48GB L3 cache, really?
"L3 embedded DRAM cache is now 48GB, double that of the z11". Is that a typo?
The systems business is largely dominated by x86-based machinery these days, but Big Blue's mainframe unit is hanging in there after five decades and is still a bit of a mint. You put in $1bn every two years for hardware development, as IBM has done with the new System zEnterprise EC12, and you take out $7bn in ridiculously …
"The z12 chip is implemented in a 32 nanometer high-K metal gate process and fabbed by IBM itself at its East Fishkill, New York foundry."
Very interesting that IBM, king of all things outsourcing and offshoring, still makes the chips for its mainframes in-house. They should teach something like this in MBA school -- never give away stuff that makes you money or risk your intellectual property! It might keep those of us in the US and Europe employed for a few more years.
Then again, when you're talking about a 700% ROI, maybe they figure it's worth it to not let a competitor compromise some contract manufacturer. Also, since the mainframe system is highly proprietary and not well known in general IT circles, I'm betting IBM has more than a few hardware nerds on staff that they absolutely must keep, and don't dare let the customers see. :-) I actually work in an industry highly dependent on mainframes, and my experience is that recent grads have no concept of mainframe computing, either because it's a total black box accessed through a web page or (shudder) terminal emulator, or because it's complex.
Oh well, I guess we can't all go back to the glory days of Digital/Sun/IBM/SGI with the order-of-magnitude margins on hardware that let those companies do things like this. Some of the older folks in the circles I travel worked for these places in the 80s/early 90s, and said they never wanted for anything, simply because the companies were so cash-rich. Of course, the flip side of that is those $25000 workstations....
"or because it's complex"
Funny, I've always thought of Mainframe as being the simplest form of computing, and the one that most closely resembles what is taught at uni. Modern desktop by comparison is over complicated and no longer bears any resemblence to the basic algorithm and input output stuff taught in higher education. Yes the hardare is different and unfamiliar but the theory mostly applies if you catch people straight out of education.
IBM makes almost all of its chips and systems in the US. Power is made in Austin and Rochester, MN. z is in upstate NY as is some of their storage. Some of the storage is made in Tucson. If you want to buy 'made in the USA' systems, IBM is that provider. Other than low end x86, it is all made in the US including the chips.
True, people buy x64 servers for a few thousand bucks a pop and then wonder why they are not getting the Unix or mainframe red carpet service and whine about the downfall of customer service. x64 is the Ikea of IT, i.e. inexpensive, but you're on your own. You have to know what you are signing up for at the outset.
This feature strikes me as a primordial Watson systems programming expert system. It is only advisory and can only process console messages produced by z/OS at this time. It doesn't matter if zAware functions completely locally or "phones home" to a Watson infrastructure for advice (unless the network is out, of course). IBM mainframe marketing has long considered itself hindered by the need to "install a systems programmer" at every customer site. This feature will likely grow into a crucial capability for winning new customers.
"... It is nice to see a company is still taking some pride in system engineering instead of figuring out how to push x thousand more half-baked x86 white boxes off the truck...."
Yes, but a company truly taking pride in system engineering would sell cpus much faster than these slow Mainframe cpus. Any high end 8-socket x86 server rivals the biggest IBM Mainframe - when we talk about number crunching.
Sure, Mainframes have far better uptime and much more I/O, but cpu wise, they are far behind x86 cpus when we talk about computing power. There is a reason you dont buy Mainframes to crunch numbers - they lack the ability. Instead a cluster of x86 are much cheaper and deliver much higher performance.
Any one got the time and skill to do a proper artists impression of the paint job they should put on the HA1 when a customer orders it fully configured with 101 GCPs ? I'd like to see the black dots over a plain white facia rather than the London logo converted to shades of grack (grey/black) for those orders only of course.
Biting the hand that feeds IT © 1998–2019