Slashdot Mirror
AMD Unveils First Zen Desktop Processor Details, Picks 'Ryzen' To Brand Zen CPU (hothardware.com)
MojoKid writes from a report via HotHardware: AMD has just officially unveiled that desktop variants of its Zen processor family will now be branded RYZEN. Zen-based processors will eventually target desktops, servers, and mobiles device, but the first wave of products will be targeted at the performance desktop market, where gamers and VR continue to spur growth. AMD is positioning RYZEN as a high-performance option and though there will be other core configurations as well, AMD has disclosed that one of the high-end options in the initial RYZEN line-up will feature 8 cores (16 threads with SMT) and at minimum a 3.4 GHz base clock, with higher turbo frequencies. That processor will also be outfitted with 20MB of cache -- 4MB of L2 and 16MB of L3 -- and it will be infused with what AMD is calling SenseMI technology. SenseMI is essentially fancy branding for the updated branch predictor, prefetcher, and power and control logic in Zen. AMD's upcoming AM4 platform for RYZEN will be outfitted with all of the features expected of a modern PC enthusiast platform. AM4 motherboards will use DDR4 memory and feature PCIe Gen 3 connectivity, and support for USB 3.1 Gen 2, NVMe, and SATA Express. Performance demos of RYZEN shown to members of the press pit a stock Intel Core i7-6900K (3.2GHz base, 3.7GHz turbo) with Turbo Boost that was enabled on the 6900K, versus RYZEN with boost disabled running at 3.4GHz flat. In the demo, the RYZEN system outpaced the Core i7-6900K by a few seconds.
The top SR7 is rumored to be around $500. Make of that what you will. Some think that this is very optimistic if the performance levels are what they seem to be. It would be very good for the consumer, though, obviously.
Ezekiel 23:20
The point is that HSA could more beneficial for this "lower-cost equipment" you (and I) are talking about: the unified virtual memory space on Carrizo and later AMD APUs makes for true zero-copy iGPU code execution! That's the very reason why I'm mentioning this. It could easily replace Intel's AVX advantage, even when considering wider Intel's AVX implementations (AVX512) and shorter-running kernels (because of the very low overhead of in-place data manipulation by the iGPU).
Ezekiel 23:20
Honestly, there's no real benefit of Intel vs. AMD in the core performance games. They're both fast enough and if you need real performance, that's where Xeons come in with 22 cores per chip and 4 chips per motherboard... if you can afford it.
There is a lot more to CPUs than just CPUs.
Chipsets:
AMD hasn't had a decent chipset on the market for years. Even though all modern CPUs tend to put the majority of the functions within the CPU itself, there are still things like the actual chipset to think about. Add to that that in my experience, the reference platform designs from AMD for their power circuitry are generally horrible and there's a real problem.
Motherboards:
What good is an awesome CPU if the motherboards are generally just crap. Motherboard manufacturers tend to make one or two AMD motherboards per generation as a token gesture. They don't expect to sell the volumes, so they slap together whatever crap they can get running and put some pretty colors into the slots and heatsinks and call it "Blaster" or something else that's horrible and peddle it off without a BIOS update ever to be seen.
Development tools:
AMD's developer website is all graphics and no CPU. To write proper software for modern hardware, it's necessary to optimize code properly. While CPUs are fast enough, good information on CPU tuning can make a huge difference. Intel for example has released extensive information on their architecture that describes the CPU front end instruction set translator well enough that it's possible to write compilers and JITs that will allow inter-core/process communication without the need for spinlocks or mutexes. Intel also documents all the information necessary to fine tune memory access by placing code on the right cores in order and taking advantage of the CPU ring bus to optimize performance and minimize cache coherency issues. AMD does not.
Where is the AMD optimizations to CLANG or Mono, etc? Where's the optimization settings within Visual Studio or GCC for AMD? These are things that they should be working on to ensure the best performance on their platform.
AMD probably made a great CPU... if we can get into a "FlaskMPEG" style performance war like we once had, it would be amazing.