AMD zen pre-release thread!

[monstercameron]

So zen, the last chance for amd in the performance market.

1st burst of news
Source: http://www.phoronix.com/scan.php?page=news_item&px=AMD-Zen-CPU-Znver1

Last week AMD published a "add znver1 processor" patch to the GNU Binutils package. This patch doesn't mention Zen explicitly but "znver1" is short for Zen after their long run of bdverX codenames within Binutils, GCC, etc, for indicating the Bulldozer architecture.

This patch reveals the AMD Zen design no longer supports TBM, FMA4, XOP, or LWP ISAs. Meanwhile the new ISA additions are for SMAP, RDSEED, SHA, XSAVEC, XSAVES, CLFLUSHOPT, and ADCX: ISAs are supported. It's nice to see with Zen that AMD will support the RDSEED instruction, which Intel has added since Broadwell for seeding another pseudorandom number generator. SMAP is short for the Supervisor Mode Access Prevention and is another Intel instruction set extension already supported by Linux.

P.S keep it civil and drop the doom and gloom.

 

[Silverforce11]

I'm not going to get hyped for whatever CPU is coming from AMD because they have shown zero capability to meet any form of hype. Under-delivery is their track record.

 

[monstercameron]

I'm not going to get hyped for whatever CPU is coming from AMD because they have shown zero capability to meet any form of hype. Under-delivery is their track record.

maybe Lisa Su straightens out the crappy marketing and overdelivers...who knows, the future is in flux

 

[Morgoth780]

I'm really looking forward to launch. From a performance standpoint, I'll consider it a success if it has better st performance than sand bridge and has more threads than has well at a similar price point.
I don't know how well that would sell though.

 

[mrmt]

maybe Lisa Su straightens out the crappy marketing and overdelivers...who knows, the future is in flux

Zen is Rory's son. Lisa's sons will arrive circa 2018.

 

[frozentundra123456]

Any information about Zen is interesting, but a "pre-release" thread seems a bit early, since it will be more than a year at best until release, and could be nearly 2 if there are any delays.

 

[Dave2150]

I'm really looking forward to launch. From a performance standpoint, I'll consider it a success if it has better st performance than sand bridge and has more threads than has well at a similar price point.
I don't know how well that would sell though.

You believe that Zen will be a success when it launches in 2016/2017, if it has better performance than Intel Sandy Bridge from year 2011?

 

[VirtualLarry]

You believe that Zen will be a success when it launches in 2016/2017, if it has better performance than Intel Sandy Bridge from year 2011?

Well, for AMD, that would be "success". As opposed to "unmitigated failure" (Bulldozer).

Don't discount the success of "good enough computing". I think that I read here that nearly all of the top laptop sellers on Amazon were Bay Trail-M based.

So don't think that AMD is competing with Sandy Bridge, so much as Bay Trail (and by then, Cedar Trail).

 

[CHADBOGA]

who knows, the future is in flux

I thought the future was fusion. D:

 

[frozentundra123456]

I think that I read here that nearly all of the top laptop sellers on Amazon were Bay Trail-M based.

That is really depressing. Wonder if Intel is intending to shoot themselves in the foot or if it just turned out that way. Seems like they are compromising both their profits and their brand image. I cant imagine anyone buying an atom or kabini laptop and really *liking* it. Tolerating it maybe, but there are so much better options available.

 

[Ramses]

Must be awfully hard to design a CPU for a market that's two or three years in the future.
I could see like a year, but two or three is ages.

 

[Shehriazad]

Eeeeeh.

This can actually go either way.

AMDs designs before Bulldozer family actually had some really powerful designs. If they can get something similar to that and then produce it in at least 16nmFinFET...then I don't see how it won't be able to compete.

I still think that Intels nm size "lead" is being overestimated. It's simply that Intels cpu DESIGNS were a lot stronger.


Let's say AMD manages to push out that Zen in 2016 in 16nmFF as a solid 95W 8core/16 thread design with singlethreaded performance of Intels chips a la 4XXXK series.

In the meantime Intel arrives at like what...10ish nm? I'm not quite up to date on their die size roadmap....either way...once you go that small...you hit quite a few technical difficulties.

I mean at least for gaming...Intels new 5690X top dog isn't actually a step forward...so going smaller doesn't necessarily mean more gaming power. Sure...those chips do certain workloads extreeeemely welll...but...ehm...the price is always an issue.

Going smaller for the sake of going smaller might not always be the best step.

And I don't think that going overly pessimistic mode on that future AMD chip is all that reasonable.

People keep saying that FX sucks...I mean sure...NOW it does. But when their 8 cores were new...Intel had no 8 threaded alternative that was actually remotely equal in affordability.


For the normal consumer, 95-100 Watt is for some reason also seen as some kind of evil wattage value that magically turns up the heat in your PC to 1000°C...when in fact some cheap 25$ heatsink is already enough to give it headroom for overclocking.



I'm not saying Zen will be a product to blow us away with its high end performance...I'm saying that Zen will correct the Bulldozer family flaws...while coming with the affordability that buyers love so much about their products.

Sure...I can accept 20% less gaming performance on the top chip...if Intels high end alternative costs 5x as much.

Not to mention by the time Zen is out...DX12 will be a thing that has hit the mainstream gaming due to it's adoption rate that is guaranteed to be far above that of DX11 back in the day.

And once that happens...raw performance will further go more into the background at least for gamers...putting the price into the foreground.


As for workloads...again...buying a 200$ product that performs 20% weaker than the 1000$ alternative is always an option.



And as always...take my speculative posting with a grain of salt...I can't see into the future...but this is just my guess at how this could possibly end up. For all I care....this could also horribly fail, causing AMD to retreat from the CPU market entirely, giving Intel a 100% monopoly until the end of all days.

 

[VirtualLarry]

I would happily pay AMD $200 for an 8C/16T CPU with Sandy Bridge-equivalent IPC or better. Maybe even $250.

That means a lot coming from me, I normally don't pay more than $100-130 for CPUs.

I think a 4C/8T SB-equivalent (2600K/2700K) would probably sell well @ $100-130.

Edit: I could care less about the IGP, if I'm spending that much on the CPU, then I'm going to invest in a matching discrete card too.

 

[Arachnotronic]

Must be awfully hard to design a CPU for a market that's two or three years in the future.
I could see like a year, but two or three is ages.

That's pretty much the design cycle for a CPU. Actually, it's more like 4 years.

Yeah, getting it right that far in advance is tricky business!

 

[Ramses]

I would happily pay AMD $200 for an 8C/16T CPU with Sandy Bridge-equivalent IPC or better. Maybe even $250.

That means a lot coming from me, I normally don't pay more than $100-130 for CPUs.

I think a 4C/8T SB-equivalent (2600K/2700K) would probably sell well @ $100-130.

Edit: I could care less about the IGP, if I'm spending that much on the CPU, then I'm going to invest in a matching discrete card too.

Yeah, bout the same. As happy as I was with an FX, saying that as a current 4790k owner even, it wouldn't take a miracle to get me to buy. Just a good improvement for a good price. And a modern motherboard to stick it on.

 

[Puffnstuff]

With a name like zen is it being built by a monastery?

 

[schmuckley]

After reading that,all I can think is: "Integer divide by zero"

 

[ThatBuzzkiller]

No transactional memory or AVX2 ?!

Ok this is definitely off my list then ...

Quite sad as this will unlikely be matching Haswell in single threaded performance ...

 

[Arachnotronic]

Does Excavator have AVX2?

 

[Roland00Address]

Well I hope it turns out good, but I have given up on hoping for AMD till I see real products.

AMD started pushing the fusion marketing buzz in 2008, and we did not have our first fusion cpu till 2011 with llano. We only got HSA 1.0 this week (2015) and what software uses it right now or in the near future? I thought the whole point of fusion was using the gpu for compute, not to have a better igp and better battery life, if that is your goal than intel beat amd to fusion a full year prior with a better performing part.

So 7 years and no results? Why should I get excited for a part that does not come out for another 12 months made by a company with a shrinking development budget, and fabs that are now 2 generations behind?

I hope it turns out good, but hope is cheap.

 

[pTmdfx]

Does Excavator have AVX2?

Yes.

 

[csbin]

"bdver4" include BMI, BMI2, TBM, F16C, FMA, AVX, AVX2, XOP, LWP, AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM

http://www.phoronix.com/scan.php?page=news_item&px=MTUxMzE

 

[itsmydamnation]

No transactional memory or AVX2 ?!

Ok this is definitely off my list then ...

Quite sad as this will unlikely be matching Haswell in single threaded performance ...

the dropping of XOT will be because of AVX2, because you know they are basically the same thing. Why care about transactional memory in a consumer CPU, its going to be forever before its used. Skyrim uses x87 and your worried about transactional memory.

The one thing we dont know is what has AMD R&D mix been, have piledriver/steamroller/excavator been done on a shoe string to give as much resources as possible to Zen or has it been more even distribution.

I personally think given most of the changes in steamroller/excavator are uarch independent, Adaptive Clocking, high density lib, AVFS and that they have stayed at 28nm (20nm would have been fine for Carrizo target) that they only spent resources developing things that could be used on the new uarch.


edit: even looking back at things like the Fix for the icache thrashing, they could have redesigned the cache increasing it ways or just a dirty make it bigger, they just made it bigger.

 

[PG]

"bdver4" include BMI, BMI2, TBM, F16C, FMA, AVX, AVX2, XOP, LWP, AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM

http://www.phoronix.com/scan.php?page=news_item&px=MTUxMzE

Why post this? It's from late 2013. The OP's link is from earlier this year and much more relevant.

 

[PG]

I'm not going to get hyped for whatever CPU is coming from AMD because they have shown zero capability to meet any form of hype. Under-delivery is their track record.

Wow, first post of the thread and you had to be negative, even after the OP's opening statement: "P.S keep it civil and drop the doom and gloom."

So much for a constructive discussion.