下一代的VR已经来到身边——小派8K与5K+图片对比评测
itomcoil 2025-04-29 01:23 25 浏览
今年6月,小派公司在全球范围征求他们公司下一代虚拟现实头盔的测试者,我有幸被邀请成为国内唯一的一位。小派参考了测试者的意见,最终推出了定型产品M2,分双4K屏的8K版与双2K屏的5K+版。从理论上说,当然是8K版比5K+版更更好,但是这两个输入的都是双2K信号,8K版在机内插值升频到双4K,5K+版是双原生2K。那么问题来了,插值升频能有效地提升8K版的实际分辨率吗?还是仅仅让画面看上去更细腻?更复杂的是,这两个型号的透镜光学系统是完全一样的,但8K版的液晶屏比5K+版的要大,边缘部分有更多像素未被利用,因此5K+版的实际有效像素有可能比8K版还要多,也就是说清晰度有可能比8K版还要高。当然8K版的像素肯定更小,窗帘效应肯定更小,至少表面看上去会更细腻。小派也一定是意识到了情况的复杂性,竟然允许先前已付款众筹8K版的用户可以重新选择5K+版。这样一来,到底是选8K版还是5K+版,已经成为众筹者最头痛最热门的难题,他们迫切希望在做出选择之前先听听测试者的意见。现在,8K版与5K+版都已经来到了我的面前......
首先声明一下,这不是广告。小派8K和5K目前还没有国内销售的计划,在国外也只能提供给已经付款的众筹用户。我做这个测试,纯粹是出于对VR的热爱。请随意体验一下我拥有的部分VR器材:
当然也未尝没有一点自豪,作为国内唯一一位拥有目前最先进的次世代VR设备用户的自豪感。是的,连王思聪也没有。
那么我一个中文系老师,为什么对VR这个高科技产品这么感兴趣呢?长话短说,我认为VR关系到人类文化的未来,包括文学的未来。有兴趣的可以看看我的这个讲座介绍:
但是VR的发展目前进入到一个停滞期,所以不少人开始唱衰VR。VR确实遇到了一些瓶颈,其中最主要的一个就是视野太窄,分辨率太低,下一代产品遥遥无期。小派8K在这种沉闷气氛下横空出世,规格惊人,打破僵局,令大家非常期待。
言归正传,下面请看我的评测,评测的内容是......英文的,因为主要是供国外的众筹用户参考,不过我的评测主要是用手持手机在头显透镜后实拍说话,所以其实光看图也能明白。我也会做一些简短的翻译。原文地址:
https://forum.pimaxvr.com/t/5k-or-8k-my-test-of-m2-through-pictures-of-aerofly-fs2-project-cars-and-virtual-desktop/8249
Hi VR Enthusiasts,
I am one of the luck testers, happy and eager to share my test result of M2 with you. I won't cover all aspects of M2 as I think SweViver and other testers will do more comprehensive and deeper reviews. I will focus on something I care most: resolution. I mainly play flight simulation in VR, so resolution has been my utmost concern (and complaint) with current VR headsets. I have watched the video of the last Berlin meeting, and I am aware that there are quite some people playing sims as me, and they are uncertain how to choose from 8K and 5K+. I will tried to answer the question with this report. As impressions can be very subjective, I will use the pictures I took through the lens for comparison. And I will give the parameters of these tests as much as possible so that you can reproduce the results at home for comparison with your own headsets.
I have done my comparison tests of 8K and 5K+ through three application: Aerofly FS 2, Project Cars, Virtual Desktop. I use Huawei P20 pro for taking pictures. P20 pro has Leica lens, manual focus, and 3X optical zoom. It's a handy tool for the job, yet the pics are not true reproduction. The pics have some noises, warps, and colorization that are not present when seeing directly through the lens. You just need to look for the most clear part of the pics for more exact reference. For each scene I took 1X and 3X zoom (Project Cars 1X and 2X). The pictures attached are all unmodified. You can click to watch in its original size.
My PiTool is 0.74C, parallel projections mode enabled, FOV normal, rendering quality 1.75. SteamVR Supersampling for 5K+ is set to 40%, rendering at 3278x2914. 8K SS is set to 42%, rendering at 3164x2899. I don't understand why 8K renders at lower resolution than 5K+. Is it proof that 5K+ is using more effective pixels than 8K? Anyway, at the above SteamVR SS they both have almost equal framerate on Aerofly FS 2.
My computer platform: Intel 4790K@4.2G, 32gb, Nvidia GTX 1080
这一段说的是我主要通过三款应用来进行对比评测,每一张图都是未裁剪的原图,并且给出了详细的拍摄参数、图像设置、场景名称,其他用户可以很容易在自己的设备上还原同一场景。
1. Aerofly FS 2.
Graphic setting: 1920x1080, all maxed out.
Scenery: Orbx Netherland TrueEarth, Amsterdam Intl, runway 06.
Airplane: Cessna 172, default startup state. Both headsets are of equal distance to the instruments.
Framerate: 8K: 36, 5K+: 37
Picture comparison8K 1x
5K+ 1x
8K 3x
5K+ 3x
2. Project Cars
Graphic setting: 1920x1080, all maxed out, all special effect such as sun flare, bloom, haze, etc. turned off.
Track: Hockenheim Classic Germany, solo quick race. I set number of opponents to 0 so that the car always starts at the same position.
Vehicle: RWD P30 LMP1, default standby mode.
Framerate: PC does not have in-game framerate counter, nor can I get Steam framerate counter to show in this game. On both headsets the framerates of PR are fairly smooth.
Picture comparison
8K 1x
5K+ 1x
8K 2x
5K+ 2x
3. Virtual Desktop
Graphics setting: 1920x1080, Virtual Desktop built-in supersampling 3.0, screen size 130 degree, screen distance 7.0, no curve, purple nebular environment.
Sample text: Times New Roman, size 10, Windows Wordpad.
Refresh rate: 8K: 80, 5K+: 91. The fresh rate of 5K+ is a bit strange, but I don't think it's a big deal.
Picture comparison:
8K 1x
5K 1x
8K 3x
5K 3x
4. Subjective observation
I let out a "Wow" when I opened Virtual Desktop on 5K+. This is the first time I saw clear texts of a complete Windows desktop on any headset, period. This is a moment to remember, and a milestone in VR history. Pimax can't be praised enough for this achievement alone. I suggest anyone who has a 5K+ should try Virtual Desktop first. I wish you could share my joy. At the Berlin meeting Pimax CEO Robin claimed 5K+ has 9% more effective pixel than 8K, but I feel it's more than 9%. Just look at the pics.
It's easy to notice from the other two tests. Look at the letter "E" on the heading indicator of FS2, and the word "Vittorius" on the roadside of Project Cars.
Yet the clarity of 5K+ does not come at no price. You can immediately notice from all the pics that the pixels are bigger and more noticeable on 5K+. How noticeable? There is a small airplane symbol in the center of the heading indicator of FS2. You can clearly count 7 lines of pixels for the wing of the airplane symbol on 5K+, but it's hard to count them on 8K, whose screen looks smoother and more refined.
这一段是主观印象,与图片是一致的,就是5K+比8K更清晰,因为5K充分利用了更多的液晶面板,而8K有一部分像素是在视场之外,白白浪费了。另外5K用的是双原生2K信号,而8K是双2K信号通过机内插值升频到4K,这样图像是更细腻了,但真实分辨率并没有上去,甚至可能造成了某些虚化效果。
Conclusion
Now we come to the answer to the golden question: 8K or 5K+, which one should we choose? I can't see much difference of FOV between to two. Under the same rendering and supersampling parameter they have almost the same framerate. So effective resolution and pixel size remain the two most cruxial factor. If you are more into simulation games as me, I'd suggest 5K+, as every pixel counts in these games. If you are more into action, adventure, RPG, sports games, etc., which relies more on overall aesthetics, I'd suggest 8K. Or I can put it in another way: if you care more about details and realism, go for 5K+; if you care more about art and beauty, go for 8K.
And finally I must thank Pimax for bringing us a long waited advance in VR!
结论:8K,还是5K+,怎么选呢?我推荐像开飞机和赛车之类的模拟游戏玩家选择5K+,因为这类游戏需要看清仪表和舱外的标志、目标,对真实分辨率要求更高。而RPG、动作、射击、冒险、体育等类游戏的玩家就可以考虑8K,因为这些游戏更讲究画面的优美细腻。那么看电影如何选择呢?其实电影中也是非真实系和唯美系,如果是看欧美电影,5K+更合适,如果是看日式动漫,那当然是8K啦。
要真,还是要美?这其实是人类永恒的选择困难。王国维先生感叹世上的事情,“可爱者不可信,可信者不可爱”。越来越觉得他老人家这个纠结的句式可以推到其他问题上,比如食物:好吃的不健康,健康的不好吃,奈何!
所以VR也是这样,没有一个设备是能穷尽世上所有用户的需求的,但我们需要小派这样的新兴公司,锐意进取,勇于冒险,不想某些挤牙膏的大厂,非把尽可能多的用户榨干才慢吞吞地出下一代,等得花儿都谢了。
谢谢小派!
相关推荐
- 最强聚类模型,层次聚类 !!_层次聚类的优缺点
-
哈喽,我是小白~咱们今天聊聊层次聚类,这种聚类方法在后面的使用,也是非常频繁的~首先,聚类很好理解,聚类(Clustering)就是把一堆“东西”自动分组。这些“东西”可以是人、...
- python决策树用于分类和回归问题实际应用案例
-
决策树(DecisionTrees)通过树状结构进行决策,在每个节点上根据特征进行分支。用于分类和回归问题。实际应用案例:预测一个顾客是否会流失。决策树是一种基于树状结构的机器学习算法,用于解决分类...
- Python教程(四十五):推荐系统-个性化推荐算法
-
今日目标o理解推荐系统的基本概念和类型o掌握协同过滤算法(用户和物品)o学会基于内容的推荐方法o了解矩阵分解和深度学习推荐o掌握推荐系统评估和优化技术推荐系统概述推荐系统是信息过滤系统,用于...
- 简单学Python——NumPy库7——排序和去重
-
NumPy数组排序主要用sort方法,sort方法只能将数值按升充排列(可以用[::-1]的切片方式实现降序排序),并且不改变原数组。例如:importnumpyasnpa=np.array(...
- PyTorch实战:TorchVision目标检测模型微调完
-
PyTorch实战:TorchVision目标检测模型微调完整教程一、什么是微调(Finetuning)?微调(Finetuning)是指在已经预训练好的模型基础上,使用自己的数据对模型进行进一步训练...
- C4.5算法解释_简述c4.5算法的基本思想
-
C4.5算法是ID3算法的改进版,它在特征选择上采用了信息增益比来解决ID3算法对取值较多的特征有偏好的问题。C4.5算法也是一种用于决策树构建的算法,它同样基于信息熵的概念。C4.5算法的步骤如下:...
- Python中的数据聚类及可视化分析实践
-
探索如何通过聚类分析揭露糖尿病预测数据集的特征!我们将运用Python的强力工具,深入挖掘数据,以直观的可视化揭示不同特征间的关系。一同探索聚类分析在糖尿病预测中的实践!所有这些可视化都可以通过数据操...
- 用Python来统计大乐透号码的概率分布
-
用Python来统计大乐透号码的概率分布,可以按照以下步骤进行:导入所需的库:使用Python中的numpy库生成数字序列,使用matplotlib库生成概率分布图。读取大乐透历史数据:从网络上找到大...
- python:支持向量机监督学习算法用于二分类和多分类问题示例
-
监督学习-支持向量机(SVM)支持向量机(SupportVectorMachine,简称SVM)是一种常用的监督学习算法,用于解决分类和回归问题。SVM的目标是找到一个最优的超平面,将不同类别的...
- 25个例子学会Pandas Groupby 操作
-
groupby是Pandas在数据分析中最常用的函数之一。它用于根据给定列中的不同值对数据点(即行)进行分组,分组后的数据可以计算生成组的聚合值。如果我们有一个包含汽车品牌和价格信息的数据集,那么可以...
- 数据挖掘流程_数据挖掘流程主要有哪些步骤
-
数据挖掘流程1.了解需求,确认目标说一下几点思考方法:做什么?目的是什么?目标是什么?为什么要做?有什么价值和意义?如何去做?完整解决方案是什么?2.获取数据pandas读取数据pd.read.c...
- 使用Python寻找图像最常见的颜色_python 以图找图
-
如果我们知道图像或对象最常见的是哪种颜色,那么可以解决图像处理中的几个用例,例如在农业领域,我们可能需要确定水果的成熟度。我们可以简单地检查一下水果的颜色是否在预定的范围内,看看它是成熟的,腐烂的,还...
- 财务预算分析全网最佳实践:从每月分析到每天分析
-
原文链接如下:「链接」掌握本文的方法,你就掌握了企业预算精细化分析的能力,全网首发。数据模拟稍微有点问题,不要在意数据细节,先看下最终效果。在编制财务预算或业务预算的过程中,通常预算的所有数据都是按月...
- 常用数据工具去重方法_数据去重公式
-
在数据处理中,去除重复数据是确保数据质量和分析准确性的关键步骤。特别是在处理多列数据时,保留唯一值组合能够有效清理数据集,避免冗余信息对分析结果的干扰。不同的工具和编程语言提供了多种方法来实现多列去重...
- Python教程(四十):PyTorch深度学习-动态计算图
-
今日目标o理解PyTorch的基本概念和动态计算图o掌握PyTorch张量操作和自动求导o学会构建神经网络模型o了解PyTorch的高级特性o掌握模型训练和部署PyTorch概述PyTorc...
- 一周热门
- 最近发表
- 标签列表
-
- ps图案在哪里 (33)
- super().__init__ (33)
- python 获取日期 (34)
- 0xa (36)
- super().__init__()详解 (33)
- python安装包在哪里找 (33)
- linux查看python版本信息 (35)
- python怎么改成中文 (35)
- php文件怎么在浏览器运行 (33)
- eval在python中的意思 (33)
- python安装opencv库 (35)
- python div (34)
- sticky css (33)
- python中random.randint()函数 (34)
- python去掉字符串中的指定字符 (33)
- python入门经典100题 (34)
- anaconda安装路径 (34)
- yield和return的区别 (33)
- 1到10的阶乘之和是多少 (35)
- python安装sklearn库 (33)
- dom和bom区别 (33)
- js 替换指定位置的字符 (33)
- python判断元素是否存在 (33)
- sorted key (33)
- shutil.copy() (33)