IT'S HAPPENING!!

Does anyone think zen chipset mobos will come in mitx form factor. There should be no reason for it not to right?

>>57359256
Of course they will. mITX is popular for oems.

>>57359256
probably with one of the poorfag chipsets 330 or whatever it's called

>>57359256
They already have am4 specifications out that could potentially be nanoitx. It would be astronomically stupid to not have them also be mini itx

>>57359256
Looking foward to this, can't wait to do a mITX Zen build.

File: 7e3[1].jpg (31KB, 600x750px) Image search: [Google]

31KB, 600x750px

>>57359271
pls be real ;_;

>>57359341
When has a non-server chipset/CPU architecture NOT had an mITX mobo?

>>57359331
Exactly my thought too. Going to try do something in a dan case with a 480 or whatever is equivalent in pricing when zen does release

Honestly hope zen drives intel out of business

a person can dream :^)

>>57359103
>pooinlootech
>moar coars

>>57359366
meant to reply to OP

File: 34oyjyx.jpg (514KB, 1563x1260px) Image search: [Google]

514KB, 1563x1260px

>>57359411
Too big?

inb4 pic related

>>57359455
that's a cute board

>>57359103
>32 AMD cores
>Barely fasts than 8 Intel cores

>>57359531
B-but Vulkan and DX12 Anon! They will save AMD! HEY GUYS LOOK AT THIS CHART, INTELFAGS BTFO!!!

>>57359531
"up to 32 cores"

>>57359531
le prove it :^)

INTEL IS DEAD

B
T
F
O
'
D

CurryTech rumor mill is not a valid source.

>>57359599
http://blenchmark.com/cpu-benchmarks

has anyone of you fanboys read the article posted as reference ? It doesnt prove shit and even states (opinion of author) that the test was conducted with one of the server processors

File: intel btfo.png (57KB, 551x895px) Image search: [Google]

57KB, 551x895px

>>57359599
it's not the original sauce
http://blenchmark.com/cpu-benchmarks

>>57359531

>souce
>my ass

>>57360068
t. Pajeet
2 rupees have been deposited into your account

>>57360087

>damage control

Fuck off and go back to >>>/trash/

>>57359103
>wccftech

Anyway I've been waiting for a while now for Zen to come out and see how good it is. I've been waiting to upgrade my ivy bridge dual core i3. I know I can just get a used i5 for around $150 but if Zen is good I want to support AMD and help break up the "monopoly" a little.

>>57359676
It's stupidly simple to insert fake results in blenchmark.

>>57359103
It's on par with a Xeon E5-2680v2, which is a 10c*2.8GHz Ivy Bridge.

Not exactly earth shattering, but a pretty damn solid sign if it's really just a ~3 GHz Summit Ridge chip.

>>57360186
It definitely is a Summit Ridge chip - the big question is which one ?
An octa-core CPU being released for the desktop market (dreaming about that) or (realistically speaking) a server cpu with more than eight cores ?

>>57360255
Considering the fact that we know the desktop motherboards are available, we know the desktop chip is right around the corner and we haven't heard a peep about any server chips, especially not 32 core ones, this is very likely the desktop chip.

>>57360255
I (pretty clearly I think) meant a single SR chip and not a MCM Naples.

>>57360255
Pretty much all of the media got this wrong, but the single 8 core die itself is called Zeppelin. Similar to how the 4 module Bulldozer and Piledriver dies were called Orochi. Rather its a name for the layout of the die itself.

Summit Ridge is the name of the single die packaged consumer CPU.
The two die MCM Opteron is Snowy Owl
The four die MCM is Naples

Will AMD have anything to counter thunderbolt?

>>57361797
Thunderbolt is a joke until proven otherwise, meant mostly for direct attach storage in overprice trashcan-esque workstations that lack onboard 10+GbE or any other way to get high bandwidth network access.

External GPU enclosures for laptops are joke, since gaming will just be bottlenecked by toaster tier CPUs.

File: mwc2014-amd-pres-2.jpg (615KB, 3126x1752px) Image search: [Google]

615KB, 3126x1752px

>>57361797
AMD came up with a solution years ago, and its a VESA standard.
DisplayPort has a sub-standard called DockPort.
One DP connector can be host to any number of devices, utilizing the full bandwidth the connection provides. It can also provide 100w of power. This was available long before any USB3.1 devices were in production.


Thunderbolt isn't used because its good. Its used because intel paid Apple to adopt it.

>>57361797
>buzzword for proprietared PCI-E and DisplayPort
why?

>two weeks before zen release
>intel launches a processor twice / thrice as fast at same price without talking about it prior
kek

>>57359103
>Intel Xeon E5 2680 v2
>v2
That's a fucking Ivy Bridge, not Skylake or Broadwell. The Xeon E5-2630 v4 is a Broadwell 10-core clocked at 2.2GHz.
2.2 Motherfucking GHz.
That's it.

Fucking wow.
Amazing, an eight-core Zen clocked at a full GHz more can barely match a 10-core Broadwell
Now watch this
https://www.cpubenchmark.net/compare.php?cmp[]=2758&cmp[]=2792&cmp[]=2794
https://www.cpubenchmark.net/compare.php?cmp[]=2758&cmp[]=2565&cmp[]=2800
It's barely on par with a 6-core Broadwell-E.

That is pathetic.

>>57362167
10 * 2.2 GHz = 8 * 2.75 GHz, genius.
If the Zen is clocked anywhere around 3 +/- 0.2 GHz, that's spitting distance to IPC parity, for a CPU with only half the DDR channels.

Nobody expects Zen to outright beat Broadwell, but these results would be much too close for comfort for Intel.

>>57359256
Duh? AM3/+ is too hard to fit on mini ITX, only a single ITX board was made

FM2+ was better but still only a few boards

AM1 is more SoC like Zen will be and it has a ton of ITX stuff

>>57362798
>AM3/+ is too hard to fit on mini ITX, only a single ITX board was made
Correction, the chipset and sufficient power delivery circuitry is really hard to cram on an ITX board.

>>57362716
And Skylake-E releases short afterwards with the much more feature-rich x299 chipset.
Dumbass

>>57362883
That isn't even a problem, its just that board manufacturers never wanted to offer a really expensive AMD board.

Some ASUS intel mITX boards have VRM on a daughter board, and they're 10~ phases for a chip drawing less power than an FX 8350.

>>57362889
> chipset mattering worth a fuck in enterprise when CPUs now getting integrated NICs in addition to PCIe/NVMe lanes

yeah those USB 3.1v2 ports will really be something, huh?

>>57359103
>16 core chip
>as fast as a 10 core chip from 2013
well if power consumption is good and the price is good it might work

>>57362952
If that's the case, there is no way they could match the pricing latest Intel chip. Sell it cheaper,maybe I could buy one.

>>57362936
>now getting integrated NICs in addition to PCIe/NVMe lanes
Not Zen, but Broadwell-D already have integrated 10Gb NICs and PCIe lanes have been around since PCIe was standardized. If you mean NVMe, then that's nothing more than 4 lanes of PCIe and a different transfer protocol. Nothing amazing.

Why does AMD use that shit brown color and poo smear logo for Zen?

>>57359366
AMD Bulldozer/Vishera

File: 1478052124399.png (522KB, 732x548px) Image search: [Google]

522KB, 732x548px

>yfw 8-core laptop cpu

>>57363151
Entirely possible, but unlikely since it doesn't have the intel badge. 8 Core/16 thread Summit Ridge could fit inside of 35w at 2.1ghz. Some mobile workstations allot way more than that for their CPU.

>>57363026
Zen looks a lot like clusters of Haswell/Broadwell-D chips.
Each Summit Ridge/Zeppelin has 2x 10GbE controllers and some SATA lanes in addition to the more standard PCIe lanes.

This is probably actually a really smart strategy.

File: 1275699848886.jpg (376KB, 1280x1024px) Image search: [Google]

376KB, 1280x1024px

Hope.

>>57359676
>http://wccftech.com/new-amd-zen-blender-benchmarks/
>>57359676
>>57359676
>>57359676
INTEL CONFIRMED FOR KILL

WE KILLED THE JEWS

>>57363402
>putting SATA lanes on the CPU
There is no advantage in this. Why not increase the memory buswidth by adding two more channels? Or swap the SATA lanes for some SAS3 ones? SAS is far more useful in an enterprise environment.

>>57363686
>SATA lanes on the CPU
Jesus i need a new PC, my memory controller is still on my fucking motherboard.

File: 1467522505791.jpg (126KB, 350x650px) Image search: [Google]

126KB, 350x650px

when the fuck is this shit show coming out so i can upgrade my compoorter?

>>57359103
>As good as 10-Core Intel Xeon
>AMD Engendering sample

Keep in mind, this might be high end CPU that costs as much as 10-Core Intel Xeon.

>>57363686
because SATA is literally just 2 differential wire pairs and DDR4 is about 300 pins?

>>57363780
It's a shit 8-core that can't even clock higher than a i7-960 from almost eight years ago. Delivers roughly the same performance as a six-core Broadwell-E at best.

File: 1406685659319.jpg (899KB, 2448x2448px) Image search: [Google]

899KB, 2448x2448px

>tfw no patches nor optimisations for future cpus features on win 7

File: 1476561393478.png (102KB, 362x259px) Image search: [Google]

102KB, 362x259px

Im glad AMDrones eat this bullshit up and buy AMDshit so I can stick with quality Intel/Nvidia hardware without fear of them being split up.

We should really take a moment to appreciate AMDrones and their AMDelusions. Thank you.

>>57363802
And SAS is four pairs. It's a simple matter of splitting the pairs to convert them into four SATA ports from a single SAS bus.
>DDR4 is about 300 pins
Well now, they should have thought of that before leaving us with two separate memory channels and no hope of a comparable quad-channel memory and relying on RDIMMs to break past 64GB of memory.
And this is supposed to be a serious contender for data-center and virtualization use in the enterprise world.

>>57363883
ok Isaac

>>57363912
Sorry, eight pairs. I was only counting the transmit wires.

>>57363849
>install Gentoo
>run Windows 7 as a VM

>>57359531
DELETE THIS

>>57364106
I love my FX8350, especially now that more stuff is taking advantage of eight cores and it somehow stayed relevant, but 32 coars is just insane, If the cpu truly is going to be 32 cores, it'll be an obsolete piece of shit before anything can leverage that.

>>57364247
>I don't know what a server is

>>57359531
does blender use more than 8 threads? is the article even real?

>>57364283
>High ass TDP AMD chips
>Server usage
Kek, c'mon Anon.

>>57364306
are you retarded anon?

>>57364318
No, I'm just saying most server companies would like a chip that does more with that 95w or whatever it is, I'm not knocking AMD, an FX is more than most people need and it is cheap.

>>57364339
Zen is on 14nm finfet, iit's going to be around the same TDP as intel chips

>>57363912
SAS isn't long for this world, or at least enough to putting controller in CPU dies.

HDDs are being relegated to quasi-cold storage for which SATA is plenty sufficient, and NVMe will be the standard for SSDs for the next decade.

We'll start to see server CPUs with gobs of PCIe lanes so that 1U chassis with 10 u.2 drive bays etc. will become the norm.

File: 1472857710460.jpg (43KB, 480x451px) Image search: [Google]

43KB, 480x451px

>>57359103
>taking CurryTech serious.
Will you ever learn /g/?

>>57364378
Intel 14nm better than Samsung/GloFo 14 nm by a significant margin.

Zen is more a TCO play than a pure perf/watt one.

>>57364405
The current gen opteron stuff is about equal to X99 stuff on TDP

and ya it'll be less for ownership, in addition to probably offering more cores/threads for less

>>57364405
>>57364378
Naples Opteron tops out at 180w for 32c/64t.
Intels current top Xeon is the E7 8890v4 with 24c/48t and its a 165w part.

>>57364656
the clock rates for Naples aren't going to be very pretty, when you consider the extra shit it has to deal with:

> 14nm process disadvantage vs. Intel
> extra South Bridge IO controllers eating into TDP vs. Broadwell-EP
> having to route data through external (interposer based?) GMI links in MCMs

despite all this I'm actually quite eager to see how the Snowy Owl parts perform and are priced though.
14c Broadwell-EPs range from $1400-$2100 depending on clocking/power, and AMD could easily start a price war in this segment as well as with the full Naples.

>>57364395
>SAS isn't long for this world
Nigga, SAS4 is releasing in less than a year, has over 22 Gbit/s transfer speeds, and - most important of all - is backwards compatible with previous SAS connectors and hardware. NVMe is just a storage protocol, SAS drives can use NVMe standards (the U.2 connector started out in life as a SAS connector). SAS will always have a place with enterprise SSDs because of this.
You can quit pulling shit out of your ass now.

>>57364400
It's not even great news, it shows that the eight core Zen is weaker than a stock i7-6850K, which has six cores.

>>57364811
Zen follows the same clock scaling as Excavator.
2.4ghz for a 32c chip isn't bad, not even remotely.

>>57364283
blender uses as much threads as it can find, meaning that hyperthreading doubles the threadcount up to, i believe as much as 16 threads

>>57364834
there is no way in fuck that 32c Naples will clock anywhere north of 2 GHz.

22c Broadwells cap out at 2.2 GHz in a 145 W envelope.

>>57364985
>I don't know what I'm talking about
Thats cute, child.
Zen is a much smaller core than Broadwell, less transistors, nowhere near as much FPU logic.
Zen has the same clock scaling as Excavator.

>Falling for AMD hype after they've lied about literally every other product they've released in the last 10 years.

>>57364995
> 8c @ ~3GHz Summit Ridge has 95W TDP
> 32c @ 2.4GHz Naples will fit in 165W

holy fuck, please try to think about how retarded what you're saying is, before you try digging any deeper.

If the 32c/165W part is clocked any higher than 1.6 GHz, I will hunt you down in the announcement thread to beg your forgiveness.
> protip: this won't actually be happening

>>57365063
>further proving you have literally no idea what you're talking about
Zen has the same clock scaling as Excavator.
Excavator's clock scaling is a known factor.

The fact that you can't figure this out in your 70 IQ subhuman brain proves you're too stupid to take seriously.

>>57365022
>being a stupid goy who buys Israel Inside with literal hardware backdoors

>>57364823
You don't even know the pricing, how can you say that it's bad just because it doesn't perform better than XYZ?

If its performance is high-tier already, and the price is very competitive, then it's pretty fucking good, specially for the market.

I rather pay 200$ less for 90% performance than a premium for a greedy company that does almost no improvements over its iterations while jacking the price incredibly high, monopoly or not, it's still a dick move and not necessary.

>>57359676
>no information on the hardware used

>>57365695
You don't need to know the performance or pricing. Just look at the history of AMD. They haven't put out anything that competes in a long time. They're a company that only exists so that Intel won't get shit on as hard for being a monopoly.

If AMD wants people to give them serious consideration when picking parts for their PC again, they need to put out something that competes in performance with the newest i5 and put it on the shelf for 80% of the price.

The put it on the shelf part is the most important. You can't buy anything but shit from AMD right now.

>>57365989
If I look at the history of AMD then I can easily consider that it will be a very good gen because John Fucking Keller was in charge for this CPU line.

>>57359256
Thats almost certain, it simply depends on the chip you choose and the cooling solution.

The "base" 8 core is probably the best one for that form factor though.

File: certified shit wrecker.jpg (13KB, 381x286px) Image search: [Google]

13KB, 381x286px

>>57366003
Jim*

>>57366076
Shit, Jim right.
Have a coworker that is called John, and since I'm sleepy I swapped shit up.

>>57366003
The best CPU in the world doesn't mean shit if they don't put it on the shelf. I don't follow this too close but I could have sworn Zen was supposed to be out by now. Keller isn't even still on their pay roll anymore right? If you have a game changing product you get it out fast. If you don't it doesn't really matter I guess. As long as you get something out often enough to keep Intel out of hot water. That's where their real money comes from.

I really hope they knock it out of the park with Zen. It will be nice to have intel not being to charge whatever the fuck they want for 5% increase in IPC and another socket. Signs are pointing to same old shit. Christmas is coming up right?

>>57366131
I'm not delusional, just hopeful that it will at least compete with enthusiast grade hardware, does not have to be the best for me.

>>57359256
I'd love a micro-STX board with an APU on it. would be a pretty swell HTPC/TV gaming/LAN box

Zen will hit barely above 3 GHz with Haswell tier IPC.

Summit Ridge is a great cheaper alternative to Broadwell-E and Skylake-E CPUs but in games Summit Ridge is slower than a Core i3-6100 or Pentium G4620.

>>57366249
A0 engineering samples already at 3.15ghz base clock with higher turbos.
No one buys an 8core/16t CPU for gaming aside from retards anyway. These are for content creation and editing.

>>57366249
Won't it be able to shut down cores to effectively operate in dual core mode for older games?

>>57359103
> 32 core Naples matches the performance of 10 core Xeon
Intel
B T F O
T
F
O

>>57366311
You do realize that Blender does not use all cores, right?

>>57366335
First of all yes it does, secondly it was consumer Summit Ridge being tested.
Don't respond to shitposting.

>>57366343
Thirdly you don't actually know anything, not the number of cores, nor the clockspeed on this blender run.

>>57366311
The Zen sample they tested was an 8 core.

>>57359455
What is that? I want one

>>57366366
No it was a 32 core that lost to a Intel 10 core AMD is such fucking garbage

>>57366375
Oh shit it said right there on the board

All I saw were two ethernet ports and I was blinded by them

>>57359455
>>57366375
>>57366390

fug its 300 euros

I've gotten even new server mainboards for cheaper

>>57366381
Nice source, too bad it smells like shit.

>>57366403
buttsalted AMD drone

>>57359676
lol I see mine cpu there, do you anons?

>>57366412
>implying
Never used AMD my whole life, you're just retarded and I called you out on it.

>>57366417
Oh so underage and a MAD shill

>>57366418
Nice glass roof there, grow up first before trying to argue about anything, shitposting seems to be your only forte.

Go on reply, you win your last word, not gonna bother with literal manchilds.

>>57366429
Glad you've seen the truth of AMD's incompetence, geez a 32 core barely beating a 10 core, how fucking pathetic lol

>>57366366
No, it was a dual core cpu. Jim told me.

intel btfo

File: 1476707969236.png (203KB, 660x498px) Image search: [Google]

203KB, 660x498px

>>57359103
>literally 1 hour after my xeon arrives
i'm returning it

>>57359103
>blender
I want a Dolphin benchmark.

>>57362034
I'm seriously hoping this will get adopted

File: 1310429820535.jpg (13KB, 300x300px) Image search: [Google]

13KB, 300x300px

>hyping another AMD product before it comes out
Nope, I've learned from the past

This looks like a poorfag thread.

>>57365223
Why do we keep this meme going when AMD is also Israel based now?

Deja vu the thread.
I remember the same excitement and bullshit every time AMD was about to launch a new chip in the past 10 years.
Yet every time it was a fucking big disappointment.
The last good AMD chips where socket 939 stuff (I still have a FX-55).

>>57366591
This. AMD hasn't been relevant since c2d.

>>57366941
It will be.

>>57359103
so they have begun, the cherry picked benchmarks, the bullshit leaks, the manufactured hype and unrealistic expectations.

it seems AMD is both the victim of its past failures and historical successes. we expect a K8 tier miracle in an industry where that's no longer a possibility.

>>57366468
Don't be this retarded

>>57366941
K8 was better than Core, on par with Nehalem, and wasn't rendered irrelevant until Sandy Bridge.

>>57366187
This
If the Polaris gpu in zen apu's ends up being pretty fast, we could have console-beating or matching performance in what is basically a NUC with no discrete GPU, not to mention the increased functionality over a console

>>57362108
Hello 2009 my old friend i7 here cuck you again

>>57368643
Polaris isn't in Zen, it's Vega

>>57368641
K8 was roasted by Nehalem, what are you even talking about?

>>57368641
>K8 was better than Core,
Are you talking about k10.5?

>>57368666
You mean Zen APUs will have Vega GPUs? Damn, that's even more promising.

>>57368666
>>57368700
Maybe this will help laptops from not sucking anymore

>>57368670
It depends on what applications you were targeting.
NUMA performance on Nehalem was pretty broken, so you had a lot of IPC messaging heavy type HPC stuff that was no-go.

But single threaded stuff Nehalem was consistently stronger.

>>57368643
>>57368700
Is Raven Ridge confirmed to be Vega (and HBM?) and not Polaris?

>>57368730
AMD/OEM relations are hilarious.

If there's no HBM on these Zen/Vega APUs then the boards will be single channel with 2133MHz DDR4 and performance will fucking suck.

If AMD does the smart thing and takes things into their own hands, puts a HBM module or two on the APU, then the memory speed or numbers o DDR channels is irrelevant for the GPU, the 256/512GBps bandwidth from HBM will be more than enough to feed a 800+ shader VEGA GPU

>>57368771
How much VRAM do you think they could fit onto that package? Surely it wouldn't be a huge amount.
Besides, don't most not-trash (read: user-replaceable, non-soldered RAM) laptops support dual channel memory?

>>57368853
They could fit in 16GB if they wanted, but they don't need more than 2/4GB for an APU.
Obviously most laptops have 2 SO-DIMM slots if one isn't soldered, however not that much people actually bother filling in the extra slot and the laptop is left with single channel setup and believe the slow speed is because the chip is bad.
This won't be a problem if HBM is there, if not AMD will have to strongarm the OEMs.

And as a personal laptop preference, please no broadcom/realtek NICs

>>57368771
I can't see AMD using 2x HBM2 for an APU, unless they dropped DDR support entirely for straight up 8/16 GB system-on-package processors.

The Polaris 10 and rumored Vega 11 get by with ~250 GB/s for 100+W of shaders already.

>>57368905
I don't think HBM for system memory is really usable on a general purpose OS like Windows or even Linux, if AMD includes HBM on package/interposer you can be sure it will be exclusively used by the GPU, I also can't imagine it being used as a pseudo L4 for the CPU like Intel's Iris Pro

>>57364283
raytracing is one of the few applications that scales 1:1 to any amount of cores

Even a single hbm1 stack would be 1GB at 128gb/s which is enough for a huge improvement for an apu. Hbm2 bringing 2/4GB stack at the same bandwidth would be more than enough with low clock/binned chips, trouble is how much does the cost increase with an interposer.

>>57369026
AMD can't afford to worry about margins right now, they need a good product even if they have 20% margins, once they get marketshare and fame back they can consider optimizing margins and profit like Intel has been exclusively doing for the last 5 years.

>>57368939
Enough applications benefit from higher bandwidth on small-ish working sets that I can see AMD exposing HBM in the flat address space with some form of explicit NUMA allocation control.

For me the best Raven Ridge design would have a single 8GB HBM2 module and a single DDR4 interface that a lot of people wouldn't even bother populating.

>>57368939
Why can't they use it as a L4? It's got more than enough bandwidth and Intel's eDRAM can't be that much lower latency than HBM.
It probably uses less power than the crystallwell eDRAM as well, I see no actual issues using it

File: performance.jpg (44KB, 1444x285px) Image search: [Google]

44KB, 1444x285px

>>57366941
piledriver was pretty damn good when the price dropped

>>57369049
>AMD can't afford to worry about margins right now, they need a good product even if they have 20% margins, once they get marketshare and fame back they can consider optimizing margins and profit like Intel has been exclusively doing for the last 5 years.

er, no, if they get shit margins then they will have to sell the product either at premium prices, or at a huge loss. And if they don't get a good enough performance then they'll shoot themselves in the foot.

So good yields is absolutely essential. It's probably why the 480 was so conservatively designed; so they can make a shitload of them cheap.

Because I don't know if you remember, but their biggest problem for the past year wasn't the lack of competing products, but that they had NO products at all.

>>57366123
are you gay for John, anon?

>>57359103
pics of Dr. Lisa

she has computerery name, so there is an example of nominative determinism in action

So I'm waiting for the performance of a cpu I can buy for $50 on ebay?

>>57359103
BASED JIM DOES IT AGAIN

>>57369026
>putting HBM in an APU that can't even take advantage of the higher memory bandwidth

>>57369929
shared memory is one of the best use cases for faster ram

>>57369958
There is no point in combining expensive technology with low-rent CPUs. Expensive GPUs and flagship CPUs? Yes, because you goyim will pay any price to get your hands on the latest and greatest.

File: 5968700.jpg (246KB, 1920x1080px) Image search: [Google]

246KB, 1920x1080px

Can anyone confirm if existing water coolers (corsair gtx 110i)will support the zen chipset?

>>57370045
The brackets will be compatible with AM3+
However
>buying Zen
>ever
Why?

>>57359377
it has less cores than the intel cpu its beating you complete mong

>>57370099
Because its better than a intel 10 core

>>57370169
No it's not >>57362167
That Intel 10-core isn't even as good as an i7-6850K, a 6-core CPU, because it's clocked so goddamn low. That means that an 8-core Zen will be behind a comparable 6-core Intel.

>>57369110
>Why can't they use it as a L4?

In order to not suck ass, a cache needs its index/tag table accessible as quickly as possible, which usually requires it be on the CPU die. Crystallwell is already pushing it with a 128 MB L4, so it's not really feasible to have a 16-32 way multi-GB cache with native line sizes. Maybe if things were swapped in/out in page size granularity, or if only 1/16th or 1/32nd of an HMB module could be pinned as L4 cache for a secondary memory interface.

Beyond that, GPU memory controllers are generally tied pretty tightly to L2 cache slices and ROPs and are more concerned with coalescing writes to conserve bandwidth than anything else, since SIMT scheduling for shaders explicitly disregards latency as the biggest priority.

It would be pretty hard to make a memory controller that served both high BW GPU concerns and low latency CPU concerns at the same time, on memory that's just designed for servicing lots of outstanding requests.

>>57370260
>fancy buzzwords no one understands, not even you
Fuck off, AMDrone
Buy an Intel or GTFO my board. Razer4Lyfe

>>57359531
The engineering sample was 8 cores and about equivalent to a 10 core xeon. Read the fucking article.

File: noice.png (392KB, 697x532px) Image search: [Google]

392KB, 697x532px

>>57370413
>equivalent to a 10 core xeon
Yeah, a 10-core from a 4-year old 22nm architecture and clocked in at just 2.8GHz. And it still loses to a Broadwell-E based 10-core Xeon E5-2630 v4 that's clocked at 2.2Ghz, which only has the same performance as a six-core i7-6850K.

>>57370260
>cache needs its index/tag table

this is a biggie. for 64B cache lines, you need ~1/10th-1/8th the cache size in not just SRAM tables but CAMs (basically parallel XOR address matching).
I'm not sure whether Intel's L4 CAMs are on-CPU or as custom logic on the eDRAM die, but HBM certainly doesn't have the functionality natively, and you couldn't easily add much L4 capacity on the CPUs' memory controllers without bloating shit up real fast.

it would be a fuckload easier to just have explicit NUMA allocation on an APU with both HBM and DDR controllers, and let software folks figure out where they want to shove what.

>>57370480
the comparisons are hand-wavey at best until we learn what the ES clocks are.

> 2.8 GHz = IPC tied
> 3.0 GHz = 9% lower IPC
> 3.2 GHz = 16% lower IPC
> ...

10% lower IPC than Broadwell will still sell amazingly well for the right price.

>>57370551
>10% lower IPC than Broadwell will still sell amazingly well for the right price.
But why would anyone want to buy a shitty eight-core for the same or higher price than a decent six-core from Intel that has more PCIe lanes, twice the number of memory channels, a higher overclocking headroom, better motherboards, etc?

ANOTHER
MAJOR
DISAPPOINTMENT

File: stop shilling.png (726KB, 643x638px) Image search: [Google]

726KB, 643x638px

>>57370592
>higher price
>more PCIe lanes
>twice the number of memory channels
>higher OC headroom
>better motherboards

Because none of these are confirmed; we don't know how Zen will perform in these areas. It might perform better, it might not. It is too early to tell, just as it is too early to extrapolate the performance of the chip from benchmarks that might or might not be fake.

>>57370220
>implying there is any measurable difference
good goyim

>wccftech

I wanna believe but fuck that site after their months of bullshit Rx480 reporting. Wake me up when a good site has the numbers.

>>57370635
>we don't know how Zen will perform in these areas
Every single engineering sample we've seen since last year were clocked UNDER 3.0 GHz. That should tell you something.
Also, the specifications for the AM4 X370 chipset boards have already been leaked... and they are disappointing compared to Intel's X99.
It will only support up to 32 lanes of PCIe (and this is including the PCIe lanes to the CPU itself, not the southbridge chipset), can only support dual channel memory (this was shown in the die shots of the Summit Ridge chips themselves), and X370 motherboard pricing has already shown that they're below comparable X99 boards and more similar in pricing to Z170 boards from the same manufacturer (ASUS).

>>57370592
> Broadwell-Es not being complete fucking ripoffs
> b-but my 6-core i7-6800K is only $450

tip top fucking cake.
Summit Ridge will murder Broadwell-E sales if they expose the 10GbE MACs.
This is as much a play on Xeon-Ds as anything.

>>57359455
That's so fucking cute, I love it.

>>57370710
>if they expose the 10GbE MACs
Consumer chips will not have this. If they did, then the leaked ASUS X370 line-up should have had at least ONE 10Gbps ethernet or SFP+ port, but not even the flagship $300+ board had it.
Stop pulling shit out of your ass. Summit Ridge will not have integrated 10Gb NICs.

>>57362716
What do you mean half the DDR channels? Doesn't an 8 core Zen have 4 channels?

>>57370905
Nope, only two and that's one per 4-core cluster

>>57370959
That doesn't make any sense. Their consumer chips are all gonna have single channel memory.

>>57359103
>wccftech

>>57371016
No, Summit Ridge (= 1 8-core Zeppelin die) is the smallest pure Zen chip they'll sell.

The Raven Ridge APU will be 4 Zen cores, plus GPU bits, plus an HBM2 interface and module.
Even if RR only has 1 DDR4 interface controller, it's not like most software on budget tier APU platforms is bandwidth limited anyway.

>>57371016
The 8c chips dont have their 2 memory channels right next to each other, but on opposite sides of the die. Each core cluster ends up having 1 memory channel right next to them, and one further away. How this affects performance remains to be seen.

The APUs however should have a more standard dual channel controller layout.

>>57359103
Stupid question, but can I upgrade directly from a FX-8320 to this?

File: Zen_Summit_Ridge_First.jpg (79KB, 369x676px) Image search: [Google]

79KB, 369x676px

>>57371016
Summit Ridge has two memory channel interfaces on opposite sides of the die and has been confirmed to support a single channel each
Raven Ridge will combine the two into a single interface to save precious die space

>>57371379
Nope, you will need a new board and ram, since its an entirely new platform. And, lets face it, AM3+ is really fucking old and outdated.

>>57360110
Its not really an if zen is good at this point, we know it is, goods baseline was at minimum sandy bridge performance, and its beating it by a fair margin.

the only unknown now is price, and amd has a range from between 150$ to 1500$ to play with if it is matching a 10 core intel.

>>57366131
The earliest we could have seen it was q4, this was based on when amd got first tap outs and assuming they would do 2-3 which are a 6 week process each time assuming they don't change anything.

q1 17 is when we expected the latest to be based on a very early this year investor thing where amd said there wouldn't likely be zen this fiscal year.

still on track, just not as soon as we would have liked.

>>57370707
x370 is equivalent to z170

anything like x99 and you'll probably have to go with opterons

>>57366285
Im getting one partially for gaming, but that's mostly because i cant afford more than 1 pc for my hobbies. would be nice to be able to render and game by allocating cores to each task, also, the proven benefit from 8 cores over 4/6 is out there, just not benchmarkable, that head room for a background application to demand 100% of a core and it not impact gameplay.

It causes intelfag to butthurt.

>>57366291
unless the game is absolute shit in its coding, 3ghz at haswell will push anything to 144fps, anything more would be wasted resources.

>>57371615
>we know it is,
That's what the rumors say BUT we really don't know

>>57366941
let me correct that, sense sandybridge. remember the first i7s were kind of shit and just re introduced a non shit threading solution.

>>57371945
we know that clock for clock, it matched intel broadwell e in blender, an intel recommended benchmark programs.

from this we can assume that it either isn't currently clocking high, or it isn't able to clock high as stock i7's because gaming could also show off its better then intel too.

for me this will be a workstation cpu that can also game and hopefully not break my bank.

I can't wait to be disappointed again! :^)

>>57363424
This picture is now mine.
It's show time

>>57372044
It will make you a believer anon :-)!

>>57372022
I wonder where the next intel architecture is, if they just rest on shrinking skylake it's gonna get ugly for them

>>57362034
Does it support PCIe.

>>57363811
>comparing clocks between completely different architectures

>>57371016

summit ridge is HEDT only and 4 core variants will probably include 2 cores and 2mb cache cut per CCX rather than cutting a full CCX.

>>57368418
This isn't a miracle, is just getting to parity to a company that hasn't really improved their products in 4 years

>>57363883
how fucking reatrded are you, enjoy getting fucked in the ass by every new nvidia driver

What's even the point of a 4 core Zen? It's the HEDT line so a 6 core should be the weakest configuration.

Wouldn't it be awesome if Summit Ridge has 2P consumer motherboards? 32 fucking threads, I can imagine running VM's for my entire house and getting rid of individual machines.

>>57374164
Probably to have a high-end quadcore that isn't an APU.

>>57366131
jim keller designs and pushes architectures, he doesn't sit and wait for their release. he wasn't needed after he did everything. and he doesn't want to be there. He's pretty much the best EE that exists on the planet for the past decade and he consults.

>>57374164
to have an 8 threaded cpu?

>>57359103
AMD always manages to tie Intel when it comes to heavily multi threaded applications.

They just fall horribly short when it comes to single core preformance

>>57371383
if anything RR will drop the AES-128 encryption units on the memory controllers.

>>57366131
>I don't follow this too close but I could have sworn Zen was supposed to be out by now.
Summit ridge was originally scheduled for October I think but was pushed back to February

If you release a cpu too early it can have terrible bugs
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=738575

>>57374235
Your 10G setup of switches/NICs would cost you more than those 2 CPUs and a $300 motherboard.

>>57372946
price is everything, from the look of it the kabby lake stuff is on average 1% more then skylake, but i would bet money one instruction set is 10-40% faster then an older one so they will use that to market it.

but if amd is clock for clock = to intel, then it can demand up to intel prices, meaning that amd's 8 core could demand up to 1100$

for me, if it costs more then 500$ and is not matching the 8 core, ill pass, if the 4 core apu costs more then 200$, its a no go and ill get an older intel xeon or if possible some moron selling a 6/8 core haswell as they all clock to 4.2

>>57373948
equal clocks and equal core counts is how we find out if chips get more or less down per cycle...

>>57374235
or just 32 threads of render capabilities from 2 likely relatively cheap cpus. would be great for a workstation/personal computer so i can render shit out on 12 cores and use the other 4 for normal tasks and not feel the hit.

>>57374293
and this should bring them at least in line with sandybridge on single thread, but the way the cores work now, i think ipc equal to intel's 8 core means it has similar single thread

>>57374337
no, we were going on tap out dates with the earliest we would see it being october and that would have been a paper launch.

>>57374293
does the calendar say 2005?

>>57359103
Equal performance in what, chip size?

>>57370501
>let software folks figure out where they want to shove what.
That went badly for Intel Itanium

>>57374344
My blood ran cold the day USB surpassed common ethernet in speed, heck 1G even bottlenecks fucking MECHANICAL HARD DRIVES nowadays.

Die already you fucking scum.

>>57374462
1tb platter 100+ mb transfer is glorious

File: 1b70c8cd-73d1-4cab-bc44-903e46d0351a.png (105KB, 255x345px) Image search: [Google]

105KB, 255x345px

HE ACTUALLY DID IT, THE ABSOLUTE MADMAN.

>>57359103
Useless info until we know which chip it was.

>>57366941
last revisions of Phenom2 were actually faster than C2D/Qs

>>57359256

Even better if it has m.2 as well

>>57374516
SHIT WRECKER
H
I
T

W
R
E
C
K
E
R

>>57359103

This will be shit like everything AMD makes.

>>57375464
This. AMD hasn't been relevant ever since Core i launched.

>>57375285

phenom 2 was also released 4 years after the core 2 lineup was released.

>>57375491
core quad was shit compared to phenom

>>57375464
>>57375491
That doesn't mean anything

Netburst launched in 2000 and core2 didn't come until 2006

File: watchitanon.png (30KB, 186x160px) Image search: [Google]

30KB, 186x160px

>>57375464

File: _85319913_poop_protest.jpg (43KB, 624x351px) Image search: [Google]

43KB, 624x351px

>The Path to Zen

File: 2010.png (1007KB, 1090x963px) Image search: [Google]

1007KB, 1090x963px

>>57368641
The Phenom II CPUs were garbage.
AMD was forced to sell their 6 core CPUs for a quarter of the cost of Intel's 6 core CPUs because they were too slow to sell at the same price.
http://cpu.userbenchmark.com/Compare/Intel-Core-i7-X-980-vs-AMD-Phenom-II-X6-1090T/m1822vsm417

The only ones who should ever look at AMD CPUs are poorfaggots.

File: 1367029579107.png (38KB, 578x712px) Image search: [Google]

38KB, 578x712px

>>57375760
>6 times the price
>50% better performance

>>57371383
>and has been confirmed to support a single channel each
Stop repeating this lie, you tech illiterate retard.
Its dual channel.

The physical location of the memory PHY has literally nothing to do with how they're connected on the BEOL.

>>57370741
I need 10GigE for good prices here in the house. I've been on 1GigE forever since like 2003 I think. It's been almost two decades FFS. Wireless is cool and all but homes should all be wired with 10GigE or GigE already.

>>57375928
Then be prepared to pay no less than $400 for a 10GbE network. The cheapest used 10 Gig AICs are about $200 each on ebay, most of the cheaper stuff are counterfeits from China. If you want a lower latency fiber network, then the minimum entry fee is about $500 since most of the AICs don't include SFP+ inserts and SFP+ wiring costs a bit more than copper.
On the bright side, a 10 Gig ethernet cable is no longer $20 for a decent length.

>>57375990
That's the point. We should already have cheap 10GigE like GigE but they keep shoving more wireless. Next step is all fiber everywhere.

Both should be advancing and getting dirt cheap. Wireless AP's all over the place even more than now. More bandwidth. Imagine cellular companies are even pushing to retard things as that business goes bye bye too.

>>57376028
I'm upgrading my LAN's backbone to 10Gig fiber, but I'm already way over-budget after buying four dual-port NICs. Don't even get me started on a switch with 10-G SFP+ bridge ports.

File: goodmorning_venom_chen.png (28KB, 585x538px) Image search: [Google]

28KB, 585x538px

>>57359103
the biggest problem with amd are the temps. if they could make a cheap, low-end, COOL cpu, i would go back to amd. but right now they are going full retard. ever since athlons amd temps were fucking insane. i had phenom shut off due temps, it was running 100+ degrees. intel i5 that i bought as replacement year later gives me 50 degrees when building llvm, even though it runs dwarf fortress at about the same speed. i'm just not willing to build a dedicated water cooling just to run amd crap.

there is no point in having 10-core xeon on desktop and there is no point to this new amd thing which probably melts boards.

I'm pretty confident it's the 8 core. We've known for months that it outpaces Broadwell-E in IPC, and AMD has delayed the release until next year. The only worthwhile reason for that would have been to improve clock speeds, and other leaks indicate that's exactly what's happened to the initial 2.8ghz summit ridge engineering sample, which is now significantly faster.

If I were Intel I'd be doing everything I could to thrust Skylake/Kaby Lake E out to the market, ASAP.

>>57376102
>phenom II cutoff temp is 65C
fuckoff lying shill. 100C would literally catch fire.
intel on the otherhand goes up to 80 before force shutoff, been this way for years.

>>57375833
You've never taken an economics course in your life, have you?

Everybody knows that when you're the ONLY company that offers the best product of its kind, you're free to set it at whatever price you want without losing customers. It's called having an inelastic demand.

File: 14780218193820.jpg (200KB, 1000x1000px) Image search: [Google]

200KB, 1000x1000px

>>57376219
>phenom II cutoff temp is 65C
it was running 65C at 0% load, amdfriend.

the point is amd should go back to being the good budget option, instead of releasing those monstrosities.

>dr lisa is a president
SJWMD

>>57376219
While the shut off should be much lower, I've seen plenty of both AMD and Intel chips hit well over 100C without turning off, nor fire. Most of which are mid-range i3 or i5 chips in laptops full of fluff..

>>57359103
x86 really has to fucking die, but I do like AMD better than Intel.

>>57376257
you're fucking retarded shitposters, anime faggotry too go figure.
no one believes your shitposts.

>>57376276
you're talking about athlons and pentiums right?
i fail to see how sturdy chips from the past are relevant to todays with temp throttles.

>>57359103
>PATH TO LOO

File: 1360368220235.jpg (80KB, 720x720px) Image search: [Google]

80KB, 720x720px

>mfw amd kids ITT aren't even old enough and haven't owned Athlon64

>>57375760
The Phenom IIs were as good as the Core 2s, but cheaper and had more modern motherboards (and you had a 6 core option as well). The only problem was that Intel had Nehalem out by that time, so AMD was not competitive on the high end anymore.

>>57376582
phenoms lacked sse4.1 and ssse3

>>57376274
She's actually competant, and very smart.
Asian tiger mom won't accept anything less than a M.D.

File: PQaHaSlSf1k.jpg (30KB, 549x280px) Image search: [Google]

30KB, 549x280px

>>57376425
>lies! amd strong!

File: 1469043199505.jpg (85KB, 595x654px) Image search: [Google]

85KB, 595x654px

GET REKT <------ lmao that was a dank meme i r89 r9001 390/3.5gb of v for vendetta ram, novidiots be like do a 1080 and walk awayy lmao, 144 wayy sli novideo grillforce gtx 480 clocked at 3.5hz 1080p 3.5 fps gimpworks edition, human eye can see only 3/5 ram anyways, water boiler titanal xp at 3.5p, shitan rascal 1080shiTi at 87.5% GTX(GetTheXtinguisher) (pisscal xp malware experience thermi burning sli bridges edition) + krappy fake shintel gore iSemen end of life cooler not included in a box edition. 8GiB? What do you need 4GiB for? Im not giving you 3.5GiB!

>>57376599
>phenoms lacked sse4.1 and ssse3

Two things that did not matter at all.

>>57376582
>The Phenom IIs were as good as the Core 2s
Horse shit

i had a 955BE for years and it couldnt even keep up with the i5 in my fucking laptop

File: 1471434695013.jpg (55KB, 849x656px) Image search: [Google]

55KB, 849x656px

>>57359103
Zen is gonna be good, boys!

File: s6 Efficiency.png (236KB, 2000x1150px) Image search: [Google]

236KB, 2000x1150px

>>57376734
That goes without saying.
An average 40% uplift over Excavator is pretty decent performance.

>>57376723
That was probably cause the i5 was two full generations ahead of the phenom 2 at minimum.

>>57359103
Just like the Fury X is as fast as a GTX 1080, right? Oh wait!

>>57376516

Fucking funny isn't it? If someone wanted to build a high end gaming system back anywhere between 1999 and 2005, they went with Athlon if they had half a brain. The original Athy, the t-bird core, the XP Barton core, and the Athlon 64 FX were monsters. Always at the top of pretty much every meaningful benchmark.

That said, I expect Zen to be in the top three, but it's unlikely to completely dominate. All it needs to do is compete, and it will be a commercial success.

>>57377110
It was faster

but not faster than the ti

>>57377149
It's literally slower than the 1070 by up to 50% (usually 20%).

Did anyone actually READ this article? It sounds like a fucking WILD FIRE of failure. They're comparing to THREE YEAR OLD CHIPS...? What the ever loving fuck is AMD doing?

This is a joke right? This is not the Zen I've been waiting for, fucking hell. Fine, Intel, nvidia, you win. I'm done with this shit AMD has no idea how to be competitive.

>>57371951
I didnt know that good threading, the second to last 15% perf increase across the board in 8 years, and being able to stably clock above 4Ghz made Nehalem a shit design.

My hex-core x5650 sitting at 4.2Ghz weeps.

>>57377149
>AMDrones believe this shit
Right up there with your zen being compared to a 4th gen i7 when 7th gen is around the corner. Holy shit this is sad for AMD and for all of us. Intel prices gonna go up after this flop.

>>57377159

Being slower or faster than the 1070 probably depends on resolution and the application, but as a general rule, yes.

One example of what I'm referring to is 4k performance in Infinite Warfare.

File: untitled-1.png (40KB, 679x985px) Image search: [Google]

40KB, 679x985px

>>57377159
wtf
even a puny rx 470 is faster than a gtx 970...

File: 1080.png (24KB, 500x490px) Image search: [Google]

24KB, 500x490px

>>57377250

>>57377183
Are you trolling?

There are three year old chips there but they best everything else including all of Intel's other chips. The fastest chip there costs $3000+ and the lowest is around $1700

>>57377183

I read the article, but avoided the comments because they're colossally fucking retarded, just like most of /g/, honestly.

File: Battlefield-1-1080p.png (638KB, 600x600px) Image search: [Google]

638KB, 600x600px

>>57377261

>>57377250
Some games show atypical performance, such as DX and D44M (on vulkan only).

>>57377261

Does anyone still game at 1080? I feel so sorry for them. I hope their food stamps don't run out.

>>57377263
>they best everything else including all of Intel's other chips
In HEAVILY THREADED WORKLOADS. Seriously, a fucking Sandy Bridge 10 core clocked at 2.8GHz? That's pathetic in terms of IPC.

>>57377297
Fury X only has an edge in 4K, though. Even at 1440p (the only resolution that matters), it performs like crap compared to its paper specs.

File: 1472634307854.png (43KB, 682x909px) Image search: [Google]

43KB, 682x909px

>>57377277
Neat

>>57377310

>tfw I bought a fury x for the same price as 1070 after seeing how it might still have potential after vulkan praise on doom.

realize it has 4gb.

am I fucking stupid? Should I get rid of it now and wait for vega?

>>57377440
Just get a 1080 or wait for the Ti, AMD is utterly incompetent and Vega will be another Fury-tier disappointment.

>>57377183

>They're comparing to THREE YEAR OLD CHIPS

And how far have we come in 3 years?

Here's a 5 year old 6 core compared to a new one.
http://cpu.userbenchmark.com/Compare/Intel-Core-i7-3960X-vs-Intel-Core-i7-6800K/1973vs3607
15% difference in 5 fucking years on the same core count.
CPU tech has stagnated so hard that making a comparison to few years old hardware is totally fine.
You're looking at a very marginal improvement over the last few years.

>>57377440
Depends how long ago you got it, what framerate and resolution you're looking at, how much you can get back by returning it/selling it, etc. Give more details.

Unfortunately, I tend to agree with >>57377492 even though I was hype as fuck for the rx line. I don't trust AMD anymore. Even though I realize a big part of the problem is nvidia jewry, when it comes down to it, it's performance + bang/$ vs anti-jewry, and I don't feel like dropping thousands on a GPU half as good as the competitor's.

>>57376710
didn't matter to GAYMEN kids like you

File: Capture.png (190KB, 572x447px) Image search: [Google]

190KB, 572x447px

>>57377309
Okay so run the benchmark on something better than those chips and put it up there so we can get a comparison of GOOD IPC.

There's other chips on the list also.

>>57377534

I bought it less than 3 months ago.

I play ultrawide 1440p

can't return it.

>>57377668
You should still get good framerate for a few years, but at 1440p, the Fury X is worse than even lower-end nvidia cards (the situation is reversed at 4k). You could try selling it now and getting a cheap card in the meantime, or waiting for the next gen, if you can get a good price on it, but if you're tight on money you might as well keep it as long as it's usable.

>>57359103
so they need 32 cores to equal 10 intel cores.

>>57377802
The test result is from an 8 core.

>>57376723
no core 2 has ever been branded i5 kiddo

>>57377802
Nothing there indicates that it has 32 cores. The best guess we have is that it's an 8 core chip because the other chips around it are 6-10 cores

>>57376257
Anon, you're really truly baka most of phenom II model run about 40-45C even with stock cooler.
Also do you not know about ACPI, CPU power state, CPUfreq? There's main difference between intel CPU there's power state mostly broken or even just lack of of CPUfreq support especially on *nix on AMD you can setting up for dynamically CPU frequency/power state changing and not have massive power drain and overheating like on intel CPU.

>>57377668

If it's not able to play at 60fps in recent games, there's something very wrong. However based on nearly every benchmark I've seen recently, it's able to perform well above that. You're fine, stop being such a drama queen.

>>57359103
Sorry, but I've been out of the loop for a good while from /g/ and tech in general

Amd poorfag cucks are still as pathetic as 3 years ago yes?

>>57379098
>Sorry, but I've been out of the loop for a good while from /g/ and tech in general

How was prison?

>>57376723
OCed phenoms are beasts.
on par with core 2 desu, and at the price, you'd have to be an idiot to give intel your money

>>57359103
yeah but fuck blender

>>57379152

>fuck blender
Say goodbye to your dick.

Who even does cpu rendering on Blender though? It's all about opencl/cuda these days

>>57379131
I was in a mental asylum, and I gotta say it was a step-up from /g/ in terms of entertainment.

>>57379165
Hi 2010

>>57379193
U wot?

File: fff.jpg (155KB, 546x564px) Image search: [Google]

155KB, 546x564px

>It's another "AMDrones massively overhype a massive dissappointment" episode .

Why is /g/ so obsessed with how fast proprietary video games run on CPUs?

>>57379397
You could say AMD stands for
"Another Massive Dissappointment"

>>57363883
"Im glad I am a mindless brand loyalist"

couldn't have put it better myself

Don't AMDfags ever realize nothing good will happen?

>>57377199
remember, everything that wasn't an i7 was a step behind stock phenom II's and even then, it took till the 920 to match even modest p2's or even core 2 quads, where it excelled was if something took advantage of the threading, but at this point, nothing outside of video rendering or ed modeling really did.

I never really look at overclocking as its largely silicon lottery, little brother has a 4XXX i7 that cant make it passed 3.9 stable and a 290X that is in the top 1% for overclockability.

but put it this way, the 2600k damn near doubled the performance of the 920 in both single core and multicore applications, and was the last massive fuckoff upgrade in cpus, its now relegated to minor tweaks in the architecture yielding 1-5% with a new instruction set that gets that 10-40% higher performance over an older one

Amd should bridge the gap to the current gen i7, but I would be perfectly happy with a 8 core at sandy bridge level performance, because intel has offered up next to nothing sense then.

>>57376723
remember, intel sandy bridge doubled performance on single core and multicore, at least from the 920 to the 2600

NEVER FORGET how massive a fucking leap sandybridge was.

>>57379147
phenoms were better then everything intel put out sans extreme editions. but even then, not that bad.

>>57375760
first i7's were about equal to phenom 2's single core, and pushed ahead multicore due to threading.

realistically at this point cores mattered as next to nothing was doing the threading or handling it well outside of benchmarks, and in core for core, amd was about 15% behind intel for a fraction of the price. seeing things like this give me lots of hope for non shitty zen prices even if they are close to intels 1000$ lineup.

>>57359103

All news is good

competition back and no intel split up
GOOD
GOOOOOOOOOOOOOOD
*rub's hands*

>>57376257
>>57376102

lets see my 6 year old thermal paste on amd stock 955 is running at... 34c with a constant 20-30% load, and I have gotten this thing down to I want to say 25 idle after a fresh clean, likely could get that current temp down to 29-31 if I changed the paste, but i'm afraid the paste is welded the heat sync to the cpu and can't risk damaging it till zen comes along.

you either fucked up the thermal paste, had a lemon, or are a lying piece of shit. take your pick.

>>57377060
during an investor meeting, eirly this year, amd said 'its surpassing expectations'

what that means we don't know, but we do know that a 40% uptick would not put it clock for clock matching broadwell E,

>>57377219
>>57377159
>>57377149
>>57377110
depending on game by game, its either just below a 980ti, or its above a 1080, real marvel at how much the api matters.

>>57379152
>>57379163
blender is a recommended benchmarking tool from intel, its why amd is using it, beating intel at their own game.

>>57380363
The statement was, regarding datacenter parts, that they were on track for a greater than 40% uplift vs the previous generation of chips.

This statement is consistent with earlier stated figures because they were comparing it against Piledriver here instead of Excavator.
It is still a 40% uplift over Excavator. This hasn't changed. AMD never said a single word that revised this.


40% is the average, and Blender is clearly a best case scenario for the arch.

>>57380479
I'm thinking that as well that its a best case scenario but it still shows well above a 40% uptick at least, depending on the application. still need more benches and non amd hosted events, but im at least optimistic that a higher than sandybridge ipc is coming, and hoping the true 8 core will at least as cheap as their i7's but prepared for up to 500$

While i hope Zen will be good, to succeed what it really needs is OEM support if they want to compete at the consumer level.
Maybe they could convince HP to drop Intel for an all AMD lineup, i know you guys don't care for them but they've probably been one of the most Pro AMD companies for a while.

Though it'd be nice if they got them into the hands of system builders quickly.

>>57380775
likely going to be full amd in apple once the laptop zen comes out.

also cant overlook heavily binned amd parts as they are stupidly fucking efficient.

>>57380371
How much were you paid to make this post?

>>57382007
ever look at benchmarks for games not on dx11? pm 11 it trades blows with a 980ti, on 12 or vulcan its bettwen a 1070 and 1080 with one where it is over a 1080

rams going to fuck it hard in a year or so, but right now, depending on price, best value you can get, next to a fury... seriously, some of them go for 250$ new

>>57382028
Another Massive Disappointment could stand for something else if they didn't pay for low-tier shills like you and focused on making non-shit products.

>>57382059
so i take it you dont... ok, be retarded

>>57376274
Shes the only worthwhile woman that is a ceo of a silicon valley company, so shes actually accomplished like useless people like that retarded cunt that ran HP into the ground and tried to run for president of the US

>>57377261
>call of duty

>>57379184
They send people with severe autism to mental asylums now?

OK everyone, lets play a game called spot the brand cocksucker.
>>57360175
nope
>>57359599
still nope
>>57359341
keep trying
>>57363883
DING DING DING WE HAVE A WINNER

NEW ZEN THREAD
>>57382556
>>57382556
>>57382556