Discussion in 'AMD x86 CPUs and chipsets' started by Frontl1ne, Sep 11, 2014.
Thats right. But lets not let facts get in the way of a good story lol
but they do need fixes for spectre which also have a performance impact
IIRC there were IPC improvements in Raven Ridge though I cant recall specifics (maybe infinite fabric, power state and/or cache efficiencies?????) and they will hopefully also incorporate these in to the next series of desktop chips which together with whatever clockspeed increases they get in the final samples (and not the engineering samples referenced in that article) will get a %10+ improvement
You just made that part up, Zen 2 is 2019 and is a new core uarch.......
CES slide scroll down to x86 Roadmap.
"Improves on Zen in multiple dimensions" that's not a completely new architecture that's shrunk and refined, both a tick and a tock in 1 release.
Yes that doesn't say what you said.
The core mirco uarch is different in Zen 1 to Zen2, Zen 3 will be the same thing again, it will not be a from scratch design. At earliest we will see a complete new uarch after Zen3 but more likely after a Zen4. By your logic skylake hasn't seen a major uarch change since the Pentium pro, because each generation only had small changes(PP, p2,p3,dothan, meron,Penryn, Nehalem, Sandy, haswell, skylake).
You've misinterpreted what I said, I actually agree with your last post on the architecture completely. We won't see a large redesign before Zen 3, but there's an interpretation out there where people think zen 2 will be the major change in the architecture. With all that's coming with zen 2 it's too much to deliver in too short of a timeframe for any company to expect an entire redesign.
I think you're misinterpreting things. Nobody is saying Zen v2 will be a new architecture, but it should add optimisations that increase IPC significantly.
Zen is a new architecture, and new architectures are not going to be perfectly optimised, so there will be a lot more low hanging optimisations to make compared to Intel's 6th/7th/8th generation architecture.
I think Intel will be behind in IPC compared to Zen v2. And GF's 7nm process will do really well, too.
Intel's 10nm process is definitely not doing as well as Intel wants. I've heard the first 10nm iteration will actually be behind their 14nm++ in performance.
I still dont understand how when shrinking in general nets improvements?
Not necessarily, particularly in regards to Fmax.
AMD were somewhat famous for struggling with node shrinks from a performance pt of view on SOI. Just about every node shrink except 45nm had lower FMax than it's predecessor. Meaning they were restricted to the lower power SKUs initially.
So, has zen lived up to the hype ? are AMD chips back in the game ?
Architecture wise, I think it exceeded expectations.
The 14nm process is mediocre in terms of clock speeds, but it does have good power efficiency.
I'm missing the overclocking ability of my 7700K rig. God I loved that thing.
With Zen, I've noticed a decent increase in productivity stuff, and a considerably lower power bill. Its like a Jaguar vs a Viper. The person who buys one, probably wouldn't be interested in the other. They both have good and bad points, but for most people only one of those two choices will be considered. I'm not nominating which is the Jag and which is the Viper, I actually don't have a choice in mind. I'm simply saying they are very similar in many ways, but very different in many ways too.. so they appeal to very different buyers.
I'm looking forward to the next iterations of Zen. The platform is young, the next chips will fit the current boards. I like my 1700 so far, though my chip isn't exactly a winner in the silicon lottery, nor is my ram . I have however discovered that this thing eats productivity stuff like video conversions and solidworks for breakfast already, so the next gen I think will be well worth the wait. For my purposes.. and that's the bottom line really, when it comes to the whole Intel v AMD rivalry. No longer is the AMD the figurative Fiat 500 that was formerly the appropriate analogy. It can now actually compete. Although to be fair, there is still a market for Fiat 500's too..
Zen is inferior in some respects to Intel, superior in others I believe. Who knows, the next gen may improve the inferior bits and surpass the best Intel have got, in ALL respects, at a lower price? Its certainly possible IMO.
Rapt in how much support AMD is getting right now. Even the Intel Fanbois ( which includes me ) have to appreciate that an actual viable competitor in the marketplace can only ever be good for us, the consumers.
Hey, how about those Vega graphics they plan on sticking into Intel chips huh? It all gets more interesting every day to me.
I loved amd since i saved up when i was a teen and bought my very own pc piece by piece and put it together an AMD Athlon XP 1800+ pally on an A7n266-vm motherboard and 256mb ddr ram @ 266Mhz
a bit of history http://archive.atomicmpc.com.au/forums.asp?s=1&c=6&t=2348
Now i'm using a dual Xeon 2670 system with 64gb of ram and it handles everything i throw at it effortlessly. But wondering if AMD is bringing anything to the table. Hoping for unlocking chips, overclocking and low power usage
by unlocking you mean cores right because every single amd chip is overclocked unlocked...just can't squeeze massive amounts out of them unless you are talking 1600/1700 etc which percentage wise are very good.
and now with the 18700x being around $480 it is middle finger to the 8700k at 60 plus more why have a 6/12 when for 60 less you can have a 8/16 AND when zen 2 or the refresh hits have a bios update and plonk it in.
now for my question.
autodesk fusion 360
I am about to put in a desktop with a 1600 for this application anyone use it or know anything I should look for in the settings (I know a lot of cad type programmes can suck with multi core but I have nio idea on this one)
why, because tthe 8700k is faster in most benchmarks by a considerable margin.
as for your 2nd question,
clock speed is king here go for the intel if you want the best performance for that application.
A ThreadRipper 1950X will give you the same 16 cores, with higher clocks, and lower power consumption than a dual Xeon 2670. Ryzen in general does have better power efficiency than Intel.
Of course, it has an unlocked multiplier, so it's overclockable. And if you want ECC RAM, ThreadRipper also works with it.
Well, it supports unregistered ECC (UDIMM), not registered ECC (RDIMM). I'd expect EPYC to support both.
not to mention the platform is a lot cheaper as well, and the performance hit after patching intel's security flaws.
For that use case depends if you're primarily doing rendering or modelling. If you're doing lots of rendering with it, lots or cores is great so an 8 core 1600 or 1700 is a good buy. If you're mainly doing 3d modelling but not much rendering say a 7600k or 8350k with a 5ghz O/C is the fastest option as it only uses 1 or 2 cores for modelling as cad programs typically do. refer:
hardwares unboxed review AFTER the meltdown patches between the 1800x and 8700k using a 1080ti this time
results were pretty much the same intel is still 20-30% faster in games.