$1723 for a 6950X

$1723 for a 6950X

Is Intel fucking serious?

Other urls found in this thread:

wccftech.com/amd-zen-cpu-performance-double-fx-8350/
youtube.com/watch?v=pEAyml3Fxgo&feature=youtu.be&t=40s
dresdenboy.blogspot.com/2016/02/new-amd-zen-core-details-emerged.html
youtube.com/watch?v=0THZo2vSInc
youtube.com/watch?v=K_JbFDe6afA
anandtech.com/show/10337/the-intel-broadwell-e-review-core-i7-6950x-6900k-6850k-and-6800k-tested-up-to-10-cores
twitter.com/NSFWRedditVideo

Watch yourself, goy, I'll have you for hate speech.

Intel are offering a quality product at a reasonable price. If you disagree you can take your 'opinions' to Stormfront.

More of a reason to wait for Zen. I really wanted them to bump the price of the 8core down, but they know retards will buy their enthusiast chips. If Zen delivers an 8core in the range of $500-$600 with at least Haslel performance, then I'm buying. I'm buying and not only making a new desktop, I'll make a little homeserver as well.

>techology

Intel seriously want you to buy a 4790K which is still comparable in performance to any of the stuff since.

I find this pricing suspect because it's not competitive with Intel's own Haswell-E chips. If you already have X99, why would you pay more money for the same CPU, and if you don't have X99 yet, why not get Haswell instead for considerably less money across the board?
There's no way the performance is anything but 5% faster at best, if only from efficiency gains in going to a smaller node, so it can't be that they're charging a premium for that.
Let's just wait for the official Computex launch.

Also, these are well into Xeon prices, so you may as well get that if you're inclined to spend a ton of cash.

i have a pair of 8/16 core/thread processors that do 3.3 ghz with a 20mb cache and 40 lanes that cost 150 bucks each.

Yes and you're going to buy it because you have no other options.

Not really that shocking.

They are "rejected" Haswell-EP chips which are expensive to make.

Welcome to what happens when Intel doesn't have competition. They've played this game before. And all you idiots sat there laughing about AMD being "FINISHED AND BANKRUPT" while ignoring the consequences of that. Once AMD is out of the picture, Intel is going jack up their prices and shift all focus away from desktop CPUs, since they'll have a stranglehold on that market. Then they'll focus on low-power solutions for tablets because that's where they actually have competition.

whats gonna happen to my 6700k

Zen is going to fucking destroy intel, they better hike up the prices as much as possible before 16 and 32 core Zen server CPUs arrive.

but what if zen doesnt?
what if it is just mediocre?
what if it ISNT EVEN mediocre?

>AMD
>mediocre

excellent meme, goy. 17 rupees have been deposited into your account

i am very serious.
this could be very bad.
if amd under delivers yet again.
this could have dire consequences.

wccftech.com/amd-zen-cpu-performance-double-fx-8350/

pic related: intel

I'm just as hyped for Zen as anyone. But if AMD isn't completely serious about its performance, they're fucked, which means you're fucked.

>wccftech.com
That is the raggiest of rumor mills. It's a coin flip whether anything they write is true or not.

I can't wait til Zen undercuts all this shit.

Unless it's got a performance parity, it doesn't matter if they undercut it. And matching a 10-core i7 isn't going to be easy.

youtube.com/watch?v=pEAyml3Fxgo&feature=youtu.be&t=40s

I will buy Zen, but let's be realistic and honest with ourselves. It'll stand toe-to-toe with the thousand dollar Haswell-E flagship 5960X, but it's doubtful the initial Zen flagship will compare to the Broadwell-E flagship. Zen will only compete with Broadwell on its second revision.

I have been steadily holding back each and every new release because the next is supposedly better. It's almost like there's no point buy now or ever...suddenly I'm depressed.

You joke, but wait until you see Intel jack their low-end desktop processors up to over $400, just like in the Athlon days.

A 40% IPC uplift over Excavator would put 8 core Summit Ridge in the range of Ivy Bridge-E to Haswell-E, with no hope of competing in heavy FPU centric workloads, and of course clocks are still in question. That definitely isn't bad, but it sure as shit isn't going to dethrone intel anywhere.
AMD might have a $500 8 core/16 thread chip that dominates at its price point, and thats realistically all people should be expecting.

>They have the monopoly

>no hope of competing in heavy FPU centric workloads
source?

For 140W, Zen could maybe hit 16 x 4.0GHz? It's not a huge core by comparison. Matching cache size would be more challenging. From what we know, that would meet or exceed performance parity. That's just theoretical, however.

The Summit Ridge chips seem to top out at 8 cores (2 x 4-modules), with A0 stepping currently running targeting a 95W TDP, testing 8 x 3.0GHz as conservative engineering samples, on the AMD Myrtle reference board (seems to be a Mini-ITX?), with one 16x PCIe 3.0 port connected to the CPU and the 4x PCIe 3.0 connected to the Promontory southbridge. This may not have a lot of relation to what they actually ship - it's an A0 stepping, they are validating and testing the tape-out, probably seeing how the packaging goes, how stable it is, how it clocks: we don't know any of this yet, and won't really until it comes out. Price will be a big factor, because as you've seen Intel are capitalizing on their lead and keeping their high-end costs very high, even as the competition starts to catch up.

They were looking at something like 16-32 core Opterons? We'll see if that actually pans out. Right now, we're just guessing based on the limited info we have available. Zen's going to go up against Kaby Lake in performance and price, and while Kaby Lake isn't going to be the 10nm step ahead Intel were hoping for, it's definitely going to be tough competition for AMD.

I think we're all sort of hoping for enough of a performance war between the two to start a price war, because that benefits everyone (except Intel's pockets).

ALU performance of early Zen cores seems to be good, right alongside Intel - which makes sense as one of Intel's big remaining multiplier patents has expired, so it may well be using the same general technique - but it's curiously slow with FMA3 for some reason? (Maybe they're still optimising the microcode on that one, maybe it's a hardware fault, or maybe it just bottlenecks the core and it sucks at FMA. Too early to tell.)

>AMD might have a $500 8 core/16 thread chip that dominates at its price point, and thats realistically all people should be expecting.
That should be enough though, most "enthusiasts" today only use i5s and quad i7s. They don't really need to compete with 10-core i7s that very few people actually buy.

Zen only has 128bit data paths in its FPU, and Su came out and said Zen would address 80% of the server market.

They can't compete against intel's monstrous AVX2 crunching FPU. By die area and transistors the FPU is the largest single part of intel's cores.

>They were looking at something like 16-32 core Opterons?
I hope so, but it's been so long since Opterons were viable that they'll probably still have trouble marketing them to businesses, which are often more hesitant to try new things than your average person, and your average person is hard enough to convince. Personally, I think they should just ditch the name and do a rebrand.

32 core Opterons are 4 die MCMs with 8 channel memory.
Thats the "disruptive" bandwidth Papermaster promised. 8 channels of 2800mhz ECC DDR4 is a lot. I can only imagine that they'll be able to use higher frequency DIMMs as they continue to improve as well.

Hm... Does Zen still use the fused 256-bit AVX design that Bulldozer uses? Is there anything else we know about Zen's FP scheduler and pipeline atm?

Ah, good point - didn't think of that! We simply don't know how it behaves there yet. Zen doesn't look like it uses the separate wide path approach of Intel's chips: which would mean less unused (albeit gated off for power reasons) transistors taking up core space, but maybe more bottlenecks - might well lag behind on things like video encoding - but maybe throwing more cores at it makes up for it? We'll have to see when it comes out, that's not an easy apples-to-oranges comparison.

I wonder how the hell it does scatter-gather?

Yep, P4 Xeons versus Opterons feel like a long time ago, but it also harks back to the Athlon 64 times when AMD chips were last truly competitive. It's not really "marketing them" to businesses exactly, but to big OEMs like HPE, Dell, and smaller ones like Supermicro. Really big companies like Google, Facebook, Microsoft, Amazon, will use whatever works and definitely won't have a problem testing them - if they are efficient enough for most of their uses, they will consider them. Still early days to find out.

A bunch of info on general core width was found in a linux kernel patch a while ago.
The FPU in Zen is 4 wide.

Bulldozer and Piledriver both had a 4 wide FPU, but this was revised in Steamroller which was only 3 wide, and had slightly higher throughput per clock. One MMX was removed, and some stuff was switched around.
All we know aside from that is that Zen doesn't use the FlexFPU design, obviously, as it isn't shared between cores.

dresdenboy.blogspot.com/2016/02/new-amd-zen-core-details-emerged.html

A lot of info about cache structures and handling in there, totally useless unless you're a technician though.

>It's not really "marketing them" to businesses exactly, but to big OEMs like HPE, Dell, and smaller ones like Supermicro.
It's not just that. Vendors are obviously not going to exclusively carry AMD, so they will always offer an Intel alternative even if AMD convinces them to ship their product. And when a tech director is given the choice between Intel, which they view as tried and tested, or AMD, many of them will just buy Intel out of name recognition. For example, one of them I'm acquainted with only buys Dell because it's what has worked, even if there are comparable alternatives for cheaper. And if their Dell rep comes forward and offers something with AMD he's going to be skeptical, because everything else they've got is Intel.

Most businesses care more about reliability than saving a little bit of money. They look at the added cost like insurance.

ZEN
E
N

You could easily pay more than double that for a top of the line Xeon.

>family

why everyone has to use slang now

>6 core 12 thread Broadwell-E
>$617
>lower binned chip with less PCI-E lanes is $434
>8 core starts at over $1000

Summit Ridge at $500 at the top end would probably be positioned pretty well. If Zen clocks high enough to remain even somewhat competitive then it'll do well for consumer HEDT.

Which has more than double the number of cores

Based Zen.

ITT: mostly gaymers who think that this will be used in reddit battletations

>>I think we're all sort of hoping for enough of a performance war between the two to start a price war, because that benefits everyone (except Intel's pockets).
AMD has had losses quarter after quarter. I very much doubt they'd like a price war. Price wars always benefit the consumer at the expense of the companies - which is a mixed blessing when one of the companies is strong and swimming in cash and one is weak and struggling.

You dont have to buy it. Free market is a thing.

So should I buy the 6800k or the 5820k :)))

>back in the day, on Sup Forums when bulldozer is going to release
>constant threads IT'S OVER INTEL IS FINISHED
>bulldozer comes out
>oh wow it's fucking nothing.

As much as I wish it was 2001 again when AMD was king I just have a hard time thinking it will ever happen again.

Zen... I want to believe.

Why would they want you to buy a 4790k?

The motherboard probably cost more than both of them combined.

thats a pretty misleading curve

They won't be king, but they'll hold their own.
Now, if Jim Keller had stuck around, and continued working his magic, yes, AMD would again surpass Intel. The guy is a rockstar. Tesla is very lucky to have him.

>AMD has had losses quarter after quarter.

Thankfully that is slowing down, but even if they get back in the black over the next couple quarters, they still have that 2 billion dollar debt hanging over their head.

There is absolutely no way that a dual die MCM is hitting 140w at 4ghz. That would be an absurd miracle which would be remembered for all of semiconductor history. It would be the most energy efficient chip ever designed. Were that possible then consumer Summit Ridge would have no problem exceeding 4ghz inside of its 95w TDP.

We know that 14nm FinFET doesn't do high voltages, we know that AM4 has the most tight power fine grain power control of any socket ever, including intels, and we know how all current ARM SoCs clock per volt on the process. In addition we've seen the super low voltages that Polaris GPUs are operating at.

I would bet Vegas money that Summit Ridge will have a base clock around 3-3.4ghz at 1.1v max with incredibly aggressive turbo speeds touching 4ghz momentarily on a single core. I'd also bet that its going to have virtually no OC headroom. Pstates on Zen are going to be ridiculous, and switch faster than anything else to date.

The Raven Ridge APU is going to be AMD's strong serial performance contender to compete against mainstream i5s and i7s.

Keller had nothing to do at AMD. He was a department head, not an engineer or low level technician. He was in charge of everyone else who was doing the grunt work on K12 and Zen. The only project we know he touched personally was Excavator which is why Carrizo was delayed after Kaveri being delayed a full year.
The high level architecture of Zen and K12 were both finished, work at already started on Zen+ as of May 2015. AMD didn't hire him on permanently to sit in a corner office and collect a salary until retirement. He was contracted to handle a specific job, and left when it was finished.

>They can't compete against intel's monstrous AVX2 crunching FPU.
Except they ALREADY can.

Bulldozer FPU might be shit, but it's actually pretty nice on 'SSE5' Floating Point - too bad most every Floating Point test is done in pure 8087 FP - which Bulldozer doesn't even do natively anymore since it's deprecated as fuck.

Wow.
Go read any Bdver programming guide, then read a Ivy Bridge or newer programming guide. Look at the number of ops per clock each can pull. Bulldozer/Vishera are 4x slower than Haswell at AVX. The way they handle 256bit vectors is incredibly inefficient, and throughput gets halved in the module when handling them. Intel's cores can natively handle 256bit vectors, and they do so at twice the rate of the FlexFPU working with a 128bit vector.

Don't try to participate in a conversation you're under qualified for.

Name a single thing, outside of Scientific computing (which 0% of Sup Forums is involved in, don't even try to lie) that is even compiled for AVX2?

>I don't use it while browsing the web so it doesn't matter!

This is strawman bullshit, trying to redirect focus from yourself for making a stupid comment. You stated something 100% factually incorrect because you're talking out of your ass.

Stick to Sup Forumsidya thread, you clearly don't belong here.

don't be a dick

I hope retail edge in winter gets us the 6900K, but almost certain it'll max out at the 6850K. 6 cores is still good enough for an upgrade I guess.

with today's slow progress you don't have to think about what's compiled to use it now, but what might be in the next five years or more.

>$400+ for the lowest-tier i7
>actually putting out a $1,700 consumer-oriented processor
Fucking jews

>Keller had nothing to do at AMD.

So, after Zen, he had nothing to do? Do you really believe that's how management works?

>He was a department head, not an engineer or low level technician. He was in charge of everyone else who was doing the grunt work on K12 and Zen.

You say that like it's a bad thing. You do realize that brilliant groups of engineers fail miserably all the time because of inept management, right? Engineers doing 'grunt work' don't make decisions in a vacuum; they do as they're told, and generally do a shit job without clear direction.

>The high level architecture of Zen and K12 were both finished, work at already started on Zen+ as of May 2015. AMD didn't hire him on permanently to sit in a corner office and collect a salary until retirement. He was contracted to handle a specific job, and left when it was finished.

Again, I highly doubt he would have nothing to do if he had stayed. Management jobs don't work like that. He would have gotten general direction from folks like Lisa, and then would have set about aligning his groups in whatever manner he wished to meet those expectations. He would have had plenty to do.

AMD didn't eliminate his position, so it's fairly obvious that someone directly under him got the post. Hopefully they know what they're doing.

I wouldn't be shocked if the prices start out high to milk early adopters, and then the top model stays high and the others slowly decline.

>So, after Zen, he had nothing to do?
Yes, nothing AMD needed his expertise for, or nothing they felt like continuing to pay him for.

He didn't have his position eliminated, and he didn't quit. He met the terms of his contract and went on to do other things.

I remember when people were hyping up Bulldozer and saying the single core performance would be almost as good as sandy.

>Double FX 8350

You could just buy an i5 and get that already :^)

Do you have a source for him working on contract?

It was on AMD's investor page when he joined the company.

>using the smiley with a carat nose

It's actually more comparable to the haswell-E series. The price point will be the real question.

>quality product
>140w

housefires

>shitty 28 lane 6800k costing as much as the 40 lane 5930k

Wow that shit better oc to 5ghz

My guess is :
Top end Summit Ridge sells at $500
Top end Raven Ridge sells at $250

Does it really cost Intel more to make more cores? Whether it's an i3, i5, or i7 xXx_ExTrEmE_xXx, they just put silicon into the machine and out comes a chip. When a new version comes out, every version should be the "best" version. The tiered system of economy chip, mainstream chip, and enthusiast chip seems artificial, like they gimp certain chips just to sell the at a lower price. When you have a new machine that makes the new best chip, why make any other chip? Unless processors are like growing tomatoes where some come out ripe and some come out rotten, where the ripe ones are i7s and the rotten ones get sold as i3s.

It cost money to turn sand into facebook machines

Intel basically hires alchemists

Wafers have a base cost, running the machines has a base cost, the raw materials have a base cost, the man hours involved in design have a base cost.

Binning is not gimping, its recouping a loss. Intel isn't intentionally disabling half an i7 just to sell an i3. Parts of the die are defective so that i7 gets turned into an i3 so they don't have to just throw it away and write it off as a total loss.

I think they're trying to squeeze as much money out of people as possible before their profit margins get smashed by Zen, ARM and power9.

>>Does it really cost Intel more to make more cores?
Yes, to a varying extent. They design a chip with X cores and make a bunch of them. The yield varies. Some work perfectly, some don't work at all, a large number work, but not completely. These often get cores disabled and sold as X - 2 core chips or whatever. Fancier version of the clock-speed binning they've been doing since forever. Nvidia also does it, as does AMD with goth CPUs and GPUs. As a process matures and yields go up, more chips will pass with no defects, and you'll have a situation where it doesn't cost them any more, most of the time, to make the top-end model.

What I'm trying to say is... you have a machine that turns sand into CPUs. The labor involved is putting the sand into the machine and taking them out. It costs you the same "amount" of materials and labor to make an i3 as a 10-core i7. Once you have finished the research and development and made a better product... why continue making any lesser product? The labor and materials are the same. You're utilizing labor inefficiently by allocating some of it to the inferior product. You should have all your machines change to making the newer version.

Zen is going to take maybe 10% of the enterprise market, ARM micro servers probably won't be a very big sell, POWER9 is going nowhere but HPC, and I have a sneaking suspicion that a few current POWER customers will be switching over to comparatively cheap Opterons.

K12 has an extremely uncertain future.

I can get my 5820k to 4.7GHz, so I wouldnt be shocked to see broadwell-e hit 5GHz on the high end of the silicon lottery, most will probably be ~4.6GHz or 4.7GHz though. With more cores it's hard to get it stable.

If this is the case, AMD might actually off an alternative solution again. I don't know if I'd actually buy a product from them given mishaps in the past I've dealt with, but I'd like to at least see the option. As it stands, Intels just a better option for both desktop and server CPUs.

think of it like babies. You don't know what you're gonna get. On average you'll get a functioning one. If you get shit luck it'll come out retarded. (amd)

So it is like my analogy of farming where some product comes out bad, but still gets sold, just not as the "premium" brand that's big and juicy and wins first prize at the state fair. The shitty ones can at least go to the pet food factory or something so they're not wasted.

actually it isn't, if you take die size into account. You can fit more tiny i3s onto one wafer than giant 10-core server chips.

also they continue making lesser products as a market segmentation strategy. The people who are happy with any old thing and won't pay extra for something faster get their own product, the people who will definitely pay more for the fastest thing they can get get their own product, so as to charge them very different prices.

Also something that's done by many other chipmakers and in almost every industry everywhere

>The labor involved is putting the sand into the machine and taking them out. It costs you the same "amount" of materials and labor to make an i3 as a 10-core i7.

Yes but then you have RD costs to recoup.
You are forgetting that intel spends BILLIONS on research and development every year.
As for the i3 and i7 , let me give you an anlogy.
Intel , like say Ford doesn't make much money on their base models, like the i3 or the Ford Fusion,
its enough to keep the lights on, but not much extra.
So what you do is for a much larger profit margin you offer other things people might like.
Like the $3000 navigation system or in this case the $1700 10core monster.
This might be where your profits come from.


>The labor and materials are the same. You're utilizing labor inefficiently by allocating some of it to the inferior product. You should have all your machines change to making the newer version.

Old products are still selling so why stop making them ?
Older tech tends to have better yeilds than newer tech.
The machines themselves cost millions of dollars each.
A new fab costs as much as a nuclear reactor these days, and then you have to replace the machines inside every few years too.

Mannnnnnnnnnnnn all this amd hype makes me crack a fatty.

But.. last time I bought into amd it was for steamroller... and we all know how that ended up.

>10core/20threas Broadwell-E $1700 instead of $1500 predicted
>intelfags were unironically expecting it to be $1000
LOL

They have two production lines.
4 core + gpu chip on newest one for laptop and low end PC. These are close to ARM SoC in size, if there was any competition they would sell for about $150. As they are a small chip partially defective ones are not likely to be common.

18 core model, made on previous gen fab. All the lga2011 products are this chip. A huge scam, high end ones are sold for $4000+, as they always make them on the older production line yields are likely very good. They're no bigger than high end GPUs While there would be some with defective cores you would expect 12 or 14 cores left over not 6 or 8. In comparison I think the 48 core cavium thunderX sells for around $500 per chip and the top of the line power8 is a little over $1000. An 18 core xeon could likely be sold at $1500 and still make a huge profit.

>ARM micro servers probably won't be a very big sell,
ARM is chasing high powered servers now, see cavium thunder X2 and qualcomm server chip.

>Steamroller

Mistakes were made, and unfortunately it wasn't worth the money necessary to correct them given that the BD family was end of line at Excavator. AMD made enormous gains over Piledriver and they ended up lost in virtually every single real world workload because L2 latency significantly increased in a trade off for improved bandwidth. All other ares of the core which improved significantly basically got swept under the rug because of that one singular fact.

The architecture could be salvaged with $10-$20 million and tens of thousands of man hours, but starting from scratch would likely be just as much work.
The real comparison that Zen has to stand up to is a theoretical 8 module Excavator chip. If Zen doesn't have higher multithreaded performance than 8 Ex modules it spells pretty bad things for AMD's efforts.

You know, I thought the same thing about cars. Why not make every car look like a Bugatti? It's not like the shape itself is expensive. Their engines are probably machined by robot assembly lines just like those of Honda Civics. Does it really cost $2 million to make a car? You're paying for the branding. Honda could easily make a car just like it for $20k and undercut them. Then competition would bring down the price of "sports-looking" cars and they'd be ubiquitous.

>Does it really cost $2 million to make a car?
Considering the parts are built to order, and are often crazy things like carbon fiber everything (including the engine block), yeah it probably costs a lot.

luxury carmakers like Bugatti also have higher costs no matter what they make because they don't make as many. The big carmakers can make cars cheaply because they roll off the assembly line by the hundreds of thousands. They can easily afford to put a million-dollar robot in their factory, because if it saves so much as a dollar per car built, it pays for itself quickly. When you make, what, a thousand cars per year, you can't do that, and the cost per car built is much higher.

6950X Benches out

youtube.com/watch?v=0THZo2vSInc

youtube.com/watch?v=K_JbFDe6afA

By not releasing any chip that gives better gaming performance for the price.

x% price difference should to x% perf difference. With these chips you're looking at hundreds of percent increase in price for 5% increase in performance.

>not doubling the price

what are you a friend of the goyim now?

>$1723
>Loses against the i7-6700K in benchmarks

Anand

anandtech.com/show/10337/the-intel-broadwell-e-review-core-i7-6950x-6900k-6850k-and-6800k-tested-up-to-10-cores