AMD Releases More Architecture Details on ZEN

clocks dont really matter its the STP that does. Majority of programs arnt multi core thread aware of 4 so 4 core, 6 core 8, core 24core, it dont mater how they performace if programs cant use the cores in meaningfull way, More core are nice but not at the sacrifice of STP which what is king still these days

Untill "ALL" programs become multicore aware STP will be remain king. which is gona be while even with the claims of DX12 and dev are lazy about that.

 

I want to see some APUS with a higher tdp cap for overclocking 4c 8t with rx460 equivalent that would almost be enuff to replace my current rig lol

 

Single thread performance is directly affected by clock frequency....as is the performance in general. It doesn't matter if the processor can do 9IPC or 9000IPC, if the clock frequency is too low for the processor's performance to be competitive.

 

.... like I said I wana know how much STP the gona sacrificed for all those cores. which directly related to clock speeds and architecture it not just about clockspeeds.

 

that leaked ES sample was at base 2.8Ghz and turbo 3.2Ghz


Imo 8core base 3 or 3.2Ghz seems very plausible, Intel 6900K 8core has base 3.2Ghz

 

None. There is no useless integrated GPU in the high-end desktop parts, unlike the 4-core i7s, whose largest part of the die is the GPU.



This part is equivalent to the socket 2011 parts from Intel, and the AM4 socket will have a similar amount of pins, if I understood correctly.

By the way, the AMD presentation was very low key and professional, I liked very much how measured everyone was. After reading the Anandtech architecture analysis, this is shaping up to be a great part. We'll get more in the coming months.

 

I said the most important piece of info we are still missing. Don't be like a political journalist by only looking at a single part of a sentence, cutting it out of context.

Of course price is also important, but it's not a technological specification.

 

Judging by the presentations, I expect 3.0GHz to be the absolute minimum.

 

A iGPU isn't useless, my 4770k's keeps up with my 780Ti with SETI work units. It's also nice having a backup to the nvidia card that gets a black screen and requires a restart now and then due to driver bugs.
AMD better deliver, Intel needs competition. Perhaps they shall win against Intel? They failed with nVidia miserably and nVidia gimped their card designs as there was no real competition expected. Just look at those small memory buses, they are obviously holding back there till the next series.
I wish there was competition, just look at what's being held back!

 

yeah, near 3ghz is very plausible indeed and would have a lot of overclocking headroom just like intels 8 core too.

 

are you kidding? amd has not failed against nvidia at all, in fact they are doing incredibly well against them. Nvidia basically has more money to spare and they're ready to even sell at no profit if it means that they can keep the performance crown. AMD is the only reason nvidia consistently releases giant chips like we had in every titan or 780/980ti.

 

Yeah I never understood this notion either. Sure, AMD doesn't have a Titan equivalent, nor do they need to. Just because they don't have the world's fastest GPU, that doesn't mean they're far behind in terms of the average person's budget. AMDs draw more power. Not a lot, but apparently enough that people think it can be used to cause city-wide brownouts. Seems to me people always linger on a company's previous weaknesses, regardless of whether the problems still persist.


@Minotaur
I find it really hard to believe that a 4770K IGP can compete against a GTX 780Ti. SETI, to my knowledge, is optimized for CUDA. And even if that were the case, it all depends on the workload. Supposing what you said is true, maybe SETI just simply isn't that heavily parallel, so a simple IGP is all you really need to make it work.

 

iGPU might be nice under abnormal circumstances (main video card busted), but I'd rather have a better CPU for 99.99% of time than a weaker CPU + iGPU for the sake of 0.01% of time. You can always get some 50 bucks basic videocard to keep in a drawer in case of emergencies if your super CPU doesn't come with an iGPU. No big deal when you know it beforehand.

 

I cannot wait for the real deal to get into Hilbert's hands to see some good, honest numbers to see how these will really perform in the wild. I want them to kick some Intel tail like the old days, but I suspect they will be very similar to the current i7 lineup. Priced right, that should still be great for everyone.

AMD, as a company has been flailing. They don't have the R&D money like Intel because they are not in very good shape financially (If you compare stocks/reports/marketshare). I've stopped hoping for AMD to bring enthusiasts the next big leap in CPUs. On the GPU side, they face nearly the same, lopsided uphill battle with Nvidia.

A disadvantaged AMD shouldn't be able to win these (King of the Hill) battles but who doesn't root for the underdog anyways?

 

Spending money on research and development in principle should mean you end up with better products as you are alluding, however this also depends on the effectiveness of that expenditure. Intel have recently announced their third great R&D failure since Haswell, they are just great at covering this with positive spin.

The first failure Broadwell, which resulted in the release of the Haswell Refresh (i5-4690K and i7-4790K etc). By the time they got Broadwell to a satisfactory point Skylake was already released.

The second and third failures are related as both pertain to Cannonlake. The second failure with Cannonlake not working as expected has resulted in the release of Kaby Lake. Kably Lake is basically a Skylake refresh, but with some additional refinemets in the CPU and chipset. Intel extended the R&D for Cannonlake hoping it wouldn't be delayed too long. This didn't work for them, and effectively has resulted in Cannonlake being dropped from mainstream consideration. This resulted in the news of Coffee Lake (second Skylake refresh), with an added sweetener to hide this fact in the form of a six core mainstream i7 variant. Cannonlake with all its expenditure will just be used in low powered devices. R&D expenditure would now be focused on Ice Lake.

Now, you also have to consider the architecture changes:
Sandy Bridge --> Ivy Bridge --> Haswell --> Haswell Refresh --> Broadwell --> Skylake --> #Cannonlake --> Kaby Lake --> Coffee Lake --> Ice Lake.

Yes, Cannonlake is before Kaby Lake, because originally it was meant to be released at that point. I've colour coded the architectures. Sandy and Ivy were always planned. Haswell Refresh was an added CPU so is in bold of the same colour as Hsawell. The same goes with Kaby Lake and Coffee Lake, same colour as Skylake as they are closely related. Cannonlake and Ice Lake are the same colour as it is likely Ice Lake is an evolution of Cannonlake rather than a complete new architecture.

So, in the progress above there are four major architecture releases with all that R&D, however two of them are 'failures'. To top this off, yes Skylake is faster than Ivy Bridge, but probably not representative of the amount of R&D spent between when Ivy Bridge and Skylake were released.

 

That doesn't make any sense, what you just told me is that a Pentium 4 @ 3.2Ghz beats out a i7 @ 3.0Ghz, because default clocks mean anything right? Wrong. Speculating that a processor is good or bad based off of the "default clock" complete and total BS.

Lets just give an example:

Are you telling me, you'll get an Intel processor @ 4.0Ghz over an AMD processor at @3.0Ghz because the AMD processor can't get to 4.0Ghz, even though the AMD processor @ 3.0Ghz beats the Intel processor @ 4.0Ghz, all because you want a higher number? How does that make sense?

"Default clocks are only meaningless if you're planning to buy the product regardless of it's performance." <---- Except default clocks mean nothing to performance, i'm not sure how people still don't understand this

By saying what you're saying, you are 100% stating my example is true. This makes no sense.

I feel like everyone still has the mentality of the Pentium 4, so many people bought it, and insisted it was better then AMDs processors at the time, because the Mhz was so much higher, even though AMDs actual performance was better.

I'm not saying AMD will beat Intel, per mhz for that matter even, or for how high the frequency will go, but really? this mentality again even after the pentium 4? have we learned nothing?

No, it's not.

A single core processor at 1.2Ghz that performs a task faster then a single core processor at 2.0Ghz does not mean that "clock frequency matters". If clock frequency mattered, the 2.0Ghz would be faster. This has happened time and time again, how is this not understood? It's all about the architecture and total performance, and nothing about frequency.

Except that it's not important at all, so..how is it the most important info we are missing? It's information that would be 100% worthless without knowing how the processor performs as a whole. They could say this processor starts out at 10Ghz, and yet that information would mean absolutely nothing, and if it didn't perform better then intels processors, it wouldn't matter that it's 10Ghz other then to people who simply want more Mhz for no reason.

 

Though what you said is true, I think what he meant to say is a 1.2GHz single core will be slower than a 2.0GHz of the same architecture. It's obvious to the point that I'm not sure why he brought it up at all though. But the simple fact is Zen will most likely operate at lower frequencies due to its higher IPC.

There is one important factor to consider. AMD implies that Zen will have an IPC 40% higher than Excavator. For argument's sake, let's say AMD released an 8-threaded Excavator AM3+ CPU at 3.5 GHz. In order for an 8-threaded Zen to compete with that CPU, it theoretically needs a minimum frequency of 2.1 GHz. Though I highly doubt AMD will release something that slow, asking what frequency the CPUs will be is a little relevant, because we can have a clearer estimation of how it will perform compared to Intel.

 

yah I think igpu is waste of space on high end cpu be it intel or amd or what ever i dont know why they are in there other then intel dont seen the need to use the space to actual improve there chips, but it "can" come in handy when and if a dGPU dies and need to be replace/rma when you dont have spare lying around.

Like I said I want AMD to pull the magicly rabbit out of the hat that make intel sweat, cause then the whole who is faster and cheaper wars can begin anew and the end user profits

 

Go zen, I really have hope for them.

 

You're so hung up on "speed" instead of the efficiency. Those architectures you listed as failures are anything but. If you look at laptops, they've been getting faster and faster year after year while the battery life and feature list has been increasing as well. This is where the focus is.

Sure, "gamers" may be disappointed because the market doesn't go along with their wishes, but that's life.