AMD "Zen" CPU Prototypes Tested, "Meet all Expectations"

[SviatA]

Huh, I would believe if some other engineers test any Zen unit and told that.
Surely, AMD will not lie, but marketers can tell anything to show that the new CPU/GPU/RAM is much better, while it is still on par or a little bit better.
Anyway, thumbs up for AMD. I'm still using an old-fashioned AMD FX-4300. But I had a choice, I would take a Phenom II rather than a Bulldozer or Piledriver.
PS: I saw that some retailers still have very old Phenom units, like PC24.de - there was an AMD Phenom II X2 3.3GHz. If that was a 4-core unit, I would take it...

 

[Athlonite]

I much prefer the wait and see method as rumours and speculation are meaningless drivel which is unsubstantiated without real proof

 

[RealNeil]

I much prefer the wait and see method as rumours and speculation are meaningless drivel which is unsubstantiated without real proof

Especially since we have to wait and see anyway,...........

 

[Schmuckley]

I won't even buy a motherboard until I see hard proof.
I got stung by Bulldozer;not happening again.
Used to use AMD stuff all the time.

 

[medi01]

Well that depends on what specifically the expectations were doesnt it?

Main promise was +40% IPC.
Perf/watt should, erm, double, but this part I remember vaguely.

...AMD publicity machine its been proven not to be Trusted

Care to name "publicity machine" that has been proven "to be Trusted", please?

The problem is we have software that needs FPU.

"fury X" also met their "expectations".

Remember, tthat 980Ti was released to spoil the Fury launch.

That's my question as well. If Zen does well it is great, but what AMD needs to do is follow up, because Intel don't stand still so AMD still need to match Intel's cadence. In the past, AMD have put out good product, but its increments after the first iteration have lasted more product cycles than they ought to

I hope that, since:
1) "Moore's law" is dead, fab development has slowed down a lot and if we see huge improvements, they are unlikely to come from silicon shrinks
2) AMD doesn't need to cover fab R&D costs
3) IPC optimizations are close to diminishing returns

AMD still has a chance.

The only thing that bugs me, is that I cannot find AMD chip, that has better IPC than my ancient 45nm i5 750.


PS
combined posts into one.

 

[RCoon]

Main promise was +40% IPC.
Perf/watt should, erm, double, but this part I remember vaguely.

Care to name "publicity machine" that has been proven "to be Trusted", please?

Remember, tthat 980Ti was released to spoil the Fury launch.



I hope that, since:
1) "Moore's law" is dead, fab development has slowed down a lot and if we see huge improvements, they are unlikely to come from silicon shrinks
2) AMD doesn't need to cover fab R&D costs
3) IPC optimizations are close to diminishing returns

AMD still has a chance.

The only thing that bugs me, is that I cannot find AMD chip, that has better IPC than my ancient 45nm i5 750.

Please do not double or triple post. There is a multi-quote option for replying to multiple people, and there is also an edit button to add additional comments to your post.

 

[HumanSmoke]

http://www.supermicro.com/products/motherboard/Xeon/D/X10SDV-8C-TLN4F.cfm
http://ark.intel.com/products/87039/Intel-Xeon-Processor-D-1540-12M-Cache-2_00-GHz
http://www.servethehome.com/intel-xeon-d-1540-performance-comparison/

If Intel can do a 8c/16t Xeon SoC in a 45-watt TDP envelope, I think Intel can do a lot more than what they're offering the run of the mill consumer.

Well, at least HEDT is continuing its incremental core addition. Seems Broadwell-E gets an i7-6950X with 10C/20T. Entry level for the new series seems to be 6 core/12 thread.

 

[Aquinus]

Well, at least HEDT is continuing its incremental core addition. Seems Broadwell-E gets an i7-6950X with 10C/20T. Entry level for the new series seems to be 6 core/12 thread.

Well, when skt2011-3 came out, the entry level Xeon was a 6c/6t chip for 200 USD, still is I think. Since then they've released a 4c/8t Xeon that has some power behind them (for E5 Xeons,) but, they're not the cheapest in the Xeon lineup, however they cost less than a 5820k.

Personally, I'm bummed that there isn't an quad-core i7 part for 2011-3 however, you can get just about every quad core Xeon option on 2011-3 for less than a 5820k, so for the 3820 in Xeon form, on skt2011-3 could get you all of your PCI-E lanes, 2 less cores, and ECC support, without a hit to cost. For anyone who was planning on using a lot of PCI-E and wasn't planning on overclocking, I would point them to something like the Xeon E5 1630 V3. It's the quad core part that should have had an i7 counter part. Problem is that it would probably cannibalize the 5820k because most people have no use for 6c/12t but, gamers and workstations alike would probably prefer to have their 40 PCI-E lanes for multi-GPU, SSDs, RAID, whatever. I know I would.

Either way, this is all beside the point. AMD has to make some serious progress on single-threaded performance to get Intel change their ways. Until something substantial happens, they're going to keep milking the technology they've had for years. All the while, it has given Intel valuable time to improve their iGPUs which they have been doing... significantly (consider the leaps in performance every generation.)