Is this why AMD has been aggressively promoting Zen?

[Azuma Hazuki]

I like how OP rather explicitly accused AMD of market manipulation--which is fraud--and still hasn't been tagged with a trolling/locked thread, which tends to be a very speedy action in the CPU/GPU forums here, due to the tireless efforts of the overworked mods fielding the endless stream of complaints that these forums seem to generate.

It makes me wonder where the majority of the reports for assumed examples of "trolling and character attacks" come from, if the example here hasn't generated the typical speedy mod reaction that would certainly happen with this thread had it been directed at OP's least-preferred silicon-manufacturing team.

Or, it's legitimately not considered trolling for reasons. I would think that non-vague accusations of fraud require actual evidence, though.

(this is not meant to be a mod callout, btw. It's a members-who-endlessly-report comments callout.) It strikes me as rather cynical that certain members are easily and quickly offended when others accuse one team of unfair, non-competitive, exploitative pricing schemes, and consider such comments as "trollish," where as comments here that suggest actual fraud are posted by and supported by (hello, tell-tale likes) those very same members that are the quickest to call for trolling in the other direction.

Caaaaaareful, the mods have a dozen kittens and the occasional baby shark when one member calls another one out Use the report feature on Arach's original post...like I'm going to do right now.

 

[kawi6rr]

Anything OP says relating to AMD I take with a grain of salt and a good laugh, he's a known AMD hater. I bought AMD stock when it was down around 1.6 and I'm laughing all the way to the bank.

Please stop the trolling.

 

[swilli89]

Yeah gotta be careful calling people out, however I do have my doubts about OP's genuineness when his signature was literally him making fun of my $10 AMD share price by March 2017 comment for a while.

 

[zinfamous]

Caaaaaareful, the mods have a dozen kittens and the occasional baby shark when one member calls another one out Use the report feature on Arach's original post...like I'm going to do right now.

I don't report posts for various reasons, but particularly in these two technical forums, these mods are heavily overworked, and, according to the mod announcements, due to a very select number of individuals that aggressively report everything. I simply don't want to add to that work load and while I do try to walk along the positive side of what is acceptable in these forums, I do find it difficult to not call a spade a spade, even if doing so violates the specific forum decorum. :\

I am 100% certain that my comment got reported by those very same members....which is the point. If it is determined by the mods to be a mod callout on top of a member callout, then so be it. I am perfectly fine with their determination along those lines. But it is very bizarre that a blatant case of a troll thread has survived so long in this forum (it has survived because the members that frequently report such cases actually support this troll; hence the lack of reporting)

I know that OP makes no secret of his preferences as many do around here, on either side, but I tend to think of him as one of the few that brings data and reasoning to the table when he makes comments. This thread however is pure trollery, which is what I find disappointing about it.

 

[Arachnotronic]

Caaaaaareful, the mods have a dozen kittens and the occasional baby shark when one member calls another one out Use the report feature on Arach's original post...like I'm going to do right now.

I actually self-reported this thread asking the mods to move it to the appropriate sub forum if this isn't the right place for this.

EDIT: I have edited the title in response to feedback.

 

[Azuma Hazuki]

There is NO RIGHT PLACE FOR TROLLING, Arach. Cut it out.

 

[Arachnotronic]

There is NO RIGHT PLACE FOR TROLLING, Arach. Cut it out.

This thread isn't an attempt to troll. What possible benefit could there be for me to "troll" on an internet forum? As a regular poster on these boards, and as somebody who enjoys the discussions here, there is zero benefit to trolling.

In this sub-forum we regularly discuss the business/financial side of these companies, and I thought this was relevant. I also provided my view of what I think is happening.

I even updated the title of the thread based on feedback from people in this thread.

 

[crashtech]

Companies aggressively promote their upcoming products, it's what they do. Perhaps a better and less inflammatory take would be that the timing of the offering is pretty good, since the buzz is mostly positive right now.

 

[Dresdenboy]

I have to agree that it was a smart financial move for them.

Doesn't this squarely contradict the still rather present opinion of AMD's incapability to do anything smart?

 

[Fanatical Meat]

This is no surprise. AMD are masters of shady accounting. With the disaster that is Polaris and the looming disaster with Zen only being on par with pre-sandy bridge IPC, the only way they can improve their financial situation currently is with hype-based marketing and dilution of shares.

I'm far from an AMD fan but given your forum name your post feels trollish

 

[krumme]

I think its pretty difficult to get a good discussion out of this crazy op stuff. Its difficult to know if its serious and to what degree.

Granted the 250% uplift is fun but generally this hype discussion is boring. We were there 6 years ago.

I think its more interesting to discuss if amd Intel nvidia have the right people in the top positions. Discussing that will also tell a good deal where we as enthusiast want things to go.

 

[zinfamous]

I'm far from an AMD fan but given your forum name your post feels trollish

ya think? Honestly, I think it's one of the resident AMD fanboys here creating an alt account as a bit of a tasteless joke, which is a permaban offense. So bad move.

 

[superstition]

I'd say the odds for that to happen are pretty much the same as Elton John suddenly turning straight.

I really don't see why Elton John being gay is the go-to analogy lately. I've seen this twice now. Can we stick to computing and not sexual politics? Thanks.

 

[ozzy702]

This is no surprise. AMD are masters of shady accounting. With the disaster that is Polaris and the looming disaster with Zen only being on par with pre-sandy bridge IPC, the only way they can improve their financial situation currently is with hype-based marketing and dilution of shares.

Hasn't Zen already proven to have better than SB IPC? Clocks are of course the other half of the puzzle and will no doubt be much lower OC to OC than SB but for 99.9% of people that doesn't matter. Ivy to Haswell IPC and 3.2-3.5ghz would be competitive assuming the price is right which it very well could be given how small the chips are.

 

[frozentundra123456]

Hasn't Zen already proven to have better than SB IPC? Clocks are of course the other half of the puzzle and will no doubt be much lower OC to OC than SB but for 99.9% of people that doesn't matter. Ivy to Haswell IPC and 3.2-3.5ghz would be competitive assuming the price is right which it very well could be given how small the chips are.

Only in one benchmark picked and published by AMD. Overall, we have no idea, although I expect it to be better than SB, maybe IB/Haswell level. But I still dont understand the myriad of posters on here who say SB/IB level IPC at lower clocks is "good enough". That level of performance has been available for years, with exceptional overclocking ability. If one assumes SB level IPC and 15% lower max clocks, that is 5820k/6800k level performance at best in MT, and those cpus are available at around 400.00. If they can push performance significantly higher than 6 core HEDT, then they have a nice price gap to fit into relative to the overpriced HEDT 8 core, but (and this is just my opinion), I think the performance will be quite close to hex core HEDT, maybe slightly better in MT, but behind in ST. Basically a wash.

 

[CHADBOGA]

There's been a lot of times I was writing a response and I was about to mention something but realized it wasn't allowed in that particular sub forum. Most of the time I just say screw it and delete the whole damn post. Don't have the energy to split my thoughts up into several categories and chase down the right forum.

The rarely spoken about downside to over the top micro management.

 

[moonbogg]

Only in one benchmark picked and published by AMD. Overall, we have no idea, although I expect it to be better than SB, maybe IB/Haswell level. But I still dont understand the myriad of posters on here who say SB/IB level IPC at lower clocks is "good enough". That level of performance has been available for years, with exceptional overclocking ability. If one assumes SB level IPC and 15% lower max clocks, that is 5820k/6800k level performance at best in MT, and those cpus are available at around 400.00. If they can push performance significantly higher than 6 core HEDT, then they have a nice price gap to fit into relative to the overpriced HEDT 8 core, but (and this is just my opinion), I think the performance will be quite close to hex core HEDT, maybe slightly better in MT, but behind in ST. Basically a wash.

So you're saying my 3930K@4.6 is as good as a stock Haswell/Broadwell-E 6 core? Damn, that's not bad. Maybe I don't need to upgrade for a LONG ASSED TIME.

 

[AtenRa]

That level of performance has been available for years,

We had Core i5 2500K performance for a long time now but still there are people buying Core i3 Skylake which is slower in MT even at default clocks.

http://www.anandtech.com/bench/product/288?vs=1783

And im sure the majority will choose a Core i5 2500K today over a Core i3 6100 (for performance only).
This "we had this level of performance for years" is flawed, just because we have 10 Core Broadwell-E today doesnt mean nobody will buy the same level of performance 2-3 years later.

 

[frozentundra123456]

yes, and do you think amd is going to sell zen for the price of an i3? Obviously it comes down to price, but if they are going to have a compelling product, they need to have something like haswell ipc and good clockspeeds. Otherwise, they are going to be just competing on price with hex core hedt and i7 quads.

 

[Glo.]

Actually the very best predictor of future performance is past performance.

Actually its great that you have quoted sentence that is out of context.

Secondly. Yes best predictor of future is past, especially when the hardware in question is completely new architecture created from scratch, on new, next generation process node.

On paper, if we look at floating point, integer etc, Zen is on the level of Broadwell-E IPC. However its hard to judge overall performance of the CPU because we do not know how cache affects the performance of Zen cores. Also, we do not know the final specification of the CPUs on the clocks front.

There is nothing in the past of AMD that can indicate how will Zen perform.

 

[The Stilt]

On paper, if we look at floating point, integer etc, Zen is on the level of Broadwell-E IPC.

Could you point out the papers in question?

 

[Glo.]

Could you point out the papers in question?

http://www.anandtech.com/show/10591...t-2-extracting-instructionlevel-parallelism/7

Is this enough?
Integer registers: 168 Zen, 168 Broadwell
FP Registers: 160 Zen, 168 Broadwell
Load Queue: 72 Zen, 72 Broadwell
Store Queue: 44 Zen, 42 Broadwell

Decode: Zen 4 uops/cycle, 4uops/cycle Broadwell
Dispatch: Zen 6 uops/cycle, 4 uops/cycle Broadwell

No, I do not understand very well CPU architectures, however I do know that all of this is important for Single Thread performance(throughput of the core). The rest parts of the architecture that are affecting performance is Cache bandwidth, and core clocks. And the block diagram itself. However I think, based just on this, that there is no coincidence that Blender work on Zen have edged out Broadwell in similar environment.

 

[AtenRa]

yes, and do you think amd is going to sell zen for the price of an i3?

I believe ZEN APUs next year will take care of that category (i3).

Obviously it comes down to price, but if they are going to have a compelling product, they need to have something like haswell ipc and good clockspeeds. Otherwise, they are going to be just competing on price with hex core hedt and i7 quads.

Hex Core HEDT is high end, its high enough margins to get a lot of $$$$ for AMD. As i have said in another forum, even if AMD only manage to compete against the cheapest HEDT 6-core models like the Core i7 6800K ($400-450), its more than fine since today AMD dont sell anything above $180-200

Mainstream Core i7 like 6700K at $350 is way overpriced for a Quad-Core + HT SKU in 2016, a 6-Core ZEN with higher MT will be perfect at $300 IF it can OC to 4GHz and beyond. And again a $300 AMD CPU is higher margins than current AMD CPUs, so win win again even if they cannot compete against 10-Core HEDT.

But think about it the other way,

Intel Introduced 8-Core Haswell in 2015 at $1000, did anyone say we had this level of performance when Intel Introduced 6-Core Broadwell in 2016 ??? No because 6-Core Broadwell is cheaper and still makes loads of $$$$ for Intel.

If 8-Core ZEN cannot compete directly against 8-Core Broadwell in Q1 2017, then it should be priced lower. Simple as that, lots and lots of people are waiting for cheaper 6/8-Core CPUs with Ivy-Haswell IPC than what Intel currently offers in HEDT platform. And dont tell me we had Ivy/Haswell 6-Core CPUs 1-2 years ago, because those where 130-140W TDP and $600+

 

[FlanK3r]

Still, Skylake-X will be there no earlier than in end of Q4 2017

 

[Phynaz]

http://www.anandtech.com/show/10591...t-2-extracting-instructionlevel-parallelism/7

Is this enough?
Integer registers: 168 Zen, 168 Broadwell
FP Registers: 160 Zen, 168 Broadwell
Load Queue: 72 Zen, 72 Broadwell
Store Queue: 44 Zen, 42 Broadwell

Decode: Zen 4 uops/cycle, 4uops/cycle Broadwell
Dispatch: Zen 6 uops/cycle, 4 uops/cycle Broadwell

No, I do not understand very well CPU architectures, however I do know that all of this is important for Single Thread performance(throughput of the core). The rest parts of the architecture that are affecting performance is Cache bandwidth, and core clocks. And the block diagram itself. However I think, based just on this, that there is no coincidence that Blender work on Zen have edged out Broadwell in similar environment.

You don't understand CPU architectures but you know this is important how? Bigger numbers are better?

What's faster, a car with an 8 cylinder engine or a car with a 4 cylinder engine?