使不上劲
回复 支持 反对

使用道具 举报

4#
 楼主| 发表于 2010-10-14 20:51 | 只看该作者
Huddy told us that the publisher re-evaluated the situation in light of the flame-war and, a short while later with no huge effort, a different version of Batman Arkham Asylum appeared on the market with perfect AA on Radeon cards. Nice.

时隔一年大家即将能看到内建 AMD MSAA 支持的 Batman AA 啦
回复 支持 反对

使用道具 举报

5#
 楼主| 发表于 2010-10-14 21:10 | 只看该作者
是不是自己三角形生成性能不行了啊~抱怨别人了
越光宝盒 发表于 2010-10-14 20:46


其实不仅是三角形吞吐的问题,而是包括像素渲染流水线设计的问题。

以前的的 GPU ROP 设计是 2x2,不代表后面的不会改。

其实就算弄成每个三角形都是 4 个像素大小,但是如果这些像素是在三角形边上,那同样会有效率的问题,因为这时候同样是相当于只有 1、2 个像素的尺寸而已。


说一下背景:
以前(怎么说的好像那么遥远,其实不是)由于大量三角形传送是比较低效率的事情,所以人们设计 GPU 的时候,都把它们 group 起来,这样三角形就能广播下去,但是三角面只会越来越多,这样的方式还是得有其他办法来改善才是。
回复 支持 反对

使用道具 举报

6#
发表于 2010-10-14 21:19 | 只看该作者
为什么太多太小的三角形在NV的架构上就没有浪费GPU性能呢?
回复 支持 反对

使用道具 举报

7#
 楼主| 发表于 2010-10-14 21:37 | 只看该作者
richard huddy 认为都浪费。
回复 支持 反对

使用道具 举报

potomac 该用户已被删除
8#
发表于 2010-10-14 21:47 | 只看该作者
提示: 作者被禁止或删除 内容自动屏蔽
回复 支持 反对

使用道具 举报

9#
发表于 2010-10-14 21:48 | 只看该作者
N卡注重复杂应用的处理
A卡认为太复杂的处理在游戏不会普遍,所以不对其优化……


我还是觉得需求和体验决定一切
回复 支持 反对

使用道具 举报

乐极生悲 该用户已被删除
10#
发表于 2010-10-14 21:55 | 只看该作者
提示: 作者被禁止或删除 内容自动屏蔽
回复 支持 反对

使用道具 举报

11#
 楼主| 发表于 2010-10-14 22:00 | 只看该作者
10几年前3D刚开始的时候,最重要的指标就是多边形生成速度,不懂是什么时候硬指标变成渲染速度了。
乐极生悲 发表于 2010-10-14 21:55


其实即使是未来,pixel shader 速度依然是最关键的性能指标。
回复 支持 反对

使用道具 举报

12#
发表于 2010-10-14 22:05 | 只看该作者
他现在除了抱怨还能干嘛!等它跑领先的时候就会说;石巨人是衡量显卡性能的标尺之一.
回复 支持 反对

使用道具 举报

13#
发表于 2010-10-14 22:12 | 只看该作者
为什么非要跑石巨人?以AMD今日之“地位”完全可以无视石巨人!
回复 支持 反对

使用道具 举报

14#
发表于 2010-10-14 22:40 | 只看该作者
又是AMD的抱怨·············
回复 支持 反对

使用道具 举报

15#
发表于 2010-10-15 00:13 | 只看该作者
石巨人的Diffusion DoF才是真正性能杀手,打开后无论A还是N,FPS都将近减半。
回复 支持 反对

使用道具 举报

16#
发表于 2010-10-15 09:16 | 只看该作者
nv为何要推出gf100?没有gf100今天跑石巨人hd5k绝对没对手.
可恶的nv
回复 支持 反对

使用道具 举报

头像被屏蔽
17#
发表于 2010-10-15 09:20 | 只看该作者
提示: 作者被禁止或删除 内容自动屏蔽
回复 支持 反对

使用道具 举报

18#
发表于 2010-10-15 09:26 | 只看该作者
导致显卡性能浪费or浪费显卡性能?
是太轻松不能显现出显卡的性能,还是无实际意义的应试的意思?
回复 支持 反对

使用道具 举报

19#
发表于 2010-10-15 09:27 | 只看该作者
石巨人是什么?
回复 支持 反对

使用道具 举报

20#
发表于 2010-10-15 09:33 | 只看该作者
估计有人会说:这是测试,你买显卡是天天跑测试的么?
回复 支持 反对

使用道具 举报

您需要登录后才可以回帖 登录 | 注册

本版积分规则

广告投放或合作|网站地图|处罚通告|

GMT+8, 2025-2-26 06:37

Powered by Discuz! X3.4

© 2001-2017 POPPUR.

快速回复 返回顶部 返回列表