Grid 2.0

马丁，
是的，我看过M60至少有一个来源（戴尔）。
我们得到特殊的教育定价，所以在这里讨论是没有意义的;
联系现在出售它的供应商以获取报价。
我的猜测是你支付的费用比M60和K2的价格要低一倍，这是不错的，因为你获得2倍的性能和M60与K2的两倍的机载内存。
至于XenDesktop / XenApp的并发用户定价，它还将取决于您的供应商 -  VAR可能比直接从Citrix获得更优惠的价格。
标价（至少在美国这里）我认为每个座位约500美元。
要使用除直通模式以外的GPU，您确实需要付费的Citrix许可证，或者如果您运行XenDesktop / XenApp Enterprise或更高版本，XenServer许可证对于托管VDI的XenServer计算机是免费的，并且允许
完全访问vGPU功能。
Microsoft许可策略过于复杂，无法在此讨论，但您确实需要为每个实例提供某种虚拟桌面许可。
再次，与合格的销售代表交谈可能是您唯一的希望。
他们中的一些团队花了很多时间帮助VAR在我们的案例中进行工作，至少需要几周时间。
为什么微软授权如此混乱是一个可能永远无法回答的问题！
我希望这至少有一点帮助。



以下为原文

Martin,
Yes, I have seen the M60 available from at least one source (Dell). We get special educational pricing, so it makes no sense to discuss it here; contact a vendor that sells it now for a quote. My guess is you'll pay quite a bit less than double the price for an M60 vs. a K2, which is not bad given you get 2x the performance and twice as much on-board memory on the M60 vs. the K2. As to concurrent user pricing for XenDesktop/XenApp, it will also depend on your vendor -- a VAR might be able to get you a better price than directly from Citrix. The list price (at least here in the US) is I believe around $500 per seat. To use the GPU other than in passthrough mode, you will indeed require either a paid-for Citrix license, or if you run XenDesktop/XenApp Enterprise or higher versions, the XenServer licenses are free for those XenServer machines that host the VDIs and will allow full access also to vGPU capabilities. The Microsoft licensing policies are way too complicated to discuss here, but you do need some sort of virtual desktop license for each instance. Again, talking to a qualified sales representative is probably your only hope. It took a team of several of them to help a VAR work it out in our case and at least a couple of weeks. Why Microsoft licensing is so confusing is a question that will probably never be answered! I hope this helps at least a little.

你好。
感谢您的回答，但联系许多销售代表和VAR非常耗时。
是否有任何公开的“成功故事”，每个座位的计算价格（HW + SW，CAPEX + OPEX）（例如简单的Windows VDI桌面，没有任何其他应用程序（> = 2cpu，> = 4gRAM）和~20,50,100个席位和
如果相关的教育，公共和商业部门）？
由于财务分析（硬件+软件，资本支出+运营支出）以及新的价格水平M60加上新的NVidia vGPU CCU许可证，它读取了大约90％取消的VDI项目，它是否会恶化到99％？
我试图剥离CAPEX（eBay for K1 / K2 GPU，瘦客户端覆盆子和开源）并消除OPEX（opensource），但最终CAPEX价格等于具有相同CPU + GPU功率的独立PC。
感谢您的意见，马丁



以下为原文

Hello.

Thanks for answer but contacting many sales representatives and VAR is time consuming.

Is there any public "successful story" with computed price per seat (HW+SW, CAPEX+OPEX) (for example simple windows VDI desktop without any other application (>=2cpu, >=4gRAM) and for ~20,50,100 seats and for education, public and commercial sector if relevant) ?

I read about 90% canceled VDI projects due to financial analysis (HW+SW, CAPEX+OPEX) with new price level of M60 plus new NVidia vGPU CCU license it is getting worse to 99% or not ?

I tried to strip down CAPEX (eBay for K1/K2 GPU, thin client raspberry based and opensource) and eliminate OPEX (opensource) but final CAPEX price is equal to standalone PC with equal CPU+GPU power.  

Thanks for your opinions, Martin

“用于CITRIX XENSERVER的GRID VIRTUAL GPU版本352.46,354.13 / RN-06927-001 / 2015年9月21日”：Tesla M60和M6支持带有Red Hat Enterprise Linux 7 / CentOS 7客户虚拟机的GRID vGPU（仅限K1而不是K1）
/ K2），作为Citrix XenServer 6.5的技术预览功能。
（只能再次预览！）
“GRID LICENSING / DU-07757-001 / 2015年9月”：NVIDIA GRID K1和K2 GPU不需要（NVidia）许可证即可运行vGPU。
这些功能是最终的还是将来可能会改变？
谢谢，马丁



以下为原文



"GRID VIRTUAL GPU FOR CITRIX XENSERVER Version 352.46, 354.13/RN-06927-001/September 21st, 2015": GRID vGPU with Red Hat Enterprise Linux 7 / CentOS 7 guest VMs is supported on Tesla M60 and M6 (only and not on K1/K2), as a technical preview feature on Citrix XenServer 6.5. (Only preview again!)



"GRID LICENSING/DU-07757-001/September 2015": NVIDIA GRID K1 and K2 GPUs do not require a (NVidia) license to run vGPU.

Are that features final or it may change in future ?

Thanks, Martin

马丁，
据我了解，您需要/获得技术预览的特殊许可证，以便能够访问vGPU功能。
您可能需要更详细地检查XenServer Dundee技术预览（XS 6.6）规范。
对于Dundee的最终版本，我很确定许可限制仍将如上所述。
但是，对于K1和K2来说，这对我来说是新闻，所以我们必须看看最终的许可最终会是什么;
这似乎是一个奇怪的选择，分割哪些GPU支持vGPU功能，哪些不支持。
至于Linux vGPU对VM的支持，在XenServer 7的最终版本中确实应该得到官方支持。



以下为原文

Martin,
As I understand it, you need/get a special license for the technical preview that makes it possible to access vGPU functionality. You might want to check the XenServer Dundee technical preview (XS 6.6) specifications in more detail. For the final Dundee release, I am quite sure that the licensing restrictions will still be as I stated above.
However, as for the K1 and K2, that's news to me, and so we will have to see what the final licensing will end up being; this would seem like an odd choice to split which GPUs support vGPU functionality and which do not. As to Linux vGPU support for VMs, that should be indeed officially supported in the final release of XenServer 7.

你好。
我现在不需要任何许可证。
我对新的NVidia CCU许可证感到惊讶（我还编辑了之前的帖子以澄清这一点）。
今年有两个新的vGPU许可证需要支付（1月份XenServer 6.5的新许可证模型，现在是NVidia的新许可证）。
我认为这种许可地狱有效地杀死了所有VDI项目。
现在，我还在等待来自NVidia的预售人员或VMware Citrix的任何人都可以总结每个座位的价格（参见上一个问题）。
谢谢，马丁



以下为原文

Hello.

I do not need any license for now. I am surprised by new NVidia CCU license (I also edit previous post to clarify that). There are two new vGPU licenses to pay in this year (new license model for XenServer 6.5 in January and now new license for NVidia). I suppose that this licensing hell effectively kills all VDI projects.

Now, I still waiting if anyone from presales people from NVidia or Citrix of VMware can summarize price per seat (see previous question).

Thanks, Martin

NVIDIA许可证？
至少使用我们的XS 6.5 SP1安装以及在XenServer / XenApp下运行的GRID K1和K2卡还没有看到。
如果M6 / M60的情况如此，我会非常惊讶，因为驱动程序随时可供下载。
我的猜测是该文件尚未更新以反映M6 / M60，因为它们是如此新颖。
我现在看看你对NVIDIA CCU许可的意思了......
我发现这个很好的博客谈到了NVIDIA的CCU许可，这是非常新的，我将不得不更详细地看一下：http：//www.poppelgaard.com/nvidia-grid-2-0
这个页面也直接来自NVIDIA：http：//www.nvidia.com/object/nvidia-grid-buy.html
让我看看我是否可以从我在NVIDIA认识的人那里获得更多相关信息。
同时，如果NVIDIA员工可以解决这个问题，那么我们很多人都会感兴趣，特别是在任何潜在的购买之前。



以下为原文

NVIDIA license? Have not seen that at least with our XS 6.5 SP1 install and GRID K1 and K2 cards running under XenServer/XenApp. I would be very surprised if this was the case for the M6/M60 as the drivers are readily available for download. My guess is the document hasn't been updated yet to reflect the M6/M60 because they are so new.

I now see what you mean about the NVIDIA CCU licensing...

I found this good blog that talks about the CCU licensing from NVIDIA, which is very new indeed, and I will have to look at this in more detail: http://www.poppelgaard.com/nvidia-grid-2-0
There is also this page from NVIDIA directly: http://www.nvidia.com/object/nvidia-grid-buy.html

Let me see if I can get more information on this from someone I know at NVIDIA. Meanwhile, if an NVIDIA employee could address this, it would be something of interest to many of us, in particular prior to any potential purchases.

是的，我在第一篇文章后发现了这个软件（我的第一篇文章是基于poppelgaard博客）。
软件包中有文档：
352.46-354.13，NVIDIA并网VGPU释放，notes.pdf
352.46-354.13-NVIDIA并网VGPU-用户指南.pdf
352.47-354.13，NVIDIA并网许可，指南.pdf
GRID许可证服务器发行说明.pdf
GRID许可证服务器用户指南.pdf
如果您购买M60 / M6并且未安装“许可证服务器”并且没有“获得有效的NVIDIA CCU许可证”，则包含以下内容：
需要许可的vGPU在获得许可证之前以降低的能力运行。
屏幕分辨率限制为不高于1280x1024，帧速率限制为每秒3帧，并且GPU资源分配受限，这将阻止某些应用程序正常运行。
获得许可后，将删除这些限制。
NVIDIA GRID K1和K2 GPU不需要许可证即可运行vGPU。
马丁



以下为原文


Yes, I found this software after my first post (my first post was based on poppelgaard blog). There is documents inside software bundle:

• 352.46-354.13-nvidia-grid-vgpu-release-notes.pdf
  
• 352.46-354.13-nvidia-grid-vgpu-user-guide.pdf
  
• 352.47-354.13-nvidia-grid-licensing-guide.pdf
  
• GRID License Server Release Notes.pdf
  
• GRID License Server User Guide.pdf
  

It contains following if you buy M60/M6 and does not install "license server" and does not have "valid NVIDIA CCU license acquired":

vGPUs that require licensing run at a reduced capability until a license is acquired. Screen resolution is limited to no higher than 1280x1024, frame rate is capped at 3 frames per second, and GPU resource allocations are limited, which will prevent some applications from running correctly. These restrictions are removed once a license is acquired. NVIDIA GRID K1 and K2 GPUs do not require a license to run vGPU.


Martin

感谢分享最新信息，马丁。
看来每个配置都不同，Virtual PC，Virtual Workstation和Virtual Workstation  -  Extended，最后一个需要支持GPU passthrough。
我将联系我们的供应商，了解有关定价的想法。
在vGPU模式下使用时，每个连接虚拟机似乎都会占用一个令牌，而如果在GPU直通模式（虚拟工作站 - 扩展许可证）中使用，则会将其分配给虚拟机（例如XenApp服务器）来实现此目的，但它是
不清楚这对于整个XenApp服务器是否足够，或者是否需要为通过此直通服务器连接的每个并发用户提供单独的许可证。
为什么世界上的公司普遍使许可变得如此复杂，而且无法提供可理解的许可程序解释？
也许NVIDIA可以提供各种许可方案的一些例子，这将是一个良好的开端，能够更好地与此相关。



以下为原文

Thanks for sharing that latest info, Martin. It appears each configuration is different, Virtual PC, Virtual Workstation amd Virtual Workstation - Extended, the last one is need for supporting GPU passthrough. I will contact our vendor to see about getting an idea about pricing.

It appears that each connecting virtual machine will take one token when used in vGPU mode whereas if used in GPU passthrough mode (Virtual Workstation - Extended license) it is assigned to the virtual machine, such as a XenApp server, to accomplish this, but it's not clear if that is sufficient for the whole XenApp server or if you need a separate license for each concurrent user connecting via this passthrough server.

Why in the world do companies universally make licensing so complicated and fail to provide understandable explanations of licensing procedures? Maybe NVIDIA could come out with some examples of various licensing scenarios, which would be a good start to be able to better relate to this.

是。
这是通过OEM与K1 / 2相同Nivida不直接销售特斯拉。
否 -  GRID 2.0仅限Maxwell（M60 / 6）和未来的架构。
它不会被添加到开普勒。
如果您在vGPU中需要Linux或4K显示器，则它仅为GRID 2.0。
定价是通过OEM和＆amp;
VAR与GRID经销商有关系。
功能是为每个用户许可的（根据运行的VM更容易想到这一点）。
有3种不同的功能集，因此您可以根据所需功能进行许可。
引导VM，它需要适当的许可证，关闭VM并将许可证交回。
直通需要许可证，对于RDSH类型工作负载（如XenApp）上的并发用户也需要许可证。
http://www.nvidia.com/object/grid-technology.html
不是。许可适用于GRID 2.0及更高版本。



以下为原文



Yes.



This is via the OEM's same as for K1/2 Nivida doesn't sell Tesla direct.



No - GRID 2.0 is Maxwell (M60/6) only and future architectures. It will not be added to Kepler. If you require Linux or 4K displays in vGPU, it's GRID 2.0 only.



Pricing is via OEM's & VAR's that have relationships with a GRID distributor.

Features are licensed for each user (easier to think of this as per running VM though). There are 3 different feature sets so you license based on the features required. Boot a VM, it takes an appropriate license, shut the VM down and it hands the license back.

A license is required for passthrough, and will also be required for concurrent users on RDSH type workloads such as XenApp.


http://www.nvidia.com/object/grid-technology.html




No. Licensing is for GRID 2.0 and onwards.

我会在几分钟内与戴尔和NVIDIA召开电话会议，并将回复更多信息。
确切的许可模式仍然在起作用，首先，GPU直通将如何处理与vGPU配置不同。
预计全部细节将在一周左右发布。
显然还会有特殊的教育定价。



以下为原文

I have a conference call with Dell and NVIDIA in a few minutes and will report back further information.

The exact licensing model is still in the works, above all, how GPU passthrough will be dealt with differently from vGPU configurations. Full details are expected to be released in about a week. There will apparently be special educational pricing, as well.

否 -  GRID 2.0仅限Maxwell（M60 / 6）和未来的架构。
它不会被添加到开普勒。
如果您在vGPU中需要Linux或4K显示器，则它仅为GRID 2.0。
为什么NVidia不支持Linux guest虚拟机用于K1 / K2 vGPU共享（K1 / K2不应该有任何技术问题支持它）？
谢谢你的回答，马丁



以下为原文


No - GRID 2.0 is Maxwell (M60/6) only and future architectures. It will not be added to Kepler. If you require Linux or 4K displays in vGPU, it's GRID 2.0 only.

Why NVidia does not support linux guest for K1/K2 vGPU sharing (there should not be any technical problem to support it for K1/K2) ?

Thanks for answer, Martin

我很困惑，因为我在NVIDIA网站上可以下载一般的Linux GRID驱动程序。



以下为原文

Am confused as I see general Linux GRID drivers available for download on the NVDIA site.

我认为这不是vGPU驱动程序，这是PCI直通驱动程序。
匹配编译的vGPU驱动程序位于虚拟机管理程序的vGPU包内。
NVIDIA-GRID-vGPU-XenServer-6.5-346.68-348.27.zip
Grid 1.x管理程序/主机包：
NVIDIA-VGX-的XenServer，6.5-346.68.x86_64.iso
NVIDIA-VGX-的XenServer，6.5-346.68.x86_64.rpm
Grid 1.x访客驱动程序（没有linux支持）
348.27_grid_win8_win7_64bit_english.exe
348.27_grid_win8_win7_english.exe
NVIDIA-GRID-vGPU-XenServer-6.5-352.46-354.13.zip
Grid 2.0管理程序/主机包：
NVIDIA-的vGPU-的XenServer，6.5-352.46.x86_64.iso
NVIDIA-的vGPU-的XenServer，6.5-352.46.x86_64.rpm
Grid 2.0来宾驱动程序（具有“有限”的Linux支持）：
354.13_grid_win10_64bit_international.exe
354.13_grid_win10_international.exe
354.13_grid_win8_win7_64bit_international.exe
354.13_grid_win8_win7_international.exe
NVIDIA-Linux-x86_64-352.47-grid.run
M.C>



以下为原文


I think that this is NOT vGPU driver this is PCI pass-through driver.
Matching compiled vGPU drivers are inside vGPU package of hypervisor.

NVIDIA-GRID-vGPU-XenServer-6.5-346.68-348.27.zip

Grid 1.x hypervisor/host packages:
NVIDIA-vgx-xenserver-6.5-346.68.x86_64.iso
NVIDIA-vgx-xenserver-6.5-346.68.x86_64.rpm

Grid 1.x guest drivers (no linux support)
348.27_grid_win8_win7_64bit_english.exe
348.27_grid_win8_win7_english.exe

NVIDIA-GRID-vGPU-XenServer-6.5-352.46-354.13.zip

Grid 2.0 hypervisor/host packages:
NVIDIA-vGPU-xenserver-6.5-352.46.x86_64.iso
NVIDIA-vGPU-xenserver-6.5-352.46.x86_64.rpm

Grid 2.0 guest drivers (with "limited" linux support):
354.13_grid_win10_64bit_international.exe
354.13_grid_win10_international.exe
354.13_grid_win8_win7_64bit_international.exe
354.13_grid_win8_win7_international.exe
NVIDIA-Linux-x86_64-352.47-grid.run

M.C>

我认为这不是vGPU驱动程序，这是PCI直通驱动程序。
M.C>
正确。
Passthrough提供GRID 1.2（k1 / 2）
vGPU for Linux仅适用于GRID 2.0。



以下为原文


I think that this is NOT vGPU driver this is PCI pass-through driver.

M.C>

Correct.

Passthrough is available in GRID 1.2 ( k1/2)
vGPU for Linux is GRID 2.0 only.

我听说过经销商的许可证费用，这是多么荒谬的钱。
如果目前的奖品是最终奖品，那么vGPU将花费很多。
对于许可证的数量，您几乎可以购买完整的工作站！
我们目前正在扩展我们的Citrix环境，但如果我们需要支付这些许可证费用，我们就再也没有商业模式......我们目前正在使用K2卡，我希望这些卡可以长期使用



以下为原文

I have heard the license cost from my reseller, what a ridiculous amount of money.
If the current prizes are the final ones, vGPU will cost way to much.

For the amount of the license you can almost buy a complete workstation!

We are currently expanding our Citrix environment but if we need to pay these license costs we have no business model anymore... We are currently using the K2 cards and I'm hoping these will stay available for a long time

对于线程回应有点慢的道歉，这是本周巴塞罗那的VMworld以及漫长的日子！
回答上面引用的问题。
您许可GPU启用主机上的用户数量，而不是整个环境。
因此，如果您的XenApp环境在启用GPU的主机上只有20个用户，则为用户提供20个许可。
还要记住，如果您不需要Maxwell卡的电源，您仍然必须选择K1 / 2，它不需要许可，因为它是GRID 1.2。
所以我建议除非你绝对需要M60 / 6的额外功率，否则你会坚持使用K1 / 2来安装XenApp＆amp;
RDSH工作负载。
由于在GRID 2.0上对XenApp的许可有很多关注。
那么请帮助我理解，今天有多少环境在耗尽CPU之前达到了K2的资源限制？



以下为原文



Apologies for being a little slow to respond on the thread, it's VMworld in Barcelona this week and the days have been long ones!


To answer the question quoted above. You license for the number of users on the GPU enabled hosts, not the whole environment. So if your XenApp environment only ever has 20 users on the GPU enabled host, that's 20 licenses for the users.

Also remember that if you don't need the power of the Maxwell cards, you still have to option of K1/2 which does not require licensing as that is GRID 1.2. So I would suggest that unless you absolutely need the extra power of the M60/6 you stick with K1/2 for XenApp & RDSH workloads.

Since there's a lot of focus here on licensing for XenApp on GRID 2.0. So help me understand, how many of the environments are today hitting the resource limits of the K2, before they run out of CPU?

嗨，杰森：
关键在于，如果M60的硬件成本增加约50％，那么K2的整体性能将提高2倍。
因此，假设每个物理服务器有2个GPU板，那么希望能够占到整个物理主机数量的一半。
只是为了玩一些相对数字，假设服务器价格为1万美元，K2价格为4,000美元，M60价格为6,000美元（不是实际价格，但足以说明情况）。
假设您需要4 x K2s，那么这是两台服务器和4 x K2板，总价值为36,000美元。
如果您只需要2个M60 GPU，那就是一台服务器，那么22000美元。
因此，如果基于K2视点，每台服务器的许可成本大于14,000美元或多7,000美元，那么使用K2配置会更好。
由于GPU直通定价显然会超过首先获得GPU直通所需的虚拟工作站扩展许可证的价格，因此每个M60引擎仍然需要其中一个，所以单独为每个服务器添加两个此类许可证会使您高于
即使没有额外的passthough许可证，服务器中的2xK2 GPU成本也是如此。
如果这些型号是正确的，换句话说，对于GPU直通，即使使用额外的托管服务器，M60甚至不会比K2 GPU的数量便宜一倍。
因此，在这种情况下，GRID 2.0 GPU直通许可模式需要彻底检修，否则任何人都不会给它作为选项的第二眼。
国际海事组织，这根本没有考虑到，在故意等待发布此产品之后，意外的许可证定价变得非常令人失望。



以下为原文

Hi, Jason:
The point is that if the M60 runs about 50% more in cost for the hardware for 2x the overall capabilities of the K2. So assuming 2 GPU boards per physical server, the hope was to be able to need overall half as many physical hosts. Just to play with some relative numbers, assume $10k for a server, $4k for a K2 and $6k for an M60 (not real prices, but close enough to illustrate the case).

Say you need 4 x K2s, so that's two servers and 4 x K2 boards for a total of $36k. If you just need 2 x M60 GPUs instead, that's one server, so $22k. Hence, if the licensing cost is > $14k or $7k more per server based on a K2 viewpoint, you're better off with the K2 configurations.

Since the GPU passthrough pricing will apparently be above and beyond that of the Virtual Workstation Extended license needed to get GPU passthrough in the first place, you still need one of these per M60 engine and so alone adding two such licenses per server puts you above the cost of 2xK2 GPUs in a server, even without the extra passthough licenses.

If these models are correct, in other other words, for GPU passthrough, the M60 doesn't even come close to being cheaper than double the number of K2 GPUs, even with the extra hosting servers. So in this case, either the GRID 2.0 GPU passthrough licensing model needs to be totally overhauled, or nobody will give it even a second glance as an option. IMO, this was not at all taken in consideration and after deliberately waiting to see this product released, the unanticipated licensing pricing turns out to be a real disappointment.

问候...
GRID 2.0许可问题是否有任何更新？
请把关于事情的一些消息告诉我们。
先谢谢你，



以下为原文

Greetings...
Are there any updates on the GRID 2.0 licensing issues? It would be nice to get some news on where things stand, please.
Thank you in advance,

大家好，
我想更多关于许可方面的内容，但是如果有关于如何使用Hyper-V的GRID 2.0的进一步消息呢？
已经有关于Azure（现在或很快）可用的GRID 2.0的公告，所以我的理由是这必须在Hyper-V上？
关于这个时间和地点的任何消息？
干杯，
戴夫C.



以下为原文

Hi All,

I'd like to here more on the Licensing side, but also if there is any further news on how GRID 2.0 is supposed to work with Hyper-V? There have already been announcements about GRID 2.0 being available (now or soon) with Azure, so my reasoning is that this MUST be on Hyper-V? any news on when and where this is ?

Cheers,
Dave C