OMG - you've invited the "car analogy" haters to take over the thread!
[doublepost=1457831326][/doublepost]
HP sells workstations, but no servers.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
2016 nMP
- Thread starter ManuelGomes
- Start date
-
- Tags
- nmp
- Sort by reaction score
- Status
[doublepost=1457831326][/doublepost]
I think, perhaps, you're reading my use of "niche" as somehow negative. I don't meant it to be. I'm just describing what I see, which is that compared to overall PC sales and revenue, workstations account for maybe 2%, if we're being generous. I think this is important particularly for Apple (and this is an Apple board) because, as I've said, I don't think Apple does niche markets any more unless they see large growth potential. To me, this explains Apple's lack of interest in the nMP.
I'd be careful about assuming that workstation sales are a huge part of that profit pie. HP's enterprise sales include tens of thousands of greige boxes and cheap laser printers, plus large format printers and the like. I am not convinced their workstation sales are anywhere near the biggest piece of that pie.
Anyway, like I said, I don't mean 'niche' in any negative term. I just don't know what else to call a sales channel which is tiny compared to general PC or mobile device sales.
[doublepost=1457833053][/doublepost]
https://www.hpe.com/us/en/servers.html
Another "can't see the forest for the trees" post.
Note that the URL domain is not "www.hp.com".... You won't find servers at "www.hp.com".
HP split into two pieces: workstations are with PCs and printers ("HP"), while servers and services are a separate company ("HP Enterprise").
Are you actually trying to argue that HPE is somehow separate from HP? Because I wouldn't know if you were serious.
[doublepost=1457834413][/doublepost]
On paper only. All the money goes into the same pot.
You are so completely wrong here that it boggles the mind.
HP.com and HPE.com are legally and operationally separate corporations.
HP is "HPQ" on the NYSE.
HPE is "HPE" on the NYSE.
The SEC has a whole bunch of regulations against separate publicly traded companies putting the money "into the same pot".
I placed a $350K order for HPE servers towards the end of November. Our purchasing department sent the P.O. to HP.com. In early December I asked HPE for a tracking update. HPE said "we have no record of that P.O.". That's when we realized that the P.O. had been sent to the wrong corporation. HP.com said "this means nothing". We tried to send the P.O to HPE.com - "this P.O. is for a different company, it means nothing".
I had to start over on the order (get new quote, create requisition, create new P.0.) - but by then the quarterly order window was closed. I had to work with the finance people to move the $350K allocation from the December quarter budget to the March quarter budget before proceeding. (If you understand that sentence, you know what a headache that can be.)
HP.com and HPE.com are legally and operationally separate publicly traded corporations. It is stupid to claim otherwise.
Last edited:
I assumed it was your standard corporate restructuring, shifting the names so you can shift the blame.
I have enjoyed your posts so far, but this is incorrect. Ferrari is very profitable, and has been for many years: http://www.marketwatch.com/story/ferrari-profit-jumps-62-2015-10-28
That's not what I said. Ferrari makes a lot of money licensing its name to video games, sportswear companies, the Ferrari theme park in Abu Dhabi, pens and even your very own carbon fiber license plate holder. Hard numbers are difficult to find, but this article, from 2010, says, "Ferrari estimates that the retail value of all Ferrari products excluding the actual cars – merchandise and licensed products – totals US$1.5 billion annually worldwide," a year when their net revenues were only about €1.9 billion. Additionally, Ferrari, like Porsche and Lotus, provides engineering services to other car makers, and they make money selling their Formula One engines and engineering expertise to other Formula 1 teams. Selling cars is only part of what they do. What Ferrari really sells is their brand, whether its attached to a pen, a car, a jacket or a Formula 1 team.
It is possible that's changed, and should you find more up-to-date figures please show them to me. But I don't see them abandoning something which brings in so much money, especially when they sell so few cars.
I wouldn’t even call them the Ferraris of the computer world anyway. Lots of people on lower incomes own Macs iDevices. They can’t afford a Ferrari.
The problem with these comparisons is the computer landscape has shifted terribly in the past 15 years. iMacs were once pretty weak education and consumer machines, and laptops were even slower and crappier, or else huge and obnoxious to lug around if they were desktop replacements.
Now, even Apple's previous "ultra thin" Macbook Air is powerful enough to do studio audio processing, the iMacs are more powerful in many respects than pro computers in day-to-day and specialized pro tasks, people actively say "that 15" rMBP is too big and heavy for me", and the price-to-performance you get for your extra cash on a pro workstation has greatly diminished. I think you're correct that the "prosumer" market definitely made PowerMacs more viable, but PowerMacs were compared to the Mac Pro far more of a prosumer-to-pro product anyhow (you could get them starting at $1800–1900 in today's dollars.) If an iMac can get you 75% of the speed of a workstation for some percentage less than 75% you're going to see the market erode or at least fail to expand, and what you're left with is the people for whom (comparatively) money is not the object—speed is, because even that 5% speed difference you paid an extra $1000 for matters for you (this is also ignoring the theoretical benefits of pro tech being top-shelf and a bit more reliable, although I dunno how much that is actually true across the PC spectrum.)
To those claiming the nMP isn't a workstation, please let me know which apps in OSX (OSX Only) benefits from dual cpu or quad gpu?
Do the nMP/osx is used for protein folding, or high frequency trading, or something heavier than Arithmetica or video editing, and of course 3D rendering.
Few pro's on OSX are concerned about a 2nd cpu, and while is somewhat valid to cry for CUDA we should recognize CUDA as the evil closed GPGPU platform we need to destroy with OpenCL, and actually the nMP is up for all the current OSX applications, and while don't offer the best 3D rendering performance it's true many pros actually rent by hours a Renderfarm For that (also those with more powerful hw use to rent gpu time when they need an final rendering).
So I think Apple is selling the workstation that OSX users need (at least at launch, as current the nMP is outdated).
A mac pro is an studio machine not an industrial Bulldozer, for those exists multi cpu workstation, compute severs and so on.
About market decline, it's has more related to delay on sensitive updates than on actual demand decline, Compute performance market moves as the improvements are released, now the industry is waiting for Xeon E5v4 as well new gpu and the new Xeon phi to release orders that's true from the Mac Pro to super computer orders both waiting for Intel/AMD To release that over promised improvement, mostly gpu and mic for compute are delaying new computer servers and workstations orders (the same reason why we don't buy a new mac pro today).
So I consider this specialized market grown as the economy growns and buy as soon new technology is released not in a regular timing as other markets but as soon they can profit from the new technology.
Few specific things are waiting hp Dell and supermicro customers: 22 core Xeon E5v4 and new Xeon Phi knights landing as well the new nVidia and AMD Pro GPU specifically nVidia is almost 2 year delayed since announced Pascal. Ok that is in short term, in mid-long term HPC should move from x86-64 to ARM, PowerPC also Intel should relaunch Itanium At least for HPC the road end for x86 it's visible now.
Do the nMP/osx is used for protein folding, or high frequency trading, or something heavier than Arithmetica or video editing, and of course 3D rendering.
Few pro's on OSX are concerned about a 2nd cpu, and while is somewhat valid to cry for CUDA we should recognize CUDA as the evil closed GPGPU platform we need to destroy with OpenCL, and actually the nMP is up for all the current OSX applications, and while don't offer the best 3D rendering performance it's true many pros actually rent by hours a Renderfarm For that (also those with more powerful hw use to rent gpu time when they need an final rendering).
So I think Apple is selling the workstation that OSX users need (at least at launch, as current the nMP is outdated).
A mac pro is an studio machine not an industrial Bulldozer, for those exists multi cpu workstation, compute severs and so on.
About market decline, it's has more related to delay on sensitive updates than on actual demand decline, Compute performance market moves as the improvements are released, now the industry is waiting for Xeon E5v4 as well new gpu and the new Xeon phi to release orders that's true from the Mac Pro to super computer orders both waiting for Intel/AMD To release that over promised improvement, mostly gpu and mic for compute are delaying new computer servers and workstations orders (the same reason why we don't buy a new mac pro today).
So I consider this specialized market grown as the economy growns and buy as soon new technology is released not in a regular timing as other markets but as soon they can profit from the new technology.
Few specific things are waiting hp Dell and supermicro customers: 22 core Xeon E5v4 and new Xeon Phi knights landing as well the new nVidia and AMD Pro GPU specifically nVidia is almost 2 year delayed since announced Pascal. Ok that is in short term, in mid-long term HPC should move from x86-64 to ARM, PowerPC also Intel should relaunch Itanium At least for HPC the road end for x86 it's visible now.
CUDA is not an open std also is vendor closed, while Opencl can be implemented on any cpu gpu also FPGA, and re-writing Cuda code for opencl is almost trivial.
I'm comparing CUDA With opencl.
Ah, FYI OSX is open source but not copy left, you can download a day audit full OSX code but legally you can't reuse this on other projects (only those portions under MIT or similar license).
No you can't download all of OS X source you can only get what's MIT'd which is just a part of the kernel.
CUDA is provided to every platform by nVidia while its not open it's available to all.
That makes you depend on nVidia, and is not available for AMD GPU neither CPU, neither FPGA, OpenCL can be implemented on anything not only pc, also you can run Opencl on phones also (theoretically) on raspberry Pi or wherever there is an gpu with drivers or an cpu.
I find it ironic that a post in an Apple forum is critical of depending on a single vendor.
Doubly so in a thread where there is concern that Apple is neglecting the high performance workstation market.
Nvidia doesn't have an open licensing system for other platforms. While it's not illegal to provide a compatibility API, don't confuse that for being "available to all." Nvidia would never willingly license CUDA to AMD.
It's hurt them a bit because OpenCL can run on CPUs too and CUDA couldn't really. That's why the only other platform port that I know of that's been blessed by Nvidia is the x86 version of CUDA that was done by third parties.
I should have said cross platform as it's what I meant. It's there for Windows, OS X, and Linux.
If you ask about CUDA, I genuinely suggest doing some research and educating yourself about what is happening in CUDA world. Let me put some things this way: Otoy is doing CUDA cross platform, cross vendor , to serve some purpose. There is huge background story happening, and happened when AMD launched one thing: GPUOpen. Nvidia is afraid of the impact it can have on their sales, and mindshare. What is more, there is huge hint in that story about who will dominate HPC in upcoming months in terms of raw performance(Pascal, or Polaris/Vega).
Nvidia may want to "open" CUDA, to close it again with Volta. That will lock whole market to their solutions. Unless - someone will come with dynamic compilers CUDA - OpenCl, that can work on this job, while working on project at the same time.
Nvidia may want to "open" CUDA, to close it again with Volta. That will lock whole market to their solutions. Unless - someone will come with dynamic compilers CUDA - OpenCl, that can work on this job, while working on project at the same time.
I agree what Otoy is doing with the CUDA cross compiler is of the highest importance (if it works and especially if it is licensed industry wide) .
Similar to what the "clone" makers of PCs did in the very early days (1980s) with using legally compatible BIOS version chips reverse engineered from the original IBM ROM BIOS . This ushered in the age of relatively inexpensive personal computers , cheaper than the Apple II and genuine branded IBM gear .
Unless nVidia and AMD upgrade their architecture to enable data access across the cores as on MIC (Xeon phi), I think this will be the last generation of cuda/opencl GPGPU, MIC architecture allows full algorithm freedom not just offloading some functions, that's why Xeon-phi is preferred on HPC Unless the application can be tweaked to run GPGPU, MIC rules.
Another revolution about MIC Comes from ARM based implementation as you know Xeon phi actually are 60+ atom 27xx cores on a single fabric, ARM vendors are seeking on the compute server market offering an integral solution (more less as the new Xeon phi knights landing) with cpu based on ARM v8 from 48 to 72 or more cores, Qualcomm seems is reading something based on its kyro core others are working on cortex a72 or a53 the only public known results are promising reaching 1teraflop on less than 1/4 the power used by an Xeon phi, Qualcomm still no comment on their peak performance or actual core count, as reference a single cortex a72 delivers about the half flops as an Xeon core on a dismal % of the power (kyro supposedly is superior) , further those chips will not suffer from Export bans from you known.
Would be interesting Apple building a similar product based on Ax cores, would be a dream having a Mac Pro with 2 ARM MIC x 128 Ax cores instead AMD GPUs ...
- Status