AMD Ryzen

[Zoraptor]

1700 is 8 core and 65W since that's what I have. You could probably get away with using the stock cooler as well, even with the 8 cores that 1700 is under 40 degrees (ambient +20) at normal use and barely touches 60 for most games. So much better than the shredded coke can screwed onto an aircraft turbine that was my old stock Intel cooler.

 

The 2700x vs 8700k tests were done with a 1070 as the graphics card, so not the usual methodology and the test set up will be gpu rather than cpu limited. Having said that, a 1070 is a more real world scenario than testing 1080p low details on a 1080Ti or a Titan as is usually done. Given indications are of a roughly 10% boost in performance for Ryzen 2 it will probably still be a little behind a fully overclocked 8700k due to the remaining clock disparity, though Ryzen's IPC will be higher if the meltdown/ spectre patches are applied on the i7.

[Valsuelm]

In all seriousness... a 1080p gaming race is a near meaningless test of those processors, no matter what graphics card you pair them with.

Edited April 17, 2018 by Valsuelm

[Zoraptor]

It's certainly meaningless as a practical demonstration of real world performance, yes. But they have to use 1080p low details with a 1080Ti if they want to test the CPU, it's the only way to make the CPU the bottleneck and even though it isn't in any way a real world scenario. When they test 4k there is very little difference between most mid to high range CPUs because it would be the 1080Ti limiting performance with a few exceptions where thread limits and the like are hit (eg Farcry 3) on i5s especially the 4 thread ones. In the wccftech article case the 1070 would be limiting performance even at 1080p for both CPUs.

 

Alternatives like synthetic benchmarks have their own set of problems with being 'real world', of course, since they're synthetic. And the focus on single thread performance is starting to get a bit old, while it is still mostly relevant now programs will only get more multi threaded, they'll never get less, and the fewer threads you have the quicker your CPU will choke.

[Zoraptor]

Ryzen 2 have been out for a couple of days now, and it's a decent improvement but not game changing. Lots of minor improvements to memory latencies, a decent bump to clock speed and the X processors coming with a (pretty good, approaching 212Evo class) stock cooler are the main changes, and a positive is that many of the new benefits are on the chip itself so will work on 300 series motherboards. The 2700X* is within 5% of a 8700k in gaming with expensive RAM and some fiddling, and given that you can get its performance without an expensive cooler and delidding- plus use a cheap MB and still get overclocking/xfr2 if you want- it's also a decent amount cheaper. It still hits a voltage wall, but at least it's approaching 4.4Ghz rather than 4.0 and XFR2 will boost to 4.35 automatically making overclocking unnecessary for many use cases.

 

*practically, a 2600/X would be better value for money for most people anyway

 

Definitely won't be upgrading my 1700, not enough improvement so I'll wait for Zen2/ Ryzen3.

[Keyrock]

The Ryzen 2 chips, according to what I've read, are clearly the best value on the market in their respective price ranges, neck and neck with their Intel counterparts in gaming and flat out beating them in most other real world tests, especially highly threaded ones.  If nothing else, this has forced Intel to bring 6 core, and soon 8 core with the upcoming new Coffee Lake flagship chip, to the mainstream market, something Intel probably wouldn't have done had AMD not stepped their game up and started competing in the mid-range and enthusiast sectors again.

 

My i7-4790K is still getting the job done, so I won't be getting these, but I am eyeing a new build somewhere roughly 1 year from now, which will make my current setup about 5 years old, which seems like a reasonable length of time to run a setup.  It's good to see renewed competition from AMD and this bodes well for what will be on offer when I do build a new rig.

[M4xw0lf]

I'm strongly eyeing an upgrade to a 2700X system. No X470 (or 400-anything) µATX boards so far, though, which is a must for me. Also ram prices continue to suck... but that can't be helped.

[Valsuelm]

I'd have already pulled the trigger if it wasn't for RAM and GPU prices...
 

Paying more for those two things than everything else combined is not something I've yet been able to swallow.

 

Perhaps if I down enough whiskey first...

[Bartimaeus]

GPU prices have been falling a lot in the last couple of weeks, though RAM prices are still absurdly terrible.

[Zoraptor]

A bit peripheral to Ryzen itself but, Jim Keller who was perhaps the most important contributor to the Zen architecture (and earlier K7/8/Athlon/ x64) has joined Intel after finishing work at Apple. Which is exceptionally good news for them since their main line CPU division seems to be having some significant problems.

 

GPU prices have been falling a lot in the last couple of weeks, though RAM prices are still absurdly terrible.

 

Yeah, the RAM prices are cartel pricing, at some point they'll get smacked by Europe again or provoke the Chinese to start producing large scale. Lots of gpu oversupply rumours going around now, but I'm not expecting too much since high RAM prices inflate GPU prices as well. Should be some 2nd hand bargains though, but mixed in amongst some overworked dross.

[M4xw0lf]

A bit peripheral to Ryzen itself but, Jim Keller who was perhaps the most important contributor to the Zen architecture (and earlier K7/8/Athlon/ x64) has joined Intel after finishing work at Apple. Which is exceptionally good news for them since their main line CPU division seems to be having some significant problems.

He was last working for Tesla. Certainly a good catch from Intel.

[Zoraptor]

I knew it was Tesla and wrote Apple for some reason- apparently he did work at Apple for a time, just not recently.

 

And since I mentioned Intel's problems, their earning call for their financial results confirmed that bulk production for their 10nm process is now slated for some unspecified time in 2019 having been initially expected in 2017- and that would be 3 years after initial production started. The competition for Ryzen 3 may well be yet another 14nm skylake iteration.

[M4xw0lf]

Without their traditional process advantage, Intel will struggle a lot more in surpassing AMD's CPU offerings. Interesting times ahead, we are blessed that AMD survived long enough to get to this spot.

[MrMoe]

I recently upgraded my PC, tossed out the old i5-2400 ( i got nearly 7 years out of it, and its still working ok) and bought an 2600x.

 

I could have gone for a little bit more single processing power with an intel system, but I think with a more multi core performance and supporting the (current) underdog I made the better choice.

 

System runs really smooth, helped that I removed my last hdd too, so I should be good for the near future.

 

Ryzen2 2600x 

16GB DDR4

RX 480

250GB + 500GB + 60 GB SSD, Xonar DSX 

and a DVD drive I never use.

[Keyrock]

TSMC officially went into 7nm semiconductor mass production today. AMD will be using 7nm for both its upcoming Zen 2 cpus and Vega gpus and may use both TSMC and Global Foundries as suppliers.

 

It's interesting to see AMD surpass Intel in the dieshrink race Intel has traditionally always won, as Intel's 10nm process has fallen further and further behind and AMD will likely have 7nm processors by the time Intel has 10nm.

Edited June 22, 2018 by Keyrock

[Sarex]

TSMC officially went into 7nm semiconductor mass production today. AMD will be using 7nm for both its upcoming Zen 2 cpus and Vega gpus and may use both TSMC and Global Foundries as suppliers.

 

It's interesting to see AMD surpass Intel in the dieshrink race Intel has traditionally always won, as Intel's 10nm process has fallen further and further behind and AMD will likely have 7nm processors by the time Intel has 10nm.

 

As far as I heard what Intel is calling 10nm is what others are calling 7nm. Don't know what that is about though.

[Zoraptor]

It's about marketing really, smaller size being better. The '12nm' process used in Ryzen 2 is more akin to Intel's 14nm++ as well. 10nm is the more accurate name for all '7nm' processes, though the whole nanometer naming scheme is beginning to get a bit silly anyway at these scales. But assuming TSMC/ GloFo can get their 7nm process working in mass production they deserve the marketing edge of '7nm' given that Intel's 10nm is so error ridden that the only chip being sold after 3 years is a low clock dual core.

 

The prospect of Intel 14nm Skylake v4 or v5 taking on 7nm Zen- probably 6 core/ ccx so 12 core desktop to 48 core TR/ Epyc and not on a low power node so more overclockable to boot- is likely why Krzanich got the chop at Intel yesterday rather than him boffing an intern or whoever. He did pretty well at gouging maximising profit while AMD was not competitive but dreadfully at contingencies in case they did come out with a 'Zen' like product and some of their damage control has been absolute cringe and poorly judged- their 28 core answer to Threadripper 2 needing 1kg of VRMs and an industrial fish tank cooler to run briefly probably got more and more prolonged coverage than the demo of it did. The Intel board certainly seemed peeved since he didn't get the customary golden handshake on leaving.

[Keyrock]

^ Not to mention that Krzanich's efforts to get Intel a piece of the highly lucrative mobile market have failed spectacularly.

[Zoraptor]

I'd also suspect some people at Intel were extremely annoyed at him cashing out most of his shares the day before the spectre/ meltdown announcement as that really did look bad- and the rest of the board may not have had the chance to. That certainly looked like little letter insider trading even if it didn't make the definition of criminal Insider Trading. In the end though I'd suspect everything else- mobile woes, the affair, share trading and the various PR gaffes- is peripheral to the 10nm fiasco as that will potentially eat deeply into Intel's core (hoho) business.

[injurai]

I had no idea about the 10nm == 7nm thing. This seems very odd to me, shadier than redefining the meaning of a KB.

[M4xw0lf]

It's been years since the 'xx nm' really meant something. You'll find very few to no structural features on current chips which actually fall into the range of some tens of nm. Now they mostly just go to smaller numbers to indicate every kind progress in manufacturing.

[injurai]

That's rather disappointing to hear, but not unexpected. I just didn't expect for marketing terms to be so domineering that I've not heard about the reality of it until now.

[Keyrock]

The bottom line is that AMD has finally caught up to Intel in an area where Intel has traditionally always been a couple years ahead.

[injurai]

Given how Intel had stalled, I was under the impression they were sitting on a few golden eggs that they'd only reveal once AMD thought they had made up ground. Though it seems Intel has mostly been taking shortcuts for their advantage. I still haven't heard of anything particularly groundbreaking on Intel's horizon.

[Keyrock]

Given how Intel had stalled, I was under the impression they were sitting on a few golden eggs that they'd only reveal once AMD thought they had made up ground. Though it seems Intel has mostly been taking shortcuts for their advantage. I still haven't heard of anything particularly groundbreaking on Intel's horizon.

Lack of competition inspires complacency, this is what drives me nuts about fanboy behaviour (well, one of the things, anyway) when they revel in a competitor struggling, unable to comprehend that this will lead to lesser products from their own hero. 

 

AMDs disastrous Bulldozer architecture left Intel in a position where they had such a massive advantage over their competition in the desktop and server markets that they had no compelling reason to invest in and innovate in those markets.  Instead, Intel diverted most of their massive R&D force to going after the mobile market.  It was a solid strategic move, the mobile market is where the big money is, it was just poorly executed.  Intel stubbornly tried to force the Intel Atom architecture, an architecture poorly suited to functioning at the very low TDPs that smartphones operate at, into the mobile market, rather than engineering a new solution.  Atom simply couldn't compete with ARM chips and no amount of Intel smashing their heads against a brick wall could make it happen.  Intel managed to only get a tiny slice of the tablet market and basically none of the smartphone market.

 

Meanwhile, Intel's desktop and server chips stagnated.  Intel's design in those markets has been largely unchanged for the last decade and demands have shifted since then.  Intel design scales poorly into high core counts and desktop and especially server demands have become increasingly highly threaded.  I think Intel got caught with their pants down.  They never envisioned AMD's Zen architecture would be such a smashing success.  AMD's Zen managed to almost completely erase Intel's previously massive IPC advantage.  Furthermore, AMD's Infinity Fabric design used in Zen scales incredibly well, allowing AMD to slap more and more CPU and memory modules together with extremely small hits to latency.  This has left Intel in a bad position.  They are currently stuck with an architecture that's rapidly becoming, quite frankly, obsolete.  In the short term they have to use tricks and hacks to squeeze whatever life they can out of their current architecture, which has led to things like the utterly ridiculous 5 GHz 28-core presentation where they conveniently "forgot" to mention the insane hoops they had to jump through to get that demo to run.  Intel is forced to use these shameful tactics because they, quite frankly, don't have anything that can compete with AMD's upcoming 32-core Threadripper.

 

Intel has to develop a new architecture and do it soon.  Their current architecture simply does not scale well enough with increasing core counts and that is where the market is going.  Plus, Intel's IPC advantage is all but gone.  Until they bring a new, highly-scalable, architecture to market, Intel is going to fall further and further behind because they simply cannot scale as well as AMD's Zen.  No amount of tricks and hacks will make it happen.

[injurai]

I'm curious what Patents AMD currently has in play with Zen, because AMD might be able to gain a vicegrip on their advantages. Intel's future might lie in the courts at this point. I guess AMD also being a big player in the GPU market marks a secondary blow to Intel who has spun their wheels with their integrated graphics.