Grid 2.0

根据上面的帖子，我认为你不能分别进行这两个讨论;
如果您想谈谈将GRID分离为许可软件产品的技术原因，那么您还必须解决这对业务案例和每个席位的总体成本的影响。
问题不在于NVIDIA选择将硬件和软件分开，因为其逻辑相当合理，但这种分离导致成本大幅增加。
即使技术很棒，你仍然需要将业务案例加起来。
虽然计算和图形之间的模式切换是一个很好的技术理念，但它仍然是一个优势 - 特别是考虑到与开普勒相比Maxwell硬件的计算能力。
问题是，如果你没有这种双重用途要求，那么所有发生的事情就是vGPU的实施成本更高。
我还认为关闭虚拟机以释放许可证的想法很难与真正的单词用例保持一致。
考虑使用K1 / K2硬件将vGPU实施到其组织中并希望扩展其部署的客户。
随着GRID 2.0的发布，该客户可能会发现自己投资于一种不再与其预算和业务案例保持一致的技术。
他们现在的答案是留在K1 / K2，但考虑到它们所基于的GPU，这些必须具有相当有限的寿命？
我认为这种模式可能会损害NVIDIA和VDI的进步，因为它可能会拖延项目或迫使客户重新考虑他们的计划。
它也可能诱使完全购买vGPU概念的客户寻找其他选择。
您可以在英国查看Scan Computers的许可定价：
https://www.scan.co.uk/search.aspx?q=nvidia+grid+2.0
假设您想要使用M60和8个用户，每个用户都有2GB双屏幕配置文件。
我需要Virtual Workstation版本，因为我需要Quadro驱动程序。
由于缺乏更好的参考点，我将不得不使用他们的M60价格，并允许使用教育定价进行许可，以免不公平：
1x M60卡的价格：£4,477.00
8x vWS许可证的价格：£4,000.00
8x vWS SUMS的价格：1,200.00英镑
这为您提供了9,677.00英镑的总成本，让8位用户启动并运行vGPU，每位用户只需1,200.00英镑（1,740.00美元）。
如果你看一下与中档Quadro GPU相比，这是性能配置文件适合的地方，这个价位非常高。
那么你当然有每年更新的SUMS。
这个价格点的最大问题不在于它与等效的Quadro卡相比如何，而是与当前一代的vGPU相比如何。
我很欣赏M6 / M60的用户密度更高，但即使您使用两张K2卡并比较价格点（使用扫描定价来确定连续性）：
2x K2卡的价格：5,372.90英镑
假设每个GPU配置文件有2个用户，这意味着每个卡有4个用户，我们的每个用户为672.00英镑（975.00美元）。
这意味着GRID 2.0虽然可能是密度的两倍，但也可能是昂贵的两倍。
如果我们采用不同的用例，请说每个GPU使用PC版许可16个用户：
M60硬件的价格：£4,477.00
32x PC版许可证的价格：£4,800.00
32x PC版SUMS的价格：1,600.00英镑
这意味着总价格为10,877.00英镑或每位用户340.00英镑（495.00美元）。
这可能看起来不是很多钱，但这个档案的目标是知识工作者和低端用户。
您可以轻松地将每个虚拟桌面的价格提高一倍。
当然，您可以说这些只是在线价格而且不太可能是客户实际支付的价格，但那么为什么首先将价格定得这么高呢？
作为一项技术，vGPU非常棒。
作为一款硬件，特斯拉M6和M60非常棒。
就最大化硬件使用而言，整个模式切换方法是一个好主意。
但是当你把它与许可证放在一起时，它就不会加起来。
我感谢每位用户提供1美元的优惠，但这是时间和数量限制 - 它可能会带来一些短期采用，但它不是一个长期的解决方案。



以下为原文

As per the post above, I don’t think you can have the two discussions separately; if you want to talk about the technical reasons for separating GRID into a licensed software product then you also have to address the impact this has on the business case and overall cost per seat. The problem is not that NVIDIA have chosen to separate the hardware and software, as the logic for this is fairly sound, but that this separation has led to a substantial increase in cost. Even if the technology is great you still have to make the business case add up.

While mode switching between compute and graphics is a great technical idea, it is still an edge case - especially given the compute capabilities of the Maxwell hardware by comparison to Kepler. The problem is that if you don’t have this dual-use requirement then all that happened is that vGPU just became more expensive to implement. I also think the idea of powering down VMs to free up a licence is difficult to align with real word use cases.

Consider a customer who has been implementing vGPU into their organisation using K1/K2 hardware, and wants to expand their deployment. With the release of GRID 2.0, this customer may find themselves invested in a technology which no longer aligns with their budget and business case. The answer for them right now is stay with K1/K2, but these must have quite a finite lifespan at this point given the GPU they are based on?

I do think that this model could be damaging for NVIDIA and the advancement of VDI, as it could stall projects or force customers to reconsider their plans. It may also tempt customers who were fully bought in to the concept of vGPU to look to other options.

You can see pricing for the licensing from Scan Computers in the UK:
https://www.scan.co.uk/search.aspx?q=nvidia+grid+2.0

Let’s assume you want to go with the M60 and eight users, each with a 2GB dual screen profile. I need the Virtual Workstation edition because I need Quadro drivers. For lack of a better point of reference I’ll have to use their M60 price, and lets use the Educational pricing for the licencing so as not to be unfair:

Price for 1x M60 card:  £4,477.00
Price for 8x vWS licences:  £4,000.00
Price for 8x vWS SUMS:  £1,200.00

This gives you a total cost of £9,677.00 to get eight users up and running with vGPU, which is just over £1,200.00 per user ($1,740.00). If you look at this in comparison to a mid-range Quadro GPU, which is where the performance profile fits, this price point is very high. Then of course you have the yearly renewing SUMS.

The biggest problem with this price point is not how it compares to an equivalent Quadro card, but how it compares to the current generation of vGPU. I appreciate that the user density with M6/M60 is greater, but even if you take two K2 cards and compare the price point (using Scan pricing for continuity):

Price for 2x K2 cards:  £5,372.90

Assuming a 2 user per GPU profile, which means 4 users per card, we’re at £672.00 per user ($975.00). This means that GRID 2.0, while potentially twice as dense, is also potentially twice as expensive.

If we take a different use case, say 16 users per GPU using the PC edition licensing:

Price for M60 hardware:  £4,477.00
Price for 32x PC edition licences:  £4,800.00
Price for 32x PC edition SUMS:  £1,600.00

This means a total price of £10,877.00 or £340.00 per user ($495.00). This might not seem like a vast amount of money, but this profile is targeting knowledge workers and low-end users. You could easily be doubling the price per virtual desktop.

Of course you can say that these are just online prices and unlikely to be what customers actually pay, but then why set the prices this high in the first place?

As a technology, vGPU is great. As a piece of hardware, the Tesla M6 and M60 are great. The whole mode switching approach is a great idea in terms of maximising the use of hardware. But when you put it all together with the licensing it just doesn’t add up. I appreciate there is a $1 per user offer but this is both time and quantity limited - it might do something to drive some short term adoption, but it isn’t a long term solution.

这是一个简单的练习来回顾数字。
初始的前期购买和年度维护都使得GRID 2.0选项比GRID 1.0或其他GPU直通配置更加昂贵，并且许多用户甚至不需要GRID 2.0的高级功能。
我最近听到一位潜在的M60客户说他们不仅考虑使用GRID K2，而且还考虑不升级他们的平台，这样他们就不会在未来几年遇到支持/驱动程序不兼容问题。
这反过来也会损害提供GPU服务器平台的供应商，如VMware和Citrix。
如上所述，此许可模式的效果将包括潜在客户完全放弃研究GPU / vGPU技术或延迟并寻找其他供应商的选项。



以下为原文

It's a simple enough exercise to go over the numbers. Both the initial up-front purchase and annual maintenance make a GRID 2.0 option much more costly than a GRID 1.0 or other GPU passthrough configuration, and many users do not even need the advanced features of GRID 2.0. I heard from one potential M60 customer very recently that they are considering not only staying with the GRID K2, but not upgrading their platform so that they don't run into support/driver incompatibility issues over the coming years. That in turn also hurts vendors that provide GPU server platforms, like VMware and Citrix. As mentioned above, effects of this licensing model will include potential customers putting off looking into GPU/vGPU technology altogether or delaying and looking to other vendors for options.

我们只能希望Nvidia能够听取他们的客户/用户的意见......



以下为原文

We can only hope for Nvidia to listen to their customers/users...

嗨erlog，
我希望我们能够做到 - 持续进行中有一些变化可以促进快速反馈并确保最具代表性的观点有机会最有效地提出和讨论这一点。
我们已经注意到Tobias在这个主题上的领导，他将参与我们的预算以获得您的意见。
我真的不能说更多。
但请相信我们正在倾听。
雷切尔



以下为原文

Hi erlog,

I hope we are and do - on-going there are some changes going on which should facilitate fast feedback and ensuring the most representative views have an opportunity to raise and discuss this most effectively. We've taken note of Tobias' lead on this thread and he will be involved on our budget to get your opinions heard. I can't really say more. But please believe we are listening.

Rachel

我们非常感谢Rachel以及NVIDIA的所有其他人。
我们都非常希望看到一些让每个人都满意的东西！



以下为原文

And we do appreciate that, Rachel, and all others at NVIDIA. We would all really like to see something worked out to everyone's satisfaction!

如果我想将M6 / M60硬件用于vGPU之外的其他东西（例如直通或XenApp），我还需要支付许可吗？
我听说过这个，但我认为这是一个误解。



以下为原文

If I wanted to use M6/M60 hardware for something other than vGPU (e.g. pass-through or XenApp), do I still need to pay for licencing? I've heard this mentioned but assumed it was a misunderstanding.

与此同时，vGPU还出现了其他技术。
我不知道他们的工作与Nvidia相比有多好，但他们的营销是这样的：
“为您的投资提供最大价值
出色的每美元性能：IT经理无需担心签订技术许可或支付任何额外的用户费用，企业可以获得VDI的全部经济利益“
与此同时，我将为我的项目订购几个K1。
希望我能得到一段时间的支持。



以下为原文

In the meantime there are other technologies emerging for vGPU. I don't know how well they work compared to Nvidia but they marketing is like that:

"Delivering the Most Value for Your Investment

Outstanding Performance Per Dollar: With IT managers not having to worry about entering into a technology license or paying any additional user fees, enterprises get the full economic benefit of VDI"

In the meantime I will order a couple of K1 for my projects. Hopefully I will get support for some time to come.

嗨伙计，
论坛帖子是一个很好的讨论论坛，但这个论坛已经很久了。
我们的新总经理/副总裁于11月进入并且已经为我们清除了预算，以便我们反馈机制，并确保他和团队中的决策者能够讨论这个并获得面对面的一般反馈。
这需要一些规划，但Tobias和其他人提供了如此详细的信息意味着我们将Tobias和其他一些重要客户/社区领导/合作伙伴分散到NVIDIA总部的一个地方，我们付出代价，他们可以面对面交流
与执行官，商业，开发，测试等面谈......我刚刚在他的酒店遇到了托比亚斯，他听到了安全的声音，我很确定愿意转达你的担忧。
我们现在至少都不可能分享讨论的细节，但我希望它能让那些花时间和精力合理地讨论并提出他们的担忧的人有信心我们不是一个不露面的组织。
这也将是一个长期的讨论，关于确保将来我们在整个产品中更好地沟通，倾听，行动等......一个与我们的企业合作伙伴保持一致但可能是GPU /硬件新手的结构，机制....更多
在即将来临;-)
在Twitter上的#GRIDdays上保持一个想法......在那里你可以关注Kreidl博士！
让我知道任何评论，关注等。
最好的祝愿，
Racehl



以下为原文

Hi folks,

A forum thread is a good discussion forum but this one is long running. Our new GM/VP came in in Nov and has cleared the budget for us to chance our feedback mechanism and make sure he and the decision makers in the team can discuss this and also get general feedback face to face. It's taken a bit of planning but the fact Tobias and others gave so much detailed information means we are flying Tobias and a number of other key customers/community leaders/partners into one place at NVIDIA HQ at our expense where they can talk face-to-face with execs, business, development, test etc... I've just met Tobias in person at his hotel and he's hear safely and I'm pretty sure willing to convey your concerns.

Neither of us is likely to be able to share details of the discussion for now at least but I hope it makes those who took the time and effort to reasonably discuss and raise their concerns some confidence that we aren't a faceless org.

This is also going to be a long term discussion about ensuring in future we communicate, listen, act etc better across the product.... a structure, mechanism aligned with our enterprise partners but probably new to GPU/hardware.... more on that to come ;-)

Do keep an idea on #GRIDdays on twitter... where you can keep an eye on Dr. Kreidl!

And let me know any comments, concerns etc..

Best wishes,
Racehl

Passthrough和RDSH使用场景都需要许可。
这同样适用于RemoteFX虚拟图形适配器和VMware的vSGA。
看待它的简单方法。
如果您使用GPU for Graphics，则需要相应的GRID许可证。



以下为原文



Both Passthrough and RDSH usage scenario's require licensing. The same applies to RemoteFX virtual graphics adaptor and VMware's vSGA.

Simple way to look at it. If you're using the GPU for Graphics, you will need the appropriate GRID license.

我们目前正在运行AutoCAD和Civil3D没有GPU，这是一个错误;
这个实施我们已经痛苦了大约一年。
我们期待着M6 / M60卡，并对许可成本感到非常失望;
许可证的成本远高于这些GPus的高端主机。
所以它要么去K2卡，要么使用别的东西。
无论它们有多好，我们都不会为M60制造商业案例。
我不知道这一切如何导致任何人继续沿着这条路走下去，至少就目前而言。
我们可以使用怪物PC为最终用户提供更高的计算能力，仅用于VDI中的GPU许可。
当然，我希望可以做一些事情，因为我认为M60卡可以带来用户的快乐，这是IT需要的。
:)我们现在正在寻找替代方案，也许一旦市场增长，竞争可以推动更高的产量并将成本降低到可接受的水平。



以下为原文

We are currently running AutoCAD and Civil3D w/o GPUs and this was a mistake; we've been suffering for about a year with this implementation. We were looking forward to the M6/M60 cards and were terribly disappointed in the licensing costs; licensing costing far more than the high end host for these GPus. So it's either go to K2 cards or use something else. We can make no business case for the M60s no matter how good they are.

I don't see how all this leads to enough volume of business for anyone to continue down this path, at least for now. We can provide far higher computing to the end user with a monster PC for the cost of GPU licensing alone in VDI. Sure hope something can be done as I think the M60 cards would bring user happiness, something IT needs. :) We are looking at alternatives for now, perhaps once the market grows the competition can drive higher volumes and lower costs to an acceptable level.

Passthrough和RDSH使用场景都需要许可。
这同样适用于RemoteFX虚拟图形适配器和VMware的vSGA。
看待它的简单方法。
如果您使用GPU for Graphics，则需要相应的GRID许可证。
好的，谢谢杰森。
这是一个有趣的区别，因为它意味着许可实际上不是关于使用vGPU技术，而是关于将硬件用于除计算之外的任何事情。
应该在这些场景中使用哪些许可以及如何实施？



以下为原文



Both Passthrough and RDSH usage scenario's require licensing. The same applies to RemoteFX virtual graphics adaptor and VMware's vSGA.

Simple way to look at it. If you're using the GPU for Graphics, you will need the appropriate GRID license.

Okay thanks Jason. That's an interesting distinction, as it means the licencing isn't actually about using the vGPU technology, it's about using the hardware for anything except compute.

What licencing is supposed to be used in these scenarios and how is it enforced?

嗨，BravoNovember：
需要运行单独的专用许可证服务器，以分配GRID 2.0使用的许可证。
如果您熟悉它，则与Citrix许可证使用的类似。
许可证分配在VM启动后立即发生，它将查询许可证服务器并分配令牌，具体取决于分配给它的许可证类型。
这对于使用vGPU的VM以及GPU直通都是如此。
我今天还希望有时计划在#GRIDdays帖子上发布有关会议的一些信息（当然，在NDA的限制范围内）。



以下为原文

Hi, BravoNovember:
A separate dedicate license server needs to be run that allocates licenses used by GRID 2.0. It is similar, if you're familiar with it, to the one used for Citrix licenses. The license allocation takes place as soon as a VM is booted up and it will query the license server and allocate a token, depending on the type of license assigned to it. This is true for both usage of a VM with vGPU as well as with GPU passthrough.
I will also plan hopefully sometime today to post to the #GRIDdays post with some information about the conference (within the constraints of NDA, of course).

不知道NVIDIA对这些高昂的每个许可证成本有什么看法。
我现在正与AMD联系购买他们的S7150 S2卡。
它们似乎更符合可靠的商业案例！



以下为原文

Dont know what NVIDIA has i mind with these high per license cost.
im in contact now with AMD for their S7150 S2 Cards.
They seem much more in line with a solid business case!

ArminL  - 您应该查看#GRIDDays博客文章并重新评估。
我想你对S7150不会印象深刻。
有时等待可以为患者带来回报。



以下为原文

ArminL -- You should take a look at the #GRIDDays blog post and re-evaluate. I think you will not be very impressed with the S7150. Sometimes waiting a bit can pay off for the patient.

我实际上认为人们应该关注并评估AMD技术;
考虑所有可用的选项是很有意义的。
AMD的方法（目前看来）看起来是一个非常具有成本效益的选择，因为它只是硬件购买，没有软件许可或年度支持。
AMD技术与NVIDIA GRID完全不同，但这两种方法各有利弊。
如果您需要Quadro驱动程序和特定于应用程序的支持，NVIDIA GRID一直都是如此，但如果您只是在寻找硬件加速虚拟桌面，那么AMD方法可能就足够了。
这项技术虽然是全新的。
该行业需要独立顾问（如那些参加#GRIDDays的人）来评估AMD技术和商业广告以及vGPU以帮助人们做出决策。
虽然我是NVIDIA的坚定拥护者，但我很高兴AMD正在进入这个市场 - 需要竞争来推动技术发展，并确保NVIDIA能够响应客户反馈。



以下为原文

I actually think people should look at and evaluate the AMD technology; it makes good sense to consider all of the options available to you. The AMD approach (as it stands) looks to be a much cost effective option because it is just a hardware purchase, there is no software licencing or annual support.

The AMD technology is quite different to NVIDIA GRID, but both approaches have their pros and cons. If you need Quadro drivers and application specific support, NVIDIA GRID all the way, but if you're just looking for a hardware accelerated virtual desktop, the AMD approach might be sufficient.

This technology is brand new though. The industry needs independent consultants (like those who attended #GRIDDays) to evaluate the AMD tech and commercials alongside vGPU to help people make decisions.

Although I'm a strong advocate of NVIDIA I'm pleased that AMD are entering this market - competition is needed to help push the technology forward, and to ensure that NVIDIA are driven to respond to customer feedback.

不到48小时，直到NVIDIA GPU技术大会！
我赢了;亲自去那里，唉，但是我会留意老鹰注意更新，并在发布信息时在这个博客上报告它们（开始一个新线程可能是一个好主意）。
您可以通过http://www.gputechconf.com/监控某些事件



以下为原文

Less than 48 hours until the NVIDIA GPU Technology Conference! I won;t be there in person, alas, but I will be keeping an eagle eye out for updates and report them on this blog as information is released (it will probably be a good idea to start a new thread). You can monitor some events via http://www.gputechconf.com/

NVIDIA GRID 2.0的新许可模式和定价：http：//www.nvidia.com/object/nvidia-grid-buy.html
更多细节要遵循......



以下为原文

New licensing model and pricing for NVIDIA GRID 2.0: http://www.nvidia.com/object/nvidia-grid-buy.html

More details to follow...

托比亚斯打败了我！
以下是对更改的快速概述，以及一些已回答的问题。
https://gridforums.nvidia.com/default/topic/781/announcements/grid-update-announced-today/



以下为原文

Tobias beat me to the punch!

Here's a quick overview of the changes, along with a few questions answered.

https://gridforums.nvidia.com/default/topic/781/announcements/grid-update-announced-today/

杰森，
好哇！
伟大的新闻和对用户社区的巨大好处。
非常感谢所有参与NVIDIA的人员重新审视GRID 2.0许可模式。



以下为原文

Jason,
Huzzah! Great news and a huge benefit to the user community. Many thanks to all involved at NVIDIA for revisiting the GRID 2.0 licensing model.

非常欢迎你！
社区通过提供大量反馈和数据点，确实帮助塑造了新模型。
虽然不是每个人都有机会面对面参与，但我们把所有事情都带到了船上并做出了改变。
顺便说一句，新代码已经发布，知识库文章正在发布。
您现在可以在此处查看许可服务器在这些新版本中的工作方式的更改
http://nvidia.custhelp.com/app/answers/detail/a_id/4077/kw/grid%20licensing



以下为原文

You're most welcome!

The community has really helped to shape the new model by providing a lot of feedback and data points. Though not everyone got the chance to be involved face to face, we took everything onboard and made the changes.

BTW, the new code is out, and KB articles are being published as we go.

You can now see the changes to the way the license server will work in these new releases here

http://nvidia.custhelp.com/app/answers/detail/a_id/4077/kw/grid%20licensing