Zen price
Images are sometimes not shown due to bandwidth/network limitations. Refreshing the page usually helps.
You are currently reading a thread in /g/ - Technology
You are currently reading a thread in /g/ - Technology
![wafer_png[1].png](https://i.imgur.com/mFRJoU4.png "wafer_png[1].png")
Assuming 30% yields on a 200mm wafer, and assuming Zen's die size will be around 200mm squared looking at the recent die shot leak, AMD will need to sell the recently revealed 8-core Zen chips for $120 a piece in order to make 50% profit margins.
>each 200mm silicon wafer is $200
>processing each is an additional $1500-$1800
>a 10mm x 20mm Zen CPU would be able to fit onto this wafer 116 times
>116 x 30% yields = 35 working CPUs
>$2000 per wafer divided among 35 CPUs = $57 spent making each Zen CPU
A price point of $199 for Zen is totally doable in the long term, although they'll probably milk it for $250-$300 at first. Maybe even more.
Unless anyone has more info on Zen's costs and AMD's typical profit per CPU, and its yields, this seems like a pretty solid result.
![index[1].jpg](https://i.imgur.com/81PC1xq.jpg "index[1].jpg")
http://www.silicon-edge.co.uk/j/index.php/resources/die-per-wafer
>>55099702
>200mm
AMD has used 300mm wafer for CPU since around 2004.
looks like shit
I'm buying Intel
monitoring
>>55099736
Let me plug that into my formula, then. That's good to hear, I haven't been able to find any info on wafer sizes.
A 300mm wafer would probably price closer to $350, with a processing cost closer to $2000.
188 per wafer, *.30 = 56 working chips
2300 / 56 = $41
>>55099702
every body expected polaris to cost $400. The most optimists thought it would be $300. About the performances no one expected it to beat a GTX 980.
Don't try to guess zen price and performances, they'll just surprise us and btfo intel like they did for nvidia.
>>55099702
>30%
aahaaahahahaaa
AMD will never compete with Intel, they are just too far behind for a long time.
GPUs are a whole different scenario thank god.
>>55099702
Wafers are delicious.
>>55099702
Sounds tasty, I'll eat computer wafer
>>55099786
>>55100037
Why you all trying to start a shitfest? OP is literally just trying to guesstimate the cost of the new CPUs.
>>55099702
I'm curious, why is the yield for working chips always so low? I thought that we were 1337 in technology as humans, why is the yield not 99%?
I know there are imperfections in silicon wafers but don't we have tech to prevent them or is there some quantum gay shit fucking with us or something?
>>55099702
Mass producing the chips was never the costly part.
It's the R&D and employee costs.
>>55100132
Dust and human stuff
>>55100101
This isn't even the best part. That 8-core Zen chip is to compete with at least stock i7-5960X performance.
A $300-$500 AMD chip is going to compete with a $1,100 intel chip.
Intel is fucking done.
>>55100152
suuuure
>>55099702
If $120 is a 50% profit margin I doubt it'll start out above $250 especially considering the price of previous AMD CPUs and the fact that AMD is trying to gain market share.
>>55100152
Well that's good to hear anon
>>55100150
Do we really need humans in processor manufacturing plants? Why not just send robots there? Is robots tech still fucking shit?
>>55100132
I worked at the design and production of certain micro and nano structures for [secret project] using wafers and lithography systems and BY far the worst link in the chain are humans. The amount of things that humans fuck up is just amazing.
>>55100175
Robots aren't perfectly clean either.
>>55100160
>>55100169
For real though Zen is going to fuck intel's shit up.
However I do worry AMD becoming jewish and wanting over $500 for their 8-core Zen chip.
>>55099948
How big is the polaris die size?
>>55100212
most likely they'll have hasklel performance on 2017
>>55100212
>For real though Zen is going to fuck intel's shit up.
Said the increasingly nervous AMD fan for 100th time this decade.
>>55100225
It's not a "what if" anymore, AMD already tested the chips and found out they will shit all over intel.
http://wccftech.com/amd-zen-cpu-performance-double-fx-8350/
That 8-core Zen chip has double the performance of the old 8-core FX processors. That's a huge fucking jump in performance.
We're talking about more than 40% IPC gains over Excavator nigga, Shit's about to go down.
the hype around zen is out of control (because AMD's marketing is based around hype when they know the product will be bad)
sticking with my 2600k until i see reviews
>>55100219
14nm but it's really 16nm
>>55100282
dude the 4790k has almost twice the single thread performance of a 8350, and the multi-thread performance on the fx series wasn't scaling as much as they must so twice of an 8350 doesn't seem like much to me.
>>55100330
Are you mentally and physically retarded?
>>55100037
>AMD will never compete with Intel
Good thing you are not in charge of AMD.
>>55100375
>>55100282
This needs to stop.
>>55100282
So slower than a 4690K in single threaded applications?
>>55100427
Not sure, this is multi-thread performance and we don't know what CPU clock frequency the Zen or Bulldozer chip was running at.
As much as I want AMD to succeed we all know deep inside that they will never reach the performance of Skylake-Kaby Lake with Zen. At least in terms of IPC which is the most important things nowadays.
Daily reminder that Zen has reached over 40% IPC jump over Excavator.
Just throwing a guess out there but that means a multi-core passmark of 8,000+... for the 4-core Zen variant... clocked at 3.5-3.8 GHz... yeah
>>55100498
Ask this guy here >>55100282 he knows everything.
Well Summit Ridge seems to be well on its way.
https://www.youtube.com/watch?v=-IyrR8KsSfQ
>>55100561
>a multi-core passmark of 8,000+
so cute
>>55100636
>comparing a 3.5-3.8 GHz chip with one that turbos to 4.4 GHz
retard
>>55100561
Holy fuck, that is a rip-off, $500 dollarinos for a sandy vag i5?
>>55100662
GHz is meaningless outside of the context of IPC and architecture.
>>55100662
>clock speed matters
retard
>>55100678
>>55100682
/g/ Technology
>>55100662
comparing a NOBODY KNOWS WHAT FREQUESNCY with a 4ghz chip whit 4.4 turbo that was released on 2014.
>>55100672
>intel
>not being a jew
pick one
>>55100636
You forgot to include smt. Zen will include it.
>>55099702
Stupid Question, but why dont the manufactures just make a square wafer for higher yields and less waste? Or does it not work this way?
>>55100636
>>a multi-core passmark of 8,000+
>so cute
your quote is talking about the 4 core variant, xen goes up to 8 cores
>$2000 per wafer divided among 35 CPUs = $57 spent making each Zen CPU
>A price point of $199 for Zen is totally doable in the long term, although they'll probably milk it for $250-$300 at first. Maybe even more.
So now we resort to pulling numbers out of our ass, right ?
Nevermind that AMD has over 9000 employees to pay.
If Zen is not some out of the ballpark smash hit AMD will trap itself once again with its pricing scheme
It will be a vicious cycle that goes like this..
AMD comes out with a new product..
it performs alright, buit intel keeps the performance crown
To stay competitive against intel, AMD makes little or no money per CPU
AMD cannot raise prices because then the competing Intel product in the same price range will kick its ass
well better go back to the drawing board and try to make a new product..
>AMD hasn't made any money in ~4 years now
For AMD to make money , Zen's sales will have to be great/great OR they will have to price their chips higher, or some balance of both.
Lets put it like this..
if Zen costs $300 it competes with the 6700K
$200 = 6500
$220 = 6600
Zen will be priced based on its relative performance, and that in turn will dictate how much (if any) money AMD makes.
>>55100707
ok ... so comparing a nobody knows what freq. not even released (NOW WITH SMT) chip with an old 2014 4ghz chip (4.4 boost).
>>55100709
it does not work this way
>>55100688
You're an idiot if you think GHz is the only thing that matters in a CPU.
You're an idiot for even comparing it across vastly different architectures.
You probably don't even know what IPC stands for.
>>55100726
So? i'm not even comparing with the current intel gen. 6800 has 6 cores, 6900 has 8 and 6950 has 10 cores.
>>55100738
Ur a idiot each wafer costs $2000 period.
Zen will be $200 and AMD can still make half descent money.
>>55100747
Gotcha. i just did some reading and found out why they end up being round and that rectangular wafers dont work.
>>55100709
It's because the pure silicon crystal they start with forms like an icicle and they slice it
Slicing it into a square would just take of small areas where chips could possibly be fabbed if they were small enough to fit, and many are
ARM chips can be as small as 50mm2 I think
On a square wafer, that's a loss of like 40% of your potential chips
I think you retards are missing the point. Zen is NOT going to beat i7 skylake and probably not even i7 haswell. What it will do is take a huge shit all over haswell i5 and maybe even skylake i5 processors. That's all Zen needs to do and all that matters. Most people don't buy i7s, most get and i3 and some buy i5s.
But performance is not why people will buy Zen processors, that is just extra frosting on the cake. The reason people will buy Zen is because it will obliterate intel in value like it always has. The 4-core Zen variant could easily be sold for $100 and AMD could still make coin.
If a 4-core Zen processor has the same performance as an i5 and costs half as much then why on earth would people still chose intel? Not just in terms of multi-core performance but single-core performance which doesn't really matter much anymore to be honest.
>>55100738
Don't forget server profits, AMD hasn't gone after servers in a very long time
Zen is going after servers just as much as the first Opteron, which made mountains of cash
>>55100703
Intel themselves recommend a price literally half of that for the 2500K, you illiterate nigger.
>>55100788
>AMD can still make half decent money
Aiming for second place ,or in this case last place. Still doesn't look good, wont help their stocks, and wont please investors.
It very simple AMD needs to completely rethink how they make their CPUs and learn to compete with intel at the SAME level or they will ALWAYS be seen as the brand you go to if you are too fucking poor to afford the Top End product.
This whole "oh ours is cheaper but also doesn't perform equal to our competitors" will not work. People that are going to spend money on components want the best of the best without going into WTFBBQ price ranges. And Intel CPU's are priced just right for people to still consider paying for them and get superior performance to boot.
The only way AMD gets more per unit sales than intel is to basically give them away at a massive loss. i.e. Intel i7 sells for 300 AMD competing cpu sells for 50 bucks. It will never happen so they need to completely start over.
wasn't there going to be a 12 core desktop zen? Because that would be the one that I want
>>55100827
Jew detected
AMD is going to be delusional and price the CPUs at the same level as intel, maybe with a $20 price cut, until the bad PR and backlash from it forces them to reduce prices by 50% just to sell anything.
they did the same shit with bulldozer and it irreparably harmed sales.
>>55100822
this
>>55100822
>What it will do is take a huge shit all over haswell i5 and maybe even skylake i5 processors.
I remember something similar being said about faildozer and those CPU's ended up getting but fucked six ways till sunday by Nehalem when AMD claimed it was going to compete with/utterly destroy Sandy Bridge.
Zen will be the same. ie barly holding its own to ivy bridge
>>55100962
Really? What was an FX-8120 worth at launch?
>>55100979
>Zen will be the same. ie barly holding its own to ivy bridge
Zen doesn't need much to destroy intel you know. Even if they only manage to reach i5 ivy bridge IPC then that
s still great news for their 8-core flagship. In the end multi-core performance is what is going to matter as DX12 becomes more adopted and devs start writing programs with efficient multi-core support.
>>55100788
are you a retard ?
Do you understand AMD has other costs than simply wafer/processing costs ?
Do you go into a car dealership and whine about the prices and then bring up that steel costs $275 per ton and so Ford should sell you the truck for $2,000 and be happy with the profits you're bring them.
Because that is how stupid you look to everyone right now.
>>55100986
$200 for 1/2 the performance of the equivalent sandy vagina gpu, aka garbage
>>55100822
>if Zen sells for less, Intel will keep the same prices!
You know Intel is just fucking around at this point, right? Literally wasting over half of their die on a GPU?
Zen not increasing CPU performance one bit doesn't really do much. So what if they force Intel's prices down? Everyone is already running at least a Sandy i5. "Moar cores" is not going to help my gaming.
>>55100636
a 6700k is slower than the 4790k kek
MORE... CORES!!!!
>>55101048
>$200 for 1/2 the performance of the equivalent sandy vagina gpu, aka garbage
???
>>55101019
Ford makes trucks out of aluminum Pajeet.
>>55100986
Just look at the review
http://www.anandtech.com/show/4955/the-bulldozer-review-amd-fx8150-tested
FX 8150 MSRP $245
2500K MSRP $216
>Single threaded performance is my biggest concern, and compared to Sandy Bridge there's a good 40-50% advantage the i5 2500K enjoys over the FX-8150.
Its been the typical AMD story for like a decade if not more, AMD sucks in single threaded stuff and only shines in heavily threaded stuff.
>>55101143
I see, that is pretty bad.
>>55101115
>$499.6
what
>>55101143
>Its been the typical AMD story for like a decade if not more, AMD sucks in single threaded stuff and only shines in heavily threaded stuff.
Except it hasn't, the lackluster single thread performance wasn't a thing until Bulldozer, that's less than 6 years
>>55100822
>Not just in terms of multi-core performance but single-core performance which doesn't really matter much anymore to be honest.
because it wont beat it in single core performance, and as retarded as it sounds the only thing my current AMD cpu bottlenecks me in is fucking Crysis 1
I'm going intel for CPU and never looking back
>>55100905
Jim"kikes hitting the floor when they see my Zen core" Keller
>>55101098
yes it is, the 4790k is great, but if you look at the image you'll see that while the 6700 has 200 points less (in single thread) it scores just 200 points less on multithread, so it seems that it scale better than the 4790 (or it'll be ~800poinst slower)
>>55101174
Okay well you do that. The rest of the world will be getting Zen because no one gives a fuck about single-core performance now that we have things like Chrome and DX12.
>>55101174
Take a look at arma 3 is like a ST performance benchmark itself.
https://www.youtube.com/watch?v=mj7wnRA1yDo
>>55101199
pretty sure the difference in score is purely because of the lower turbo speed, the 6700k is ~10-15% faster than haslel clock for clock in most reviews.
>>55101233
arma is probably the worst eastern-european shitcode the world has ever seen. it's an OOP mess with zero consideration for cache locality or parallelization, because it's based on code written in 1997 when both cpu caches and multiple cores were expensive luxuries.
>>55101236
O_O 4.2 turbo instead of 4.4 WHY?
AMD is just shilling for the Koreans that are still unsure of going all out in the CPU industry. Going to be interesting what happens if Zen is successful.
>>55101266
probably for lower voltages. 4400 -> 4200mhz is a 5% clock speed decrease on an architecture that brought a 10-15% performance increase across the board.
basically, intel is just trying to jew people with power efficiency because they have no pressure to sell higher core count CPUs or CPUs with beefier clock speed/larger cache/etc because of the lack of competition in the desktop space from amd, via and arm's licensees.
>>55101172
ok my mistake about the time frame.
it is hard to find good quality reviews from back then so I can't comment further about pre-bulldozer
Plus of course Intel really, really sucked back then, it was all about the ghz for intel back then
>>55101266
likely thermal limitations..
but again.. don't get discouraged, clock for clock skylake *is* faster
>>55099702
>30% yields
>>55101256
I know, it's the only game i play, they've done some changes on the gfx recently and it works a little better (not much) and they're going to launch a new dlc with a more realistic map, but still drops to 30fps in their machine so...
https://youtu.be/RunKiT2svRM
>>55101300
i've heard that the ihs -> die thermal compound on the skylake chips is pretty bad and that they're not using enough amount of it, so it could be thermal limitation or just for power efficiency like >>55101298 said.
>>55100330
*8150
ftfy
>>55099702
Summit Ridge is a $500 CPU competing against $600~ i7Es.
Stop regurgitating this $200 bullshit. AMD isn't trying to be the budget company any more, they don't want razor thin margins on their products.
Raw cost to produce die doesn't factor in packaging, validation, transport, or headroom for vendor profit margins either.
>>55101137
yeah aluminum frame and brake drums and engine block and.....
>>55101353
as i said replacing the thermal paste between ihs and die on skylake reduces ~18º on load, not like I'm going to delide one to test, but it confirms that intel is using bad quality paste on skylake.
>>55101472
6700k cost ~300$ so zen must perform a lot better than that to cost 500$
>>55101552
People have no reason to buy high performance CPUs anymore, the smartphone/tablet market killed everything desktop. Nobody is buying a new computer every 2-3 years and hell, tons of people dont even own desktop PCs.
Intel is responding to market saturation by trying to maintain or raise margins as mucb as possible. Its a long term strategy to stay afloat as long as possible. AMD not being competitive really has little to do with it anymore, and besides 80-90% of the CPU market is in low power and/or mobile devices. Its entirely why intel tried so hard to focus on power efficiency over raw performance for their arches, even though it just isnt possible for x86 to scale as low as ARM
>>55101606
sadly ARMs will find harder and harder to improve performance as they get closer to x86 performance.
>>55101552
Its a more complex issue than that. The TIM provably is pretty mediocre, but the biggest issue is actually the height of the IHS over the die. Intel in packaging have been so sloppy that they're leaving a considerable gap between the die and heat spreader.
If you delid a nonsoldered recent intel CPU, adjust the shim height, and put the heat spreader back on without changing the thermal paste you'll still see a huge reduction in your temp delta.
intel is just cutting corners everywhere to pinch pennies. Making $50,000,000,000 a year in revenue obviously doesn't exclude you from being a cheap fuck of a company.
>>55101578
Zen is not a CPU line, its the name of a core architecture. Get it right or don't bother posting.
Summit Ridge is not competing against mainstream i5/i7 CPUs. It is competing against the heavily threaded i7Es with 6-8 cores. The reason why intel released a 10 core i7E with Broadwell-E is so they'll remain on top in all work horse benchmarks like rendering and encoding/transcoding video. That is what Summit Ridge is aimed at, not pissant entry level desktops and average consumers. It is HEDT, only one step below a bonafide workstation.
>>55101303
Honestly, This. As much as I hope AMD Releases their zen at ~$250 dollars the yield was rumored to be around 10% last I heard. It's really just not feasible.
>>55100937
>12 core
haha
I legitimately laughed at this. Thanks.
It's basically MOAR CORES again.
But this time with better IPC.
I'll keep buying Intel like I usually do every few years because I want better performance in DX11 games and emulation.
>>55100937
Actually I saw from a reputable leak that it would be a 16 core max on a single chip, not sure if it's for desktops or servers but the server boards will be able to allow two CPU slots and support 32 cores, 64 threads.
>>55101985
Even though you're shitposting, I'm curious if you've considered the Dx12/Vulkan performance benefit.
>>55102009
Yeah, but that's in the future, besides I have a feeling that Nvidia wants DX11 around as long as possible because it will be to their benefit.
>>55102003
There was proof that Zen is going to get up to 32 cores max. Although that's obviously for the server end.
http://www.extremetech.com/computing/230305-rumors-suggest-amds-zen-servers-will-pack-32-cores-serious-heat
>>55102003
You didn't see shit.
The Summit Ridge die, which is the exact same die configuration used in their new Opterons, is only an 8 core part.
16 core and 32 core Opterons are MCMs.
Each 8 core die has two memory PHY along the outside edge, and two GMI links. Dies are placed side by side with the GMI links facing one another in an MCM configuration.
1 die = dual channel memory
2 die = quad channel memory
4 die = octa channel memory
>>55102033
Most likely, although they'll have a harder time stopping Vulkan. That's Microsoft's job.
>>55101985
Now is a good time for more cores. Things are optimized for more cores now,. Just about everything has 4 cores, from mid end phones to the cheapest windows tablets to i5 & i7. Zen will force Intel to make i5 6core or improve IPC a lot.
Also, the IPC is a huge leap for amd instead of a step back.
>>55102075
>Zen will force Intel to make i5 6core
Things only retards say: The Post.
You might as well say that Peterbilt is going to force Honda to start building semis while you're at it.
>>55102075
Ah, the more cores = better meme
>>55099702
Couldn't they fit more dies on each wafer if they made them more square-shaped, instead of wide rectangulars (picture is somewhat like 2:1 width:length ratio), assuming similiar area per die?
Sorry, I'm not too good at math...
>>55102170
yes but they would need to completely redesign the core and it might not work the same.
>>55102170
The shape of a die is dictated almost entirely by the structures featured. You have a little bit of wiggle room with how things are placed, and how dense the logic is in certain spots, but you can't radically alter the layout. Everything on the front end of line is connected on the back end of line. You can't arbitrarily rearrange complex logic.
>>55100282
Oh yay! Now theyre back at phenom II level and we get an 8 core now oooo!
>>55100101
Any "CPU" manufactured by AMD is guaranteed to be shit and not be nearly as good as advertised.
Feel free to screenshot this comment and come back to it when it actually releases, it will be a huge disappointment and won't do anything at all to push AMD a little further ahead.
>>55100152
Wasn't bulldozer the Intel killer?
>>55100212
Yeah okay, we will talk once the chip is out, it sits in a compatible motherboard with a good chipset, then run some benchmarks and only then will we have any sort of objective foundation to build any further discussion upon. Right now all we can do is sit here with open mouths drooling on our lap and speculate about things that haven't happened yet.
>>55100326
He wasn't talking about the transistor size and fab process, he wanted to know the physical dimensions of the die itself.
I can't answer that, because i don't know.
>>55100400
Totally respond with a silly meme, that ought to really show him!
>>55102507
And despite that, he was still hilariously wrong.
This board is full of complete fucking morons.
>>55102543
Well, do you know the dimensions of the Polaris die? I'm curios and a quick google search has turned up nothing.
>>55102479
50 for rupees for you my son
>>55100037
Intel has stagnated, allowing AMD to catch up
>>55100282
Hopefully that means my 8350 will last for a while.. but I'm thinking of selling my mobo, RAM, and CPU if Zen is going to be cheap.
>>55100626
Yup, APUs will be pretty sweet. How long til 1080p 60FPS, the GPU next die shrink? I bet that res is 720p
>>55100738
>For AMD to make money , Zen's sales will have to be great/great OR they will have to price their chips higher, or some balance of both.
Now that they're finally off the 28nm, wasting silicon, they'll definitely pull a profit, but how much remains to be seen. If the chips are good, AMD will profit hand over fist. If they're ok, they get operating income and then some.
>>55102556
Thank you i can now buy chapati with rice many thanks kind sir
>>55102588
Do we know what kind of socket and chipset it will be on?
It kind of makes no sense to use a brand new CPU architecture with year old motherboard chipsets and sockets.
>>55102479
10 Goy Points™ have been deposited to your account - collect more to win a 10 Dollar coupon on selected Intel® Products!
>>55102548
They haven't been released yet, but I would guess around 150mm2. Zen cores aren't nearly as big as intel's, and if you roughly estimate DDR PHY on 28nm vs 14nm it would come out to about that. Though PHY aren't a uniform size, so its only an approximation.
>>55102630
Its socket AM4. Summit Ridge and Raven Ridge are both SoC, but can make use of an external chipset for added SATA, USB, and dedicated NVMe. The chipset is code named Promontory.
>>55102588
>Yup, APUs will be pretty sweet. How long til 1080p 60FPS, the GPU next die shrink? I bet that res is 720p
Nothing is known about Raven Ridge yet, other than its an APU with 4 Zen cores. It would be possible for them to have an IGP comparable to Polaris 11 on die with that chip, but I'm not counting on it. I gave up having high hopes for anything AMD years ago.
I hope they keep their Polaris/mainstream mindset with the pricing.
>>55102723
Yes, AMD is totally going to make a workstation CPU for the masses. Because thats something that the average person wants.
Makes perfect sense.
>>55102543
no, he's right. Samsung's 14nm finfet is actually 16nm. I wish I could find the RedGamingTech video I saw that in.
>>55102666
http://niceme.me
>>55099702
>calculating IC production costs based on the raw materials
>>55102755
Trying to correct people who know more than you do is just pathetic.
Samsung's 14nm process uses their 20nm planar BEOL, both Samsung and TSMC adopted this approach as a cost saving measure. Hence why they are called "transitional" FinFET nodes. It is not 16nm, and in terms of FEOL feature size Samsung does actually reach a gate length of 14nm nominally.
>>55102807
ok, that's cool. I was wrong. The more you know. But why are you such a fucking dick about it?
>>55102769
That's completely valid. The yields are what make it expensive.
>>55102882
Why do you think its okay to open your mouth when you have no idea what the fuck you're talking about?
Would you walk into a museum and talk about how Picasso actually painted with water colors? Stick to /v/ where you belong, cretin.
>>55102910
I bet you use ArchLinux.
>>55102927
>do something stupid
>get called out
>whine about people pointing out your stupidity
>respond with meme bullshit
Yeah you're definitely a little /v/ faggot.
>>55102939
Yeah, definitely an Arch user.
>>55102939
Different person, Idiot.
>>55099702
>A price point of $199 for Zen is totally doable in the long term, although they'll probably milk it for $250-$300 at first.
>they'll probably milk it for $250-$300 at first.
Nigga please. If you are seriously expecting anything lower than $500 for an 8 core processor, then you are really deluded.
>>55102910
Picasso did paint in watercolour though...
>>55102902
It's somewhat valid as soon it's used as a comparison to very similar products with nearly the same total production numbers, lithography methods, yields and R&D costs among other things. Do you even know how many masks this thing needs? They don't exactly disclose the entire manufacturing and R&D process.
>>55102956
I bet it's Richard Stallman, the fucking faggot.
>>55103007
Nah, he doesn't post here. It's too inane. This thread is exactly the bullshit he was talking about.
>>55103000
Virtually all Picassos works were in period oil paint, or even common house paint.
Grasping at straws, anon.
>>55103012
I can't blame him at all, until the mods ban all of the shitposters this board is no better than /v/
>>55103035
If you're the same anon, it's time to stop. You need to get back to work on your undoubtedly 'very important' project.
>>55102711
4 zen cores with vega gpu and hbm 2 would be pretty sweet
>>55103128
I'm anxious to see what they'll do with their 2018 parts. Zen+ cores, guaranteed to have high density HBM2 on package, or some other 3D stacked memory. It could still be 14nm, just a different process variant. That alone will be incredibly interesting.
Samsung and GloFo are still collaborating, but they're also each developing their own 10nm and beyond nodes. The kicker here is that Samsung is only aiming for mobile, but GloFo has all of IBM's IP now. They'll likely bring to market the 14/10nm SOI FinFET IBM designed for their future POWER chips.
If that actually were the case then it would be a revolutionary step in semiconductor history. Virtually ultimate channel control without needing any more complex topography like GAAs. A GPU the size of AMD's Hawaii could fit along side a few CPU cores and come together in a nice 50w package.
>>55101168
It's a -100% off sale. Get it while you can!
>>55100037
>GPUs are a whole different scenario thank god.
I hope you are not referring to RX480. The card that isn't out yet and there are no proper benchmarks for it. And with the "price tag" that nobody but big resellers will be buying it for.
>>55100774
>6800 - $440
>6900 - $1100
>6950 - $1750
>>55100561
lol the x4 is light years behind the 2500k.
what the fuck is wrong with that screenshot?
>>55102910
>>55102807
lol the turbonerd from mom's basement.
just fucking leave, please.
Intel has already won, whatever amd makes Intel have the money to make something twice as good.
>>55104542
Its Passmark, /v/'s token benchmark of choice. Go figure.
Excavator is up to 90% behind Haswell in anything heavily leveraging the FPU. In X265 encoding nothing AMD has can even hope to compete against a prosumer intel chip.
I doubt AMD is going to make significant gains on this front. In both die area and transistor count the FPU is the single largest unit in intel's i Core arch. We know from linux kernel patch notes that Zen only has 128bit data paths in its FPU, so it isn't natively handling 256bit vectors.
They might compete in more int bound things, especially if they have any kind of clock advantage. We'll have to wait and see. ES silicon that has been popping up is fairly low clocked. 2.8ghz~
Lower IPC AND lower clocks than intel chips would be terrible for AMD.
>>55104552
Shitpost more.
>>55100164
If they were this cheap they'd be shooting themselves in the foot, everyone would buy the CPUs and nobody would upgrade for the next 5 years
>>55099786
>>55100037
>>55100054
>>55100073
>>55100400
>>55100409
Kys shills/shitposters.
>>55099702
where is the R&D cost
>falling for AMDs shady shiesty deceiving marketing tactics
https://www.youtube.com/watch?v=04ITA1_XoqM
>>55103418
Zen chips - ???$
zen performance - ???
zen release date - ???
Will they support DDR4, PCIe 3.0?
>>55107136
Yes.
>>55104967
>Kys
Kill yourself.
>>55106841
Explained a long time ago by AMD themselves. The 1080 wasn't rendering correctly and actually wasn't working as hard due to lack of full snow rendering. The 1080 had the unfair advantage -- fewer obstructive particles simply made that image look richer.
>>55106841
>Being this late
>>55100175
I dont think you fully grasp just how small we're talking here in terms of chip destroying imperfections.
good god this thread is fucking dumb
>>55101236
heres the thing senpai, the latest intel chips have a single core that clocks far better than all the others, so single threaded tests arent necessarily as meaningful for them as previous chips single threaded benchmarks were
>>55099948
retards thought it would be 400$, anyone with a brain put it at 200-300$ for the 480
the zen is replacing the 8350 sku, that launched at 200$ (i remember it being 350 but i could be wrong)
from what i remember the process puts the zen at half the size of the 8350
amd could legitimately sell these for 100$, they could also sell it for 200$ and still pull more profit from them they they ever did with the 8350.
also, you are assuming only 30% yield, where as i assume they will get far higher and cut down/disable cores to keep prices low on the consumer side.
if zen is good and is sandybridge level, multicore they would beat out intel's current 6 core offerings, if zen is haswell/skylake level (some amd employees said at investor meetings its the first time since they joined they are matching if not exceeding their competition so grain of salt there) they would beat out 6 core soundly and rival the 8 core possibly be in between it and the 10 core
amd has a price range that goes from 200-400$ at a 'mainstream' level, or could go full jew and match intel price performance.
but amd is writing off the consumer computer side of their business when it comes to cpus, they could focus aggressively intel comparing prices with the enterprise market.
the range and factors that go into cost are to high...
>>55102420
second gen bulldozer was already phenom 2 single core
>>55107071
performance - sandbridge minimum due to what they said and them telling us it's performing above expectations.
price - 200-1300$
release - october 16 - april 17
>>55100905
If amd was out for blood, i would price it as low as i can, because amd would still pull massive profit over what they currently do, but intel would need to lower their prices to compete.
amd can live on the lower margins, do you think intel could?
>>55111036
Steamroller was at Phenom 2 release single core. They made small improvements for the hexacore and Llano. Excavator caught up to those.
>>55100937
server chips are 16 core
cern leaked a 32 core that put 2 chips on one socket.
>>55100324
>sticking with my 2600k
based gen but i want to do some upgrades this year
>>55104581
all depends on how much lower the ipc is.
>>55106841
He even says in the video that he knows that the textures are procedurally generated. How thick can you get?
>>55100175
Humans can clean themselves better than robots.
>>55111194
>cern leaked a 32 core that put 2 chips on one socket.
no they didn't. cern leaked that there is software support for up to 32 cores, the only known 32 core zen chip is the EHP, which is 32 cores + huge gpu + hbm in one package, which would be for a supercomputer or xeon phi equivalent product.
opterons will be two of the HEDT 8 core dies on an interconnect like they were in the past, this is a pretty large improvement from the bulldozer opterons which were 4 modules per die.
>>55109375
>heres the thing senpai, the latest intel chips have a single core that clocks far better than all the others,
turbo is for single threaded tasks, so if they had 'one really good core' per chip then they would have increased turbo clocks, not decreased them.
Bumping and buying to tick intel goys.
>>55100636
:^)
