Looking at those block diagrams properly, I am very happy as a developer to see that we are getting true 8-lane wide hardware in this generation. Whilst Bulldozer onwards have supported the AVX extensions the FP unit was only properly 128-bit wide so only 4 32-bits or 2 64-bit data types at a time. The simulation work i'm involved in wields quite a lot of 3+D vector calculus and while we revert to 32-bit floats where possible to cram one vector per SIMD op it'd be really nice to be able to use doubles and not lose any significant throughput.
Our standard compute server is a 4 * 16 core Bulldozer chips (we combine up to 4 of these servers) and some of the floating point paths in not frequently parallelised algorithms are frequently beaten by my 15W TDP Haswell MacBook Air which makes me sad.
Hardware with true 256-bit intrinsic support throughout would push me to using doubles everywhere and let us remove some rather dodgy precision hacks in places. Next wish from AMD: give me something like IACA to give me a way to micro-tune critical algorithm paths please
SkyNet Confirmed
Hmm, I think you're being a tad unfair there - as I've said above there's definitely still applications where the spread of many, high-clocked cores of an old AMD processor beats the more efficient/modern Intel. Heck, in my own household we've stopped using the SB i5's for Handbrake "rips" because my (much older) AMD 1090T can pretty much do them in half the time. In which case a more contemporary FX8250 would be amazing. I'm sure Cat could do "chapter and verse" on where AMD's offering are better...
Pardon my ignorance (last time I did vector calculus dev Intel was having trouble getting their chips to add up properly and the ole "Top supercomputers in the world" list was pretty much a Cray exclusive club), but wouldn't using graphics cards as compute engines be better - not only better FP hardware, but also dedicated vector units? Or at least according to the ex-colleagues who are still in that area (CAE). I'm just being nosy... (and a bit nostalgic for better times).
Anything integer/logic the current generation of AMD chips scream through. I love running bits of code like that through them.
GPGPU would almost certainly be the best way to go about modern vector work. But we don't really have the kit for it or any programmers trained in it. The majority of our researchers are also tied to IDL which will thread array calculations but not put them into a GPU kernel. It'd be nice to use GPGPU but I don't currently have the time or budget to spend on it. Plus, I can't imagine what the slowdown for 128-bit floating points calculations in GPU code is, given the often at least 4x slowdown going from 32-bit to 64-bit (if the kernels support this at all).
crossy (25-09-2015)
DanceswithUnix (25-09-2015)
I don't want to buy into the hype yet. It would be very welcome, but high expectations is half the equation to disappointment
I wonder if AMD has any ambition in challenging Intel in the laptop market though, and whether Zen can do it.
The 8350 really isn't a bad cpu - so long as you don't care about its power draw. For the money it's actually pretty ridiculously good value and will handily beat an i5 in a lot of stuff.
I think the main issue with them is that no gamer with basically any i5 would swap to one, and that's a lot of people. Intel are kinda finding the same issue with those of us who stubbornly refuse to upgrade on Sandy Bridge.
It's gonna take 8 good cores and a reasonable price, or 4 cores + HT at a reasonable price before I even think about a new CPU, and I'm pretty sure I'm not the only one thinking that.
I do hope Zen brings consumers a major performance boost and AMD some good fortune. AMD of the last 10 years has been about coming late to the party, matching their competition and some cases beating it by small fractions then the competition releases their next gen putting AMD behind by a year again.
They need to change their philosophy back to early 00's and not only need to meet the competition but smash it. Intel has trickle fed updates since Nehalem and could have quite easily released better CPUs but they've been graced with little competition for too long milking minor updates. 9 years on and they're still only 4 core on mainstream with no sight of change...
AMD do at least sit closer to nVidia with their GPUs but still always late to the party
At this point I don't believe a word that AMD says. All the hype and bs coming from them before a new product release only for the product to be meh or just rubbish has made me lose all faith in them. I really hope (and we really need them to be competitive) they can turn it around with Zen but I shall wait until they actually launch before getting interested.
Last I heard Asmedia where designing future Amd chipsets. http://www.kitguru.net/components/cp...g-deal-report/
Main PC: Asus Rampage IV Extreme / 3960X@4.5GHz / Antec H1200 Pro / 32GB DDR3-1866 Quad Channel / Sapphire Fury X / Areca 1680 / 850W EVGA SuperNOVA Gold 2 / Corsair 600T / 2x Dell 3007 / 4 x 250GB SSD + 2 x 80GB SSD / 4 x 1TB HDD (RAID 10) / Windows 10 Pro, Yosemite & Ubuntu
HTPC: AsRock Z77 Pro 4 / 3770K@4.2GHz / 24GB / GTX 1080 / SST-LC20 / Antec TP-550 / Hisense 65k5510 4K TV / HTC Vive / 2 x 240GB SSD + 12TB HDD Space / Race Seat / Logitech G29 / Win 10 Pro
HTPC2: Asus AM1I-A / 5150 / 4GB / Corsair Force 3 240GB / Silverstone SST-ML05B + ST30SF / Samsung UE60H6200 TV / Windows 10 Pro
Spare/Loaner: Gigabyte EX58-UD5 / i950 / 12GB / HD7870 / Corsair 300R / Silverpower 700W modular
NAS 1: HP N40L / 12GB ECC RAM / 2 x 3TB Arrays || NAS 2: Dell PowerEdge T110 II / 24GB ECC RAM / 2 x 3TB Hybrid arrays || Network:Buffalo WZR-1166DHP w/DD-WRT + HP ProCurve 1800-24G
Laptop: Dell Precision 5510 Printer: HP CP1515n || Phone: Huawei P30 || Other: Samsung Galaxy Tab 4 Pro 10.1 CM14 / Playstation 4 + G29 + 2TB Hybrid drive
There are currently 1 users browsing this thread. (0 members and 1 guests)