*** AMD "Zen" thread (inc AM4/APU discussion) ***

[CAT-THE-FIFTH]

yes, so its like 60%+ over Piledriver.

Wow.

 

[cereal3]

yes, so its like 60%+ over Piledriver.

shockingly **** piledriver chips or zen engineering marvel you can be the judge

 

[Troezar]

If Ryzen does as well as it is looking I'll definitely be switching back. Just need some good motherboards to go with them.

 

[billybob89]

It's interesting how they haven't said anything about the turbo. Either it's very clever marketing holding the information back till its realised so they are not giving much info away for Intel. Or maybe they are having problems with it low speed gain etc.

Hope it's the latter.

Also read somewhere (can't find the link) that they may be poor overclocks. But does that matter if they appear to be performing so well?

 

[Harlequin]

how the living **** does anyone know how they overclock?? more clickbait fake news!

 

[StarShock]

It's interesting how they haven't said anything about the turbo. Either it's very clever marketing holding the information back till its realised so they are not giving much info away for Intel. Or maybe they are having problems with it low speed gain etc.

Hope it's the latter.

Also read somewhere (can't find the link) that they may be poor overclocks. But does that matter if they appear to be performing so well?

do you mean you hope its the former?

 

[8 Pack]

Who even cares about stock. So many on about turbo. Come on it depends on average cpu top clock. Then compare performance using these frequencys.

 

[h4rm0ny]

shockingly **** piledriver chips or zen engineering marvel you can be the judge

Glass half-empty or glass half-full? You can be the judge.

do you mean you hope its the former?

I doubt it. Some people the hate is all that keeps them going.

 

[Boomstick777]

Who even cares about stock. So many on about turbo. Come on it depends on average cpu top clock. Then compare performance using these frequencys.

Yeah man I wanna see Intel and AMD's best running at max clocks and compare those. Need to wait for something official.

 

[Kmodo]

Don't have as much disposable income these days. Think I'll get Vega on day 1 and see how much my 2600k bottlenecks it, then save for Zen.

 

[~>Dg<~]

yes, so its like 60%+ over Piledriver.

hahaha my god i knew it . by the time its released it will be 110 percent power the sun and create a cure for cancer.

from what i seen its closer to 40 percent.now work that out from the chips they talking about which have been slower than a i5.then you end up with something around a 6700k in normal performance. maybe quicker in multithredded stuff.

 

[humbug]

hahaha my god i knew it . by the time its released it will be 110 percent power the sun and create a cure for cancer.

from what i seen its closer to 40 percent.now work that out from the chips they talking about which have been slower than a i5.then you end up with something around a 6700k in normal performance. maybe quicker in multithredded stuff.

Your off the rails... We know a lot about its performance, we know clock for clock core for core its as fast as Sky Lake or Broadwell, yes it will be faster than a 6700K in multi-threaded in the same way a 6900K is faster than a 6700K in multi-threaded.

What we are trying to figure out here is how much IPC up lift the performance we know about is from Piledriver, it works out at about 60%+

 

[FoxEye]

Don't have as much disposable income these days. Think I'll get Vega on day 1 and see how much my 2600k bottlenecks it, then save for Zen.

Going to be an expensive start to the year, for sure, if both Vega and Zen deliver. (Shame I have to put that qualifier on the end, but my head tells me not to count AMD's chickens before they've hatched, laid a few eggs, and survived their first winter...)

Wish I hadn't bought a 480 now; someone on the MM is going to get a bargain

 

[CAT-THE-FIFTH]

Your off the rails... We know a lot about its performance, we know clock for clock core for core its as fast as Sky Lake or Broadwell, yes it will be faster than a 6700K in multi-threaded in the same way a 6900K is faster than a 6700K in multi-threaded.

What we are trying to figure out here is how much IPC up lift the performance we know about is from Piledriver, it works out at about 60%+

He has lost the plot,its a Ryzen 8C/16T sample running at 3.1GHZ to 3.3GHZ and the 8C/16T Core i7 6900K is running 10% faster whilst running at 3.5GHZ to 3.7GHZ or thereabouts.

According to AT forums it is a motherboard validation sample and does not have Turbo properly working - it also lacks that new Turbo implementation AMD mentioned at their preview.

Also,if you look at the charts,the A12 9800 and FX8370 are more or less the same for games,indicating the overall games suite is not really taxing more than four cores and values single core performance.

Then if you look at some of those games,like ARMA III,they do horribly on AMD CPUs.

The 3.1GHZ to 3.3GHZ Ryzen sample is 32% faster than a FX8370 running at 4.0GHZ to 4.3GHZ or thereabouts in a games suite where 4 of the 6 games use two to four cores and do not scale with SMT.

That alone would hint IPC in games has gone up well over 40% and apparently at one of the interviews with an AMD engineer,it was meant to be 48% over Excavator.

The magazine who leaked the results is CanardPC and the chap who is the editor leaked the first Athlon 64 results months before release,and I can still remember all the denial from some people at the time when reading tech forums.

Either way it is far above the SB or IB level IPC I expected,so looks like AMD will be having something decent - after all even Intel said Ryzen would be "disruptive" until they launched their 10NM CPU range.

 

[alex_123_fra]

I hope AMD are already working on 10nm or 7nm processes as Ryzen may be quite short-lived with Intel planning Cannonlake for 2H 2017.

 

[CAT-THE-FIFTH]

I hope AMD are already working on 10nm or 7nm processes as Ryzen may be quite short-lived with Intel planning Cannonlake for 2H 2017.

BTW,you might want to read this and it was published in a print magazine:
https://www.reddit.com/r/AMD_Stock/comments/5k31sr/i_bought_canard_pc_there_you_go/

They have internal sources at Intel and the climate there is currently awful apparently, employees are very discouraged and some feel like "they have nothing left to lose". They are making big profits now by totally screwing up the future. According to the magazine, Intel is "probably in the most delicate situation it has had to face to this day". The CEO is reducing cost so much that he's managing engineers like "supermarket cashiers", he doesn't care about taking the time to train them.

Krzanich is very impatient and eager and keeps changing his mind about projects. If a new architecture isn't created in like a couple weeks, he gives up and cancels the project... he keeps sending contradictory instructions to the teams.

"Fab Hell": Intel is likely going to have a 6 month delay on 10nm. Worse, even Cannon Lake is not expected to feature any significant architectural improvement. Basically Intel was just hoping AMD would keep not competing with them.

Krzanich is apparently a disaster, and he won't be able to stay CEO for long. Apparently some people have heard him yelling from the next building when he was angry. But he seems unaware of him being perceived so negatively. Employees at Intel hope Murthy Renduchintala will replace him ASAP, and he seems much more capable and is slowly refocusing Intel in the right path, but basically R&D is ****ed atm and there will be a huge empty space until about 2019. Apparently, Krzanich completely underestimated the possibility of an AMD comeback.

(I didn't really understand that point I'm not expert enough) but apparently x86 is going to disappear sooner than expected, it'll be replaced by ARM and Intel is panicking about that.

Intel is currently working on a "multichip package" (MCM) integrating an Intel CPU and an AMD GPU. So this is confirmed guys.

In the picture before the last picture I posted, it is said that Kaby Lake has exactly the same IPC as Skylake. No improvement as to perf/watt but there seems to be more room for overclocking.

Personally I think AMD has a good chance currently we know they are working on Zen+ which probably their APUs will have.

But if it is true,Intel is going to integrate an AMD GPU with an Intel CPU,that would be a bigger deal than even Ryzen!!

Apparently from what I have read earlier,Intel has trimmed people from its GPU design teams - if that is actually true,it might explain why they might do this.

Plus,the whole Intel CPU and AMD GPU bundle thing was rather intriguing.

The thing is Intel even on 14NM can produce more volume than AMD - they will sell more than AMD can sell at full pelt.

However,if AMD were to even catch even 10% of the server market and increase consumer CPU sales at the same time,for a small company like AMD that would be a massive deal for them.

For AMD to even get close to a Core i7 6900K would massively uplift their perceived value as a brand - it will be the first time since the days of the Athlon 64 AMD has managed to even compete with a £1000 Intel CPU.

 

[Angilion]

shockingly **** piledriver chips or zen engineering marvel you can be the judge

I'd go for neither. The people making the decisions at AMD chose to accept lower IPC for a higher degree of parallelism. They misjudged what would be the most common uses, so they ended up with a CPU that's quite good at uses that turned out to be not very common. Not shockingly bad engineering. Just the wrong focus on the design. With Zen, they changed the focus to be more towards IPC. Not an engineering marvel. Just the right focus on the design.

 

[nashathedog]

 

[AllBodies]

I hope AMD are already working on 10nm or 7nm processes as Ryzen may be quite short-lived with Intel planning Cannonlake for 2H 2017.

Yeah they've said they're working on GloFo's 7nm next, and planning for 2018-2019 release.

They haven't said anything about consumer chips yet (I think), but they have said there'll be a 32c/64t server processor called Naples using Zen+ architecture.

And Zen+ is meant to be more of an Intel style revision, so probably 5-10% more IPC than Zen, and a couple of extra hardware features. But also 7nm looks to be shaping up to be a very strong node, and it wouldn't surprise me if we get 12 core CPUs capable of 4+ GHz on all cores.

 

[Harlequin]

i can see that happening - GloFO are going for a full node shrink and not a hybris ( Samsung 14nm is a hybrid = 14/20nm hybrid 20nm is Samsung unreleased process for back end , but 14nm for front)

so GloFO 7nm will be FinFET , tsmc are going 10/20 hybrid as a stepping stone to 7nm , intel are likely going 7nm as well (10nm is delayed) and sadly UMC are out of the game now