>AMD Zen ‘Summit Ridge’ CPUs and Intel ‘Kaby Lake’ CPUs Spotted in AIDA64 Changelog
It's fucking happening!
http://wccftech.com/amd-zen-summit-ridge-cpus-intel-kaby-lake-cpus-spotted-aida64-changelog/
>>53645287
Kek, intel at it again, no hexa/octa cores
At least they'll fit in the same sockets, for once
>>53645444
Just pay 1000$ for an octacore, goyim.
>>53645287
7 more months of waiting
>>53645287
40% ipc over excavator
how would that compare against intel?
i like how wccftech is still stating zen will have 8 cores when their original source that removed (they no longer cite a source) off of their announcement from a few weeks ago cited a 4chan post here on /g.
>>53645672
sandy bridge performance or so.
>>53645693
lmao this exactly
here is the original source they used on their announcement
https://arch.b4k.co/g/thread/52492812/#52493743
they removed it fast after people noticed it linked to 4chan.
>>53645672
haswell level
>>53645672
Excavator is about equal to Haswell.
Broadwell is 5% over Haswell
Skylake is 5% over Broadwell
Kaby Lake is 5% over Skylake
Cannonlake is 5% over Kaby Lake
Ice Lake is 5% over Tiger Lake
This means that Zen is about 10% over Tiger Lake so expect Intel to die soon.
>>53645750
>haswell level
Its worth remembering that 40% increase is regardless of node. Jumping to 14nm might let them tweak the design a little bit more and get to skylake levels.
THIS IS IT BOYS
>>53645750
Didn't AMD make a gcc commit and it showed 8 cores?
>>53645764
hehe memes
>>53645764
>Excavator is about equal to Haswell.
Absolutely not.
Excavator is substantially slower than Sandy Bridge per clock. Look at any Anandtech comparison between Kaveri chips and Sandy-Haswell. Excavator is only on average 5% faster than Steamroller in Kaveri.
>>53645693
AMD's own internal slides that got leaked stated Summit Ridge was an 8 core part.
Sweclockers even revealed this back in 2014 before anyone else had heard of the codename.
>>53645783
>"get in, onii-chan"
>>53645693
>Currytech uses 4chan as a source
RUMOR
U
M
O
R
MILL
I
L
L
>>53645771
The process node has literally nothing to do with the IPC of the arch.
>>53645787
AMD Zeppelin (Family 17h, Model 00h) introduces an instructionsretired performance counter which indicated byCPUID.8000_0008H:EBX[1]. And dedicated Instructions Retired register(MSR 0xC000_000E9) increments on once for every instruction retired.
+ core_complex_id = (apicid & ((1 << c->x86_coreid_bits) - 1)) >> 3;
+ per_cpu(cpu_llc_id, cpu) = (socket_id << 3) | core_complex_id;
>>53645801
There isn't a Kaveri chip with level 3 cache so you don't see the real power of the architecture.
>>53645836
Not inherently, but there are advantages to a smaller node that make make some things more viable. Having (say) more die space available for L2 cache might let AMD slap on (random number) another 12kb compared to their initial design.
years back keller has stated that they'd try to get the best of both worlds working. aka keeping up the focus on multithreaded 'muh cores' systems while increasing single thread performance.
octacore cpu inc
>>53645850
The arch is so latency heavy all the way from the first caches to the memory controller it doesn't even need L3.
>>53645850
L3 wouldn't do much. AMD had to increase L2 latency in Steamroller in a tradeoff for higher bandwidth. That change alone killed off most of the IPC gains. As a bandaid in Excavator they simply cut the L2 in half to try and improve latency a bit while maintaining the higher bandwidth. An L3 added to this design would be so slow it couldn't be effectively utilized without incurring a net penalty to performance overall.
>>53645912
Caches are balanced, you don't just add more and get more performance.
>>53645787
>>53645801
8 cores doesn't mean cheapo $250 processor so many users keep believing and sites like wccftech keep implying.
haswell technically scales to 8 cores on the consumer parts but only on the x99 platform and only for the low price of $1,000.
no doubt amd will have 8 core zens. amd has showcased 16 core zen parts for opterons. but the consumer line? most likely similar to shittel in terms of pricing and core count.
plus the fact that wccftech used 4chan has the basis for their article is hilarious.
even ocn picked up on the 4chan source fast.
http://www.overclock.net/t/1594483/wccftech-amd-zen-8-core-95w-summit-ridge-cpus-launching-in-october-already-going-through-validation#post_24985515
>>53645955
>8 cores doesn't mean cheapo $250 processor so many users keep believing and sites like wccftech keep implying.
It means whatever AMD considers effective to cut Intel, they don't need Intel's margins and frankly I thank God for that, Intel charges you $300 for a 120mm^2 14nm quad core die, they make massive money per wafer, AMD will charge you $300+ for a 8 core 300mm^2 CPU.
AMD needs marketshare, not trying to pull an Apple.
We need to make more claims.
Find a way to get wccftech's attention.
Then wccftech can become what we all want it to be: 4chan
>>53646033
Excavator will be equal to Skylake since AMD will include a water cooler with every CPU sold.
>>53645999
Summit Ridge is going to have a die around 125mm2 to 150mm2, same neighborhood as Raven Ridge.
Samsung's 14nm process has a greater than 70% area scaling advantage over industry 28nm bulk processes. A Carrizo APU would only be aroun 75mm2 in 14nm.
>>53646033
That post was made by the same autist who shows up in every AMD thread. Hes a pathological liar who pretends to work for AMD. Its pretty likely that its some forum cretin like Seronx, they're both just as mentally ill.
That being said, all he was doing is regurgitating things we already know. Except the dipshit thinks Samsung will be AMD's primary fab, and this absolutely isn't the case.
Summit Ridge *is* an 8 core die, the top binned parts have a 95w TDP. We've known this for two years already.
>>53646076
> Its pretty likely that its some forum cretin like Seronx
Is it wrong I know who that is? Do I need to spend more time on /g/ to refine my delusions?
>>53645999
you forget amd is still the same company that QUICKLY jumped on the $1,000 cpu bandwagon back during the athlon 64 days when they had an extremely competive product with intel.
it was called the athlon 64 FX. they even released super expensive lga fx series parts to be ran in a dual socket motherboards.
their first dual core parts, the athlon x2, went all the way up to $1,000.
then we look at their gpu divison. the same company that sells $1,500 dual gpu parts, $600 cards that get faced raped by nvidia's $600 parts, and $300+ house fire parts known as the 390 and 390x. but they charge those prices because the performance is still extremely competitive.
the only time amd slashes prices is when their ass is getting handed to them on a silver platter. like their entire faildozer line. the only reason why they sell their cheapo 8350 for $150 is because it barely keeps up with i5 in multitasking and i3's in single threaded performance.
don't expect zen to be cheap. they won't be releasing $350 8 cores with haswell or broadwell level of single threaded performance. expect them to be $800. cheaper than intel's but still expensive.
>>53646076
>Summit Ridge is going to have a die around 125mm2 to 150mm2, same neighborhood as Raven Ridge.
You don't know if Summit ridge will have more than a single die configuration.
I imagine the 4 core APUs coming later will be in that range, a bit bigger if AMD doesn't cheap out on shaders.
But the nonGPU summit ridge will surely be bigger than 250mm^2 if they want any decent core performance, 8 cores at 150mm^2 would be far too small individually and not very performant.
>>53646072
AMD is going to provide a 'fridge' with Zen like what Coolermaster has with their V10 heatsink.
200W of cooling power!
>>53646125
AMD will include a lifetime free supply of liquid hydrogen with their CPUs
>>53646110
Don't give me history lessons, I'm not underage.
AMD could wave their dick around in 2005 because the marketshare wasn't 5:1 for Intel at that time and they frankly had a superior product.
Now they're the underdog in every respect.
>>53646124
>I imagine the 4 core APUs coming later will be in that range, a bit bigger if AMD doesn't cheap out on shaders.
Summit Ridge is CPU only.
The APU is Raven Ridge. Summit Ridge is only a single die, and its 8 cores. Stop trying to talk about things you're laughably clueless about.
If Zen cores were as large as 6mm2 in 14nm silicon, they'd be substantially larger than an entire Module from any of the BD family architectures. The die is not any bigger than 150mm2.
You have no idea how area scaling works.
>>53646155
>liquid hydrogen provided to every home like natural gas as a utility
Livin' the future dream!
>>53646165
they're an underdog in the gpu arena and still charge nvidia level prices.
YOU CANNOT FORGET THEY HAVE SHAREHOLDERS. if they have a competitive product, shareholders want to milk that shit for all its worth. that means HIGHER PRICES.
>>53646186
Hahaha, Broadwell core sizes 7mm2, you expect AMD to be competitive with even smaller cores and a inferior process?
You deranged, shithead?
>>53646225
Meanwhile in real life the equivalent amd card to an nvidia card costs $200 to $300 less on average.
>>53646110
U know what's up. AMD fanbois are delusional if they think that the pent up demand won't artificially raise the msrp. Happened with the underwhelming skylake. And who really knows if floflo will be able to keep up.
also from the SWEDEN YES article:
>Nu kan SweClockers avslöja att först ut med arkitekturen AMD Zen blir familjen Summit Ridge med processorer innehållande upp till åtta kärnor. Samtidigt tas ett jätteskutt från 32 nanometer (för dagens FX-serie) ned till 14 nanometer, vilket tyder på att Samsung eller Globalfoundries står för tillverkningen. Framstegen ska inte minst sänka strömförbrukningen. Dagens snålaste modeller i FX-serien landar på 95 W TDP, någonting som sträcker sig upp till 125 och hisnande 220 W för värstingmodellerna. För Summit Ridge gäller däremot "upp till 95 W", vilket talar för att det kommer finnas strömsnålare varianter.
>Now vBulletin reveal that the first with the architecture AMD Zen gets the family to Summit Ridge with processors containing up to eight cores . At the same time taken a giant leap from 32 nanometers ( today's FX series ) down to 14 nanometers , which suggests that Samsung and Globalfoundries are manufacturing . Progress will especially reduce power consumption . Today's most economical models in the FX series landing at 95 W TDP , anything that extends up to 125 and 220 W for breathtaking monster models. Summit Ridge same is " up to 95 W ," which suggests that there will be more power-efficient variants.
>processors containing up to eight cores
>up to 95 W for Summit Ridge
http://www.sweclockers.com/nyhet/19954-amd-zen-forst-ut-i-summit-ridge-for-14-nanometer
yeah doesn't sound like $350 8 cores. sounds more like $250 - $350 quads, $450 - $550 6 cores, and $650 - $750 8 cores.
a quad zen apu would be fantastic for gaymens. broadwell level of performance along with a radeon level gpu that can crossfire with a dedicated radeon gpu. squeeze out free extra performance with every radeon user.
they also are leaving room for higher wattage processors.
>>53646240
>look how retarded I am.
Zen does not compete with Broadwell per clock. This is 100% categorically false. This is utter bullshit that Redditors have been spouting and its based on nothing other than a guesstimate from a Passmark score.
Lisa Su explicitly said their Zen based Opterons would be competitive in 80% of the server market, they are going to lose out in 20% of all common server workloads. Zen cores do not have an FPU to rival intel's in size, and they will not match them in performance.
The area scaling difference between Samsung's 14nm and intel's 14nm isn't that drastic either. Intel has a considerably smaller HD SRAM cell, and their CPP is tighter. Those are their biggest advantages/
>limited to 95W TDP
Fuck off, give me 140W, ramp up those turbo clocks.
Fucking electricity jews thinking I give half a hooter's shit for 40 watts.
>>53646292
>apu anything
>videogames
You're retarded m8, nothing to be surprised about though, coming from a nigger. Sweden YES!
>>53646305
Intel shills in full damage control i see. Zen confirmed for intel killer!
>>53646262
lmao are you high?
$620 fury x vs $620 980 ti
http://www.newegg.com/Product/Product.aspx?Item=N82E16814202155
http://www.newegg.com/Product/Product.aspx?Item=N82E16814487223
$510 fury vs $510 980
http://www.newegg.com/Product/Product.aspx?Item=N82E16814202187
http://www.newegg.com/Product/Product.aspx?Item=N82E16814487079
$450 390x vs on sale 980's
http://www.newegg.com/Product/Product.aspx?Item=N82E16814121973
http://www.newegg.com/Product/Product.aspx?Item=N82E16814125696
$320 390 vs $320 970
http://www.newegg.com/Product/Product.aspx?Item=N82E16814202164
http://www.newegg.com/Product/Product.aspx?Item=N82E16814487088
the list goes on and on
>>53646347
you do know amd apu's allow for crossfiring with a dedicated radeon gpu right?
with amd finally catching up to intel in cpu performance it makes their apu's more worth whiled.
a quad core broadwell level of performance with a apu radeon + dedicated radeon in crossfire = free extra performance.
it would make amd more attractive to gaymers since the integrated gpu won't go unused and they can toss in extra performance at a click of a button.
>>53646371
[reality denial intensifies]
>>53646398
>>>/9gag/ the place where people as smart as you congregate.
>>53646342
>he want pre factory overclocked amd product like the 3xx and the fx with a aio kit
>>53646416
intel shill out in full force today
>AMD 8 core
>95W
>Intel 8 core
>140W
AMD is a joke, their single core performance will be 50% slower just looking at the TDP
>>53646342
I don't care about electricity. I do care about heat, however.
>>53646489
Then don't buy a 250-300W GPU you stupid fuck
>>53646433
The real issue with those stupidly overclocked centurion fx chips isn't the aio they came with (once they stopped being OEM parts) as you can air cool them. No the real bitch is unliek a gpu a cpu doesn't bring its own motherboard and those fx chips really, really showed what happens when you stress motherboards.
Its why (iirc) only asus and asrock make mobos rated to handle them (and I wouldn't touch the asrock one).
>>53646478
Upper end of intel's TDP in their high core count parts is there for AVX. The pathways for those large vectors use a lot of power.
Summit Ridge will perform in the same relative range, but loose out in FPU ops. They'll likely clock a bit higher than their intel counterparts as well.
>>53646489
more electricity = more heat
so you do care about electricity.
>>53646499
And an oc 5820k will put out 200. Guess what the 5960k puts out.
>>53646509
>GloFo
>thinking they'll hit 4.2GHz+ with their shit Samsung process without blowing up
Good joke
>>53646549
>>53646548
The cost. Quit being pedantic.
>>53646562
At least it's better than 220W AMD HOUSEFIRE using 500W
>>53646551
>I have no idea what I'm talking about
Base clock for Summit Ridge will likely be a little north of 3ghz. No one is expecting 8 cores at 4.2ghz to be within a 95w envelope.
Go to /b/ if you want to shitpost.
>>53646562
That is total system power consumption,
>>53646562
Is that system wattage or solely cpu? I have a hard time believing even an OC'd 5960x can suck down more juice than some of the thirstiest gpu's in existence today.
To wit: 398w isn't too far from what a 295x2 pulls.
>>53646566
to care about heat is to care about power consumption. more power consumption more heat.
no matter what you care about electricity if you care about heat.
>>53646632
There are many on this board that seem to care about the cost savings of lower power parts. (Erroneously) I was mistaken, apparently, in thinking you might be one.
Have you taken your Ritalin today?
>>53646616
it is indeed total system power consumption. but if you subtract the motherboard, hard drives, and idle gpu, so roughly ~100 - 150 watts off, it gives you a good indication of the power draw from the cpu.
a 4.5ghz (OC ALERT! OC ALERT!) 5960x drawing ~400 watts at load would net you around 250 - 300 watts.
a 3.5ghz 5960x drawing ~300 watts would net you around 150 - 200 watts at load.
original source: http://www.kitguru.net/components/cpu/luke-hill/intel-core-i7-5820k-haswell-e-6-core-cpu-review/8/
they used prime 95 and measured at the wall.
another source at stock clocks, not oc: https://www.guru3d.com/articles-pages/core-i7-5960x-5930k-and-5820k-processor-review,8.html
>>53646701
5 pills so far
>>53646110
>then we look at their gpu divison. the same company that sells $1,500 dual gpu parts, $600 cards that get faced raped by nvidia's $600 parts, and $300+ house fire parts known as the 390 and 390x.
Which use HBM, which is very ineffective to manufacture, thus expensive, and has bad yields. If they were using GDDR5 they would sell them for less. They can't sell them at a loss, they aren't a service brand.
>>53646626
well the 5960x is intels 8 core. it comes stock at 3ghz. at 4.5ghz thats a full 1500mhz increase.
so yeah i can easily see it sucking down electricity like crazy.
>>53646741
the 295x used gddr5. the 390 series uses gddr5.
they're all priced competitively to nvidia's.
i replied back with links to show this.
here they are again:
fury x
http://www.newegg.com/Product/Product.aspx?Item=N82E16814202155
980 ti
http://www.newegg.com/Product/Product.aspx?Item=N82E16814487223
fury:
http://www.newegg.com/Product/Product.aspx?Item=N82E16814202187
980:
http://www.newegg.com/Product/Product.aspx?Item=N82E16814487079
390x
http://www.newegg.com/Product/Product.aspx?Item=N82E16814121973
on sale 980
http://www.newegg.com/Product/Product.aspx?Item=N82E16814125696
390
http://www.newegg.com/Product/Product.aspx?Item=N82E16814202164
970
http://www.newegg.com/Product/Product.aspx?Item=N82E16814487088
380
http://www.newegg.com/Product/Product.aspx?Item=N82E16814127878
960
http://www.newegg.com/Product/Product.aspx?Item=N82E16814487091
no one is really cheaper than the other.
>>53646416
>no counterpoints
>lmao 9gag xD
>>53646347
>apu anything
>videogames
The huge majority of non-mobile gamers use APUs.
>>53646759
I'm more concerned about the sort of motherboards thats can feed a chip 400w - even for the enthusiast range those have to be in the minority.
>>53646791
370
http://www.newegg.com/Product/Product.aspx?Item=N82E16814127886
950
http://www.newegg.com/Product/Product.aspx?Item=N82E16814487157
goes on and on. you can cherry pick and find one or two that are on sale and even cheaper and even nit pick HURR X IS FASTER THAN X! but they're all within the same price range of each other.
>>53646849
one of the reasons why x99 boards are expensive are not just because of popularity, but also because they are designed for higher wattage than intel's mainstream z series are.
>>53646110
>jumped on the $1,000 cpu bandwagon back during the athlon 64 days when they had an extremely competive product with intel.
You forget that Athlon 64 had 2x the IPC over the P4 in some cases, and that Intel was also selling $1000 Extreme Edition chips.
>>53646902
ok?
that just proves the point even more that amd will gladly jack up prices if they release competitive products.
which is expected.
>>53646812
>there are people who actually believe this
Who exactly is the target audience for these two chips?
>>53647515
>low TDP
>no GPU
What do you think?
>>53645444
intel is the only one with true hexa and octa core cpus on the market
>>53647608
For the small price of $999, kek
Fucking jewtel
>>53647608
>doesn't understand what a core is.
I'll give you a hint: its whatever anyone says it is as no company has ever defined what a core is and by certain definitions (like in that case against AMD) some of the real old chips don't qualify as cpus due to lacking the hardware to make up a "core".
Speaking of AMD
Polaris patch just landed for EIGHT (8) PCI IDs
http://phoronix.com/scan.php?page=news_item&px=AMD-Open-Sourced-Polaris
Looks like a Polaris is gonna be more than just 2 dies
>>53647759
It will be:
>Baffin (Polaris 11)
440
M440
450
M450
460
M460
>Ellesmere (Polaris 10)
470
M470
or something like that.
>>53645783
>Dat trigger control
Good man.
>>53645444
You can get an SB octacore for $70 right now.
