Hacker News Reader

Comment by ianbicking | Hacker News Reader

All these legacy answers don't really make sense for this Z17... it's a new mainframe supporting up to 64T of memory and specialized cores for matrix math/AI applications. I have a hard time believing that legacy workloads are calling for this kind of hardware.

It also has a ton of high availability features and modularity that _does_ fit with legacy workloads and mainframe expectations, so I'm a little unclear who wants to pay for both sets of features in the same package.

View on HN

The legacy workloads are in some sense legacy, but in another sense the term legacy is misleading because at larger shops with capacity growth there has been no shortage of modernization of the various frontends and addition of new logic into the existing programs. The application maybe has the same name and core business as it always has, but is still growing in size and under active development. The idea now is that you might as well do the same thing and build in AI inference to those existing applications. Which is something they started implementing first with the last generation Telum chip which added some basic tensor units. This time around they are adding vector extensions 3 (think round 3 kinda like how AVX evolved) and that tensor processing extension 2 to the instruction set. Plus they are selling a discrete chip now that uses those same instructions and connects over PCI, which will probably be more than enough given that the goal is to never train on mainframes and only inference.

I've heard that the AI features are used by banks for fraud detection. I guess some banks are also growing their transaction volume.

I agree that many mainframe workloads are probably not growing so what used to require a whole machine probably fits in a few cores today.

Thanks for mentioning. This makes sense.

You won't see mainframes doing AI training, but there is a lot of value in being able to do extremely low-latency inference (which is why they have their NPU on the same chip as the CPUs, just a few clock-cycles from the cores) during on-line transaction processing, and less timing-critical inference work on the dedicated cards (which are a few more nanoseconds away).

Additionally, IBM marketing likes the implication that mainframe CPUs are 'the best'. If you can buy a phone with an AI processor, it only makes sense that your mainframe must have one too. (And IBM will probably charge you to use it.)

They aren't the best, they are different, and the best in certain areas. In terms of cache size they are way out in front. That architecture has very very fat cores. That's the main thing that makes them different (not better or worse) than something like x86_64, is the focus on single thread performance and cache rather than more cores. If you have a transaction processing app that you can't shard and that scales mostly with single thread perf, then they are pretty well suited for that use case.

I'm speaking more to the marketing checklist sense, IBM wants features that the others don't quite have yet. This is luxury branding imo a great degree, customers are paying a lot for a super-premium product, so they need to think they are getting 'the best'.

In terms of actual functionality, no matter how good it is, its not price-competitive.

Not just cache sizes, but also tricks like using any available core-local cache as the next level "off-core" cache.

Reply on news.ycombinator.com