[Boards: 3 / a / aco / adv / an / asp / b / biz / c / cgl / ck / cm / co / d / diy / e / fa / fit / g / gd / gif / h / hc / his / hm / hr / i / ic / int / jp / k / lgbt / lit / m / mlp / mu / n / news / o / out / p / po / pol / qa / qst / r / r9k / s / s4s / sci / soc / sp / t / tg / toy / trash / trv / tv / u / v / vg / vip /vp / vr / w / wg / wsg / wsr / x / y ] [Search | Home]
4Archive logo
So, how do you /g/entoomen think Zen will be?
If images are not shown try to refresh the page. If you like this website, please disable any AdBlock software!

You are currently reading a thread in /g/ - Technology

Thread replies: 193
Thread images: 29
So, how do you /g/entoomen think Zen will be?
>>
I'll be better than Excavator.
>>
I just bought a lenovo y50-70 laptop so I really can't afford anything in the next 6 months, but I wish it's good for amd because we need a healthier competition on x86
>>
>>52860729
how the fuck to we know SMT hasn't been in the picture for AMD for years but at least it will put things back into being competitive again bring back the GHZ wars
>>
They're all going to shit, Intel has the ME spyware chip and backdoored rdrand, and and is now implementing a similar thing with its PSP chips and rdrand implementation.

Welcome to a world where the botnet is built right into the silicon, is undetectable, has direct access to memory and network, and is powered by something as powerful as an arm chip.

Welcome to a place where your CPU will perform heuristic matches to dissident material in memory and report you for free thinking, where you can't even trust crypto because it already has your keys and the algorithms are backdoored anyways

Lookup:
Bullrun NSA program, Intel ME, RDRand, dual_ec_drbg
>>
>>52860865
That's why I'm hopeful for Zen.
>>
>>52860865
An ARM TrustZone processor is not a backdoor, it is literally the exact opposite. It is a more robust method of implementing trusted execution.

Spreading absolutely ignorant babble like this is a bannable offense outside of /b/
>>
>>52860878
Zen is AMD's full migration into the botnet. You can expect the NSA has paid millions for no questions asked access to hardware designs

>>52860891
It runs unauditable code and runs on top of the iommu, giving it full access to any part of the system. Do you really want to give root the NSA so simply?
>>
>>52860911
B-but anon, how will I use ddr4 without botnet?
>>
>>52860930
Hopefully pure arm chips that don't take microcode will be mainstream soon, or maybe power7 will save us all.

Who knows. Libreboot keeps a list of systems that will roughly allow you to avoid much of the botnet but they're all old.
>>
>>52860911
>even more shitposting from a tech illiterate retarded child
>e-everything I don't understand is a backdoor!

Hilarious, you underage spastic. It is functionally no different from KNOX security, a system so robust that the government itself approve its use for their agents. It is a bolster to security, not a hole. TrustZone coprocessors have been in use in devices for years, and there has never been a single issue raised with them by any privacy advocacy group. Its only you tech illiterate child shitposters on /g/ who make these outlandish claims.
>>
>>52860951
Is my i7-4950HQ safe?
>>
>>52860967
Anything that has an Intel ME chip has a chip with mystery code running at ring 0 with DMA and network access, so no

>>52860965
The 'government' also approved DES for its secrets after backdooring it with s-boxes.
>>
>>52860988
What if I flash a new BIOS and install Gentoo?
>>
>>52860988
Nice non sequitur, kid.
Go shit up a pokermon thread and stop spreading FUD about a totally unreleased core arch.
>>
>>52860993
Microcode in the bios is signed and opaque. Flashing a libre bios will not work as the CPU will not work without an Intel signed official blob giving it instructions. Coreboot is libre but still includes these blobs because there is no way to get around them.

Some systems even detect when you've tried to disable the ME and shutoff after 30 minutes automatically
>>
>>52861012
I don't really believe you senpai desu.
>>
>>52860988
>ring -1 on arm blocks all ring 0 demands on command and access its own memeory

ring-2 makes ring -1 hidden from OS and user interface and only accessed though dev memory PCI-E chip

that face when
>>
HEY GUYS.
BULLDOZER IS GOING TO BTFO INTEL THIS YEAR.
FINALLY DECENT PERFORMANCE FROM AMD.
LEAKED BENCHMARKS SHOW IT SHITTING ALL OVER INTEL'S 2500K.
I'M SO EXCITED JUST LOOK AT THESE HYPE THREADS.
>>
>yfw some faggot steals signature made for syrillian and slaps it on as the amd zen logo
>>
>>52861050
>>52861055
welcome back to 2011
>>
>>52861050
in multi core compute bullozer can lel
and it still can.

single care performance is going the way of the dinosaurs
lel.

VR will make intel burn.
lel
>>
>>52860865
Holly crap, I thought this post was a paranoid /g/ rant, but this is real

remote admin for a PC using a flakey web interface even when the power is off, wtf is intel thinking
>>
>>52861083
I've seen benchmarks that show Intel beating AMD in multicore performance. Not only that, the benchmark also showed that Intel WAS FASTER WITH HALF THE NUMBER OF THE FUCKING CORES.
>>
>>52861083
VR is a really interesting case along with 4K in general.
Pushing tons of pixels places such an imbalanced load on the system, the GPU is doing the bulk of the requested work, so it really masks the differences in CPU performance from one processor to another. Granted this will eventually change as GPUs become sufficiently powerful, but thats years away.

>>52861130
Intel's core architecture has massive, ungodly powerful FPUs. AMD's cores never really have, and the FlexFPU used in the Bulldozer family was designed only to be "good enough" while saving space. Anything workload making heavy use of the FPU is going to see intel pull way ahead in performance, and unfortunately for AMD lots of misc integer ops are handled by the FPU as well.
The Bulldozer derivatives tend to do fairly well in strict highly threaded integer bound workloads, Apache bench for example, but not man programs on the client side of things favor AMD chips to that degree.
>>
>>52861130
>pass mark
yes

it was 3 boardwell and 1 haswell and 1 skylake
all i7

but on the lower end there nothing.
so intel only cares about multi care performance on i7

wow and i though intel was number 1
On every level its pathic it really shows that intel care more about servers and high end for the cash nothing else.
>>52861323
sorry but amd made those cpu's in 2011 there was nothing compeditive now this is why we need zen.

and not to mention there is only one server CPu with AMD that has any power at all.
it shouldn't even be on this graph.

and AMD haven't built any high end CPu's since the 9590 whitch is old as fuck 2011 tech.

what ever happened to full line ups beating competitors?

think is bullbozer is not too far off only 1028 points behind that should not happen in the high end for a processor that old.

Only the newest stuff is betting the old line of processors that means intel don't give a shit enough for a clear front on the consummer side this is why everyone is saying the cpu market is stagnate.

this is why we need zen
things are slowing down
>>52861323
if you run it on intel xeons you can. but this is consumer tech and games and xeons can't do both all that well.
>>
>>52861548
pic.

just confirming what in said above.
>>
>>52860729
WHEN IS IT COMING OUT REEEE

I want to upgrade from my sandymeme i7, not to another intel though
>>
>>52861640
according to currytech we'll be lucky to see it by the end of the year, which sounds about right daysu sempie.
>>
>Zen logo is a big fat 0

top kek
>>
>>52861548
Your post is an indecipherable mess.
There are no server chips shown in that bench
The Vishera line is not from 2011
The FX Centurion chips are just a different binning of Vishera

Lay off the booze, Maori.
>>
File: fm3-am4 socket.jpg (97 KB, 1024x724) Image search: [iqdb] [SauceNao] [Google]
fm3-am4 socket.jpg
97 KB, 1024x724
>>52861640
Summit Ridge is launching end of the year.
>>
FOCKEN GARBAGE, DOUBLE GUNS COCKSUCKA
>>
File: chippuh.jpg (4 KB, 104x122) Image search: [iqdb] [SauceNao] [Google]
chippuh.jpg
4 KB, 104x122
>>52861914
>>
>>52860965
https://en.wikipedia.org/wiki/Trusted_Computing#Criticism
NSA shill pls go and stay go. What we're discussing is a coprocessor running code that you cannot know, cannot control, and which has access and control over your entire computer.
>>
>>52861926
>more shit posting from a tech illiterate retarded child
You literally didn't even read the article you posted. You just desperately googled around trying to find something relevant, and shit your little childish pants at the first sight of something barely relevant.

The criticisms raised there are on the concept of so called trusted computing itself, not on any specific implementation. Worrying about the potential future use of a technology is not even remotely the same as *proving* a fault in an existing technology. You cannot just claim something is a backdoor. Claims require proof. You have nothing except infantile shitposting.

There is nothing inherently wrong with ARM's TrustZone system, and saying its a NSA backdoor is nothing short of a total fabrication. You're fearmongering and spreading FUD against AMD because you're a little childish shill. Go back to your pokemon threads where you belong, you underage uneducated child.
>>
>>52861978
>criticisms raised there are on the concept of trusted computing itself
>there is nothing inherently wrong with ARM's trustzone system
Do you even listen to yourself? No, silly question, of course you don't.
>>
>this thread
>>zomg i need something that can top out bench marks so i can shitpost.

amd apu here underclocked cause i dont need 4ghz to shitpost.
>>
>>52862084
Keep on shitposting, you tech illiterate retard.
You don't have any ground to stand on whatsoever.
>>
>>52861116
It makes sense if your it. So you can spy on your boss otherwise welcome to the world you have chosen.
>ZEN will beat Intel for like a week
>rigged benchmark by intel
>heavy incentives to buy Intel
>and will not get the market share it deserves
>amd will be kill
>>
>>52860804
You gosh darn better be
>>
I expect it to perform slightly above Ivy Bridge levels in IPC, clock a bit higher stock, run less hot, have the power use of an Haswell and have twice the "core" count of intel parts. Add modern chipset features on the new motherboards and it would be a better pick than an 6600k for the same price. Would atleast force jewtel to bring hyperthreading to i5's and bump the i7's to 6 cores and 12 threads
>>
Around haslel performance. Can't wait to get i7-5960X performance on the cheap.
>>
>>52864124
>mfw amd btfo intel price-wise in that comparison
>>
>>52860951
>maybe power7 will save us all
How the fuck would a CPU architecture that's been outdated for three years save us from something that has superior performance in EVERY respect?
>>
>tfw 3570k
>tfw nothing AMD or intel puts out in the next 10 years will be worth upgrading to
>>
>>52864309
ma nigga
>>
AMD's last chance to redeem themselves a bit.

If Zen is a failure, they can close the gates.

They are already dirt cheap and no other company is showing interest in buying them up.
>>
>>52860729
I expect Zen's performance to fall somewhere between Haswell and Broadwell for a substantially lower price than comparable Intel chips, which is all I care about given that Intel's gains since Sandy Bridge have been utterly abysmal.

Zen being good is the only thing that will move Intel from their 5-year lingering at $300 = 4 [email protected]
6 or 8 core Zen will be >90% as good as 6 or 8 core Skylake/Kaby Lake and cost a lot less.
>>
>>52864403
>>52864335
Yeah this really looks like a sink or swim scenario for them.
>>
>>52864403
>I expect Zen's performance to fall somewhere between Haswell and Broadwell
You're setting yourself up for a massive disappointment.

> for a substantially lower price than comparable Intel chips
Thats just pure delusion.

Top binning 8 core Summit Ridge is absolutely going to have an MSRP over $300, around $500 is likely. Summit Ridge is not a mainstream desktop chip, its aimed at intel's i7E line.
Their APUs are their mainstream desktop parts, and Raven Ridge is the biggest one ever. Even Kaveri had a raised MSRP, and that was a tiny uplift in performance over the refreshed Richland. They're going to be north of $200 on the high end.

AMD isn't going to compete with intel in AVX performance, so they won't be building an FPU around that. That means any tertiary integer instructions will also fall behind in performance. The strong suit of Zen is multicore scaling, and socket to socket scaling for enterprise. They're not going to win out in raw IPC, or FPU performance, but they may have a bit of a clockspeed advantage from being a slightly smaller core.
>>
>>52864403
>between Haswell and Broadwell
wasn't Zen supposed to only be "at least as good as Sandy Bridge"? seems a tad more realistic
>>
>>52864674
Zen is suppose to have 40% better IPC over excavator and 45% better IPC over steamroller.

So according to this benchmark >>52864124 Zen is suppose to be around haswell IPC.
>>
>>52864674
It's SUPPOSED to be about 40% faster than excavator, which would put it on par with Haswell. As of right now, only some engineers working for AMD actually know how fast it'll be.
>>
>>52864529
Not even AMD is delusional enough to think that APUs are enough for AAA games.
>>
>>52864529
>You're setting yourself up for a massive disappointment.
This
Zen will not even compete with Haslel in single threaded performance. Multi-threaded performance is going to get BTFO by Skylake and Kaby Lake i7s because of Intel's more mature SMT design and you bet your ass that AMD is going to price their Zens at the same level as Intels.
A quad-core Zen with SMT is only going to be on par with an Intel i5, and yet cost as much as an i7 because "muh premium upmarket price move"
>>
>>52864751
Guess what?
A Raven Ridge APU will utterly shit all over any Vishera chip when it comes to gaming.
>>
>>52864882
>4 year old chip gets beat by current year APU

Wow it's almost like progress is actually being made, who would have thought.
>>
>>52864814
Then only fanboys will buy their CPUs.
>>
File: 678696.png (5 KB, 142x142) Image search: [iqdb] [SauceNao] [Google]
678696.png
5 KB, 142x142
It's shit like everything ATI/AMD has ever done since Athlon64.
>>
>>52864403
>Intel's gains since Sandy Bridge have been utterly abysmal.
Because its only x86 competitor still hasn't catched up to that ancient Core design. Intel has been focusing on mobile and server processors.
>>
File: leo.jpg (73 KB, 300x237) Image search: [iqdb] [SauceNao] [Google]
leo.jpg
73 KB, 300x237
>>52864935
Jim Keller designed Zen and Zen+ and left AMD. He designed Anthlon64.
This implies Zen will be kick ass.
>>
>>52864972
Past glory is not a guarantee for future success.
>>
>>52864979
Keller's glory is hardly "past"
Apple's A7 and A8 chips were Keller products.
>>
>>52864996
>reduced to making i-shit
How the mighty have falled
>>
>>52860729
I'm ready for the influx of shills and rectum ravaged Intel fanboys
>>
>>52860729
I think it will be late and underwhelming.

But I'd like to be wrong.
>>
>>52864996
You mean chips they apple then turn away from with a year? Kek
>>
>>52860729
I just want a CPU that lets me use ddr4 ram and win7.
>>
File: YuExZ.jpg (183 KB, 800x1000) Image search: [iqdb] [SauceNao] [Google]
YuExZ.jpg
183 KB, 800x1000
>>
>>52864979
>the best predictor of future performance is past performance
Not a guarantee, but there's no other information to go with
>>
>>52861978
>Thinking anything is safe from Freedomland or it's alphabet agencies...
I remember when I was young and innocent also anon
>>
>>52865128
Just to add to this, I think the tech industry has subverted this line of thinking more than any other. Just off the top of my head, core 2 broke so radically from the usual development curve that you couldn't have anticipated it by just looking at previous products. Likewise, bulldozer also broke from our expectations that we could reasonably put together from previous developments.

Yet again we find ourselves on the brink of another large shift. I remain cautious, mostly because AMD have only been talking about IPC and not raw performance. Are we to expect something with similar IPC to competing Intel products but with a sub 3GHz clock rate? I say this as an AMD fanboy, which is why I'm more hyped for Bristol ridge than Zen. We know what to expect from it and it promises to be a solid step forward, Zen is much more of an unknown.
>>
>>52865347
Why would they clock it so low? Other than power requirements, which should be significantly lower than Bulldozer thanks to the significant reduction in transistor size, there's no reason to nerf the clocks like that.
>>
>>52865394
>>52865347
People here really need to stop speaking of Zen as if its a chip.
Zen is a core architecture.
Summit Ridge and Raven Ridge are the two consumer chips coming out which use Zen cores.

Summit Ridge is an 8 core chip aiming to compete with intel's i7 Extreme line of processors. High core counts and lots of cache means clocks have to be limited to maintain a reasonable TDP. This is perfectly fine for this class of chip because they're targeting nothing but highly threaded workloads like rendering.

Summit Ridge isn't going to be a high clocked part. Raven Ridge will probably have higher serial CPU performance from hitting higher clocks.
>>
>>52865458
While I realise that it's an architecture and not a single chip, it is an important question. We know that the entire range isn't going to be clocked the exact same but wait will stay within a general range for a general set of products. This becomes clear when you look at their current lineups for different segments.
>>52865394
Clocking it below 3GHz may be a bit extreme but just keep in mind, they've only mentioned IPC, not clock rates and definitely not absolute performance. It not clocking high enough can be just as bad as lacklustre IPC, which again is why I'm cautious but not worried. I'll jump aboard the hyper train once we get a few leaked benchmarks, by which time I'll have a compatible motherboard running a Bristol ridge APU.
>>
>>52865458
>summit ridge is going to be a 225W TDP housefire anyway
>less single AND multi-threaded performance than an equivalent Intel i7... from three years ago
AMD's so fucked
>>
>>52865719
>32nm to 14nm in one shift
nah m8
>>
>>52865394
>>52865458
>>52865690
Another thing to keep in mind, sometimes they don't always know what clock speeds they can wring out of the silicon until its they have good sample chips using the final process.

Bulldozer was hyped up as running at very high clocks (5-6GHz) early on and this was supposed to overcome the fact we knew it would have slightly lower IPC compared the Phenom II chips. Unfortunately it fell flat on its face and first gen Bulldozers were clocked practically the same as the Phenom II chips.
>>
File: frodo why.jpg (93 KB, 400x399) Image search: [iqdb] [SauceNao] [Google]
frodo why.jpg
93 KB, 400x399
What would happen if Zen is a flop and amd tanks,
What would happen to the x86-64 license?
Will intel have to break-up due to anti-monopoly policies?
>>
Bristol Ridge has already been leaked and it's awful.
>>
>>52865755
>AMD R&D budget only a fraction of Intel's
>AMD can't even fabricate their own parts
>AMD had to cut corners by copying their own shitty Jaguar processor to turn it into Zen
>muh dieshrink
Zen is going to be a disaster of epic proportions. They don't have the money, the people, and the time to make it a decent chip.

>>52865781
Just as I expected. And you all thought Zen was going to be AMD's savior.
>>
>>52865781
Those are mobile SKUs, retard.
>>
>>52865797
Do you even know what you're talking about?

>Zen is anything like Jaguar
>Jaguar is not based on Bulldozer
>implying
>>
>>52865778
piss factory
>>
>>52865811
Did anyone say something else?
>>
>>52865848
Have you even compared the Zen architecture with Jaguar?
>>
It always seems like AMD has some technology up its sleeve. Looking at their record reveals that it's always disappointing and generates undue hype.
>>
>>52865781
How is that awful? Looks like a reasonable improvement to Carrizo while using the same architecture and process. Base clocks alone are a large improvement.
>>
>>52865903
GCN was pretty impressive, especially in retrospect. Either that, or Kepler, really, REALLY sucked for Nvidia to need an entirely new architecture so soon.
>>
>>52865934
We can already see that 20 extra watt will only give 300 extra MHz base and 100 MHz turbo.
If the desktop chips run at 65W TDP as predicted then they will still be very slow thanks to their shitty IPC.
>>
>>52865959
I've had a 660ti for years and still love it.
>>
>>52865877
Jaguar did the same bullshit Bulldozer did, two ALUs and one FPU form a "module", which they count as two cores. From everything I've read, heard, and seen, Zen is going to be using one ALU and one FPU per core, and they're not doing the module shit anymore.
>>
>>52865903
For the size of the company AMD does an unexpectedly fantastic job developing IP, they just completely fail at executing. They're trying to play a big game without the money and man power to do so. Focusing on a few key segments is the only way they're going to stay alive, and its likely what they'll do in the coming years.

AMD is going to pull an IBM and start focusing entirely on the enterprise market while they prepare for an exodus of the consumer market. It wouldn't be that terrible if you could still buy server boards and Opterons off of newegg or microcenter.

>>52865934
Ignore shitposters.

>>52865985
>3ghz base clock vs 2.1ghz base clock at 35w
>hurrrf only 300mhz boost for 20 moar watts guys!
You're flat out illiterate
>>
http://hardware.slashdot.org/story/16/02/08/0753255/linux-kernel-patch-hints-at-at-32-core-support-for-amd-zen-chips

32 cores
2
c
o
r
e
s
>>
What the fuck is happening with the stocks again, it's going down even more.
>>
File: 1330293660774.gif (2 MB, 391x237) Image search: [iqdb] [SauceNao] [Google]
1330293660774.gif
2 MB, 391x237
>>52860729
2% better than previous architectures
progress is over, moores law is dead
>>
File: core compare.jpg (204 KB, 1600x983) Image search: [iqdb] [SauceNao] [Google]
core compare.jpg
204 KB, 1600x983
>>52866011
>I have literally no idea what I'm talking about: the post
Jaguar cores are not modules, they do not use the FlexFPU, every single core is independent.

Stay in /v/, child.
>>
>>52865848
>>Jaguar is not based on Bulldozer
You're not implying that jaguar and bulldozer are in any way related, are you? Because that would be retarded. Especially since AMD initially designed bulldozer to scale down to jaguar TDPs but came up with jaguar as a replacement when they saw bulldozer couldn't deliver.
>>52865797
Please tell me you're not going by the EXTREMELY simplified block diagrams. While it wouldn't be unexpected (or imprudent) for AMD to reuse IPC in making a new design calling one a copy of the other is laughable.
>>52865781
>Solid incremental update on a chip made to be an incremental update
I don't see anything wrong here.

It'll also be a huge leap over Kaveri, since Carizzo never made it to the desktop market beyond a chopped down Athlon not made for desktop TDPs.
>>
>>52866011
> From everything I've read, heard, and seen, Zen is going to be using one ALU and one FPU per core, and they're not doing the module shit anymore.

Also this is just too hilarious for words. You're a total tech illiterate retard. You don't even understand the words you're trying to use.

Zen is a 10 wide core.
4 ALUs, 2 AGUs, 4 FPU pipelines. This is straight out of the kernel patch.
>>
>>52865985
I'm sorry, what? Are you saying that the architecture being optimised for low TDPs is a sign of it being bad? Even if we ignore the poor power/clock scaling, the IPC improvements, increases in clock rate and vastly improved graphics portion is huge compared to Kaveri, the last desktop APU. Heck, the colour compression ALONE would make a 65 watt Carizzo chip worth it over a 95 watt Kaveri chip but there are many other dissimilarities preventing that from happening, most notably that Carizzo is an SoC while Kaveri is not.
>>
>>52860967
core2 were last safe Intels
>>
>>52866134
over 20 times the Judaism found in Bulldozer for a 15% bump in core performance and a 6 billion percent rise in shekel caching
>>
>>52861024
its true. Google "disable Intel ME" and watch how its literally impossible. Look on wikipedia what ME is. Its literally designed for full control via remote access
>>
>>52866075
>AMD initially designed bulldozer to scale down to jaguar TDPs but came up with jaguar as a replacement when they saw bulldozer couldn't deliver.

No they didn't.
Bulldozer was originally a much larger server specific architecture with a larger FPU and 3 ALUs per int core instead of two. There are still some test chips out there floating around. This architecture never was designed with low power in mind, it was meant to be highest possible integer performance.

Jaguar is a very, very tiny incremental upgrade over Bobcat. Bobcat itself is literally a refined smaller version of K10.


Every single one of you on this board needs to stop talking out of your asses.
>>
AMD is opening up their hardware and Intel is doing the opposite. Skylake and newer require a firmware blob to get working. If Zen chips even come close to matching Intel's performance then they've got a guaranteed sale from me.
>>
>>52866263
I'm trying to find some of those old slides, you know, the ones claiming bulldozer would scale from 10 watts to 150. Bulldozer was built to be scaleable. Unfortunately, this was never realised until carrizo came out with chips capable of scaling down to 12 watt TDPs.
>>
>>52866391
It is literally impossible for one transistor library to scale to that degree. You're imagining bullshit.
>>
>>52866263
>Bobcat itself is literally a refined smaller version of K10.
And you claim others are talking out of their ass? I'll bet you think irony means "sort of like iron" too.
>>
File: bobcat-k10.png (137 KB, 1400x482) Image search: [iqdb] [SauceNao] [Google]
bobcat-k10.png
137 KB, 1400x482
>>52866484
>shit eating child totally unfamiliar with the topic at hand opens his stupid mouth

That is literally the design path that begat AMD's Cat cores. Bobcat is a refined and simplified K10. They spent as little money as possible revamping their old core arch to create an Atom competitor.
>>
File: hmmm.png (11 KB, 470x454) Image search: [iqdb] [SauceNao] [Google]
hmmm.png
11 KB, 470x454
https://www.reddit.com/r/Amd/comments/44hrj7/a_theoretical_calculation_about_of_zens_single/

I know I know >reddit, but its easier to link than pasta.
>>
>>52866558
>Broadwell is 6.45% faster than Haswell
>Skylake is 4.24% faster than Broadwell

This would make Skylake ~11% faster than Haswell which is bullshit.
>>
>>52866528
Bwahahaha, you're going by the block diagrams? Those overly simplified things? You're as bad as the anon who thinks Zen is just a development of jaguar.

Yes, they reuse IPC when building new cores, because all of these old tech giants have a vast library from previous architectures, it's what makes developing something new for them far cheaper than it would be for an outsider. No, that does not mean that one architecture is the same as an architecture it shares features with.
>>
>>52866635

In some workloads it is.
>>
>>52866635

>skylake
>faster than broadwell

A huge slab of L4 cache disagrees.
>>
>>52866558
None of those figures quoted are anywhere near right, not that you could reduce CPU performance to a single figure to make such a comparison.

>>52866640
Look through the Bobcat Hot Chips presentation demonstrating in depth its front end, decoder, register file, and pipeline. I made a 100% factual statement. Bobcat *is* derived from K10, and this a *fact* that AMD engineers themselves spoke about when first detailing the architecture.
They proudly touted it being able to handle 90% of common X86 workloads because of its K10 pedigree.

You are hilariously confusing the term IPC for IP.
IP is intellectual property
IPC is instructions per clock

You not knowing this is too funny for words. You look like a child.
>>
>>52866711
>IPC is instructions per clock
cycle tbqhwyf
>>
File: anon pls.png (704 KB, 781x779) Image search: [iqdb] [SauceNao] [Google]
anon pls.png
704 KB, 781x779
>>52866711

>IPC is instructions per clock
>>
>>52866775
>>52866755
>first year CS drop outs who don't know what a clock cycle is

This board gets better every day
>>
File: iloveheatherfeather.jpg (541 KB, 3000x2614) Image search: [iqdb] [SauceNao] [Google]
iloveheatherfeather.jpg
541 KB, 3000x2614
>>52866711
>IPC is instructions per clock
>>
>>52864979
>>52865036
>>52865074
>Cucked by Intel
>not wanting competition
>>
>>52866798
I know and I didn't say you are wrong like >>52866775 but I have never heard anyone say instructions per clock
just pointed you in the direction of calling it instructions per cycle my senpai
no hate my friend :^)
>>
>>52866263
This 100%
>>
>>52866391
Bullshit from you or AMD's marketers. Bulldozer was designed as a server component marketed by retards as a desktop CPU.
>>
File: CL7LYlQUkAAqeM8.jpg (40 KB, 600x431) Image search: [iqdb] [SauceNao] [Google]
CL7LYlQUkAAqeM8.jpg
40 KB, 600x431
>>52866798
>all this damage control
>>
>>52866899
>he literally doesn't know what a clock cycle is

Infantile shitposting isn't going to make up for how stupid you are
>>
>>52866804
>AMD
>Competing with intel since Sandy Bridge came out
Pick one.
>>
>>52866711
>You are hilariously confusing the term IPC for IP
Actually, that's down to overly aggressive autocorrect but hey, you'd never make a typo, right?
>IPC is instructions per clock
Yeah.
>Bobcat hotchips presentation
Where they do mention that it has things in common with their older architectures, including things been doing since k7 but never saying it's just a development of an earlier architecture. Think you need to go rewatch it.
>>
>>52867074
>blaming spell checking for outing yourself as a dumb shitposting /v/tard
>grasping at straws because you're completely out of your league

Just give up already, you brain damaged spastic.
Trying to argue when you have no technical knowledge whatsoever is a complete waste of time.
>>
>>52866755
It is clock. Also, gtfo with that black fake meme language.
>>
>>52867131
>please oh please tell me already that I won that important internet debate on an anonymous image board please
>>
>>52867131
>Having no arguments
Please, tell us where in the video they outright state that bobcat is a development of a previous architecture, as you have claimed.

As for going on about the whole IPC IP thing, alright, even if we do grant you that, it doesn't make your situation any better. It would mean that a blatant idiot is seeing through your bullshit. Trying to hide it with insults is really not helping your case.
>>
> 40 % IPC Improvement over Bulldozer
Notice it's "Bulldozer", not Excavator.

FPU/SIMD is being doubled so I do expect double the performance there.

Actually, to me the biggest improvement is the move to 14nm, as being stuck on 28nm has been AMD's Achilles heel. Though arguably, battery life has become "good enough" for most notebook products as long as you recharge everyday. The allowable TDP for chips can be reduced for the same performance too, as ultrabooks and 2-in-1s are all the rage in notebook design.

Like this Yoga 2 11in I'm typing on, many of these products lack the active cooling of a bigger product, which means the processor's max potential performance can only be in short bursts. Even light gaming is kill on this thing.

Also, like the Anandtech article about Carrizo in consumer products stated, AMD really needs to become more involved in how OEMs use their product. It just gives them a bad name when OEMs use a single channel mobo (remember Carrizo-L compatibility) and limit the cTDP to 15W.

I would love a Carrizo or Bristol Ridge system with full 35W TDP and dual channel memory in a 14in chasis with 1080p screen.
>>
>>52867214
Actually, it is over excavator. As for the 35 watt machines, the performance gain is minimum over the 15 watt version so it's not as much of an issue as some make it out to be.
>>
>>52860865
What. The. Fuck.
>>
>>52865781
BristolRidge is not Zen but Excevator.
>>
>>52860804
ahahah lol
how is ur cpu life goin ?
>>
>>52867214
Yeah, AMD is going to have to go their own way or contract somebody to make a decent premium laptop for them and show people what can be done if they want to dig themselves out of the "AMD is cheap trash" pit.

Right now AMD begs manufacturers to make a decent $700-800 laptop and they say lolno and continue to make 17" laptops with 1366x768 displays and one of the cheap jaguar parts in it so they can sell it for $330 in Walmart. Meanwhile OEMs are jumping over each other to make $1600 Intel laptops.

I think just one good reference design would go a long way.
>>
>>52867214
>>52867450
HP EliteBook 745 G3. The TDP is set to 15 watts but that doesn't make a huge difference in terms of performance.
>>
>>52867500
Isn't that still limited to single channel memory. That and it costs way too much.
>>
>>52867450
>if they want to dig themselves out of the "AMD is cheap trash" pit.
Retail cunt here.

In my store, the only thing that goes off the shelves is AMD APU powered laptops.
All the kiddies know that those laptops can play their games, and are affordable, so the parents agree.

Rarely seem them back in apart from maybe a damaged screen or cracked case to be serviced.

People have learnt that Intel are the "cheap shit" when it comes to mobile graphics performance.
>>
>>52867500
That might be the best product being sold with one of the chips in it but its still pretty lackluster.

I tried speccing it against an Intel 840 and found getting an equivalent build was next to impossible but I wasn't really seeing a cost savings even though an i5 is somewhere around $120 more than the Carrizo chip.

The 725 vs 820 wasn't really coming out any better.

Oh and using their build too really fucks everything up, building the exact same thing as their $1200 version with the 1080p IPS display suddenly costs $1600.

All in all HP's site is complete garbage for attempting to buy one of their products.
>>
>>52867522
It has to memory slots but yeah, it seems to only come with one populated. As for the price, is $700-800 not the price range you were going for? It won't get you the SSD version with the fully enabled a12 chip and the 2560x1440 screen but it's definitely a step in the right direction.
>>
>>52867529
lolno, you either have no idea what you're talking about or you're in a poor neighborhood.
Intel Core i5 and i7 with Nvidia graphics cards are all the rage in the gaming laptop market. Apple laptops all use Skylake or Broadwell processors. All the decent high-end stuff that moves in volumes are Intel. Only the poor buy AMD.
>>
>>52867599
>That might be the best product being sold with one of the chips in it but its still pretty lackluster
It's got the build quality of an enterprise laptop, you can get it with an SSD, a 2560x1440 screen AND it's cheaper than the Intel equivalent. What part, exactly, is lacklustre? I completely agree that HPs website is a fustercluck but that doesn't detract from the laptop itself. Generally speaking it's better to buy laptops in general through a third party anyway, with the notable exceptions of Dell and Apple.
>>
>>52867689
Gtfo. No shills allowed.
Back to >>>/reddit/ kid
>>
>>52867450
Beyond a referene laptop AMD should go after stuff like the Raspberry Pi and Nvidia's Jetson. Make a relatively inexpensive dev board with a Zen/Polaris SOC and good open source support.
>>
>>52867861
They could easily do that with their cat core processors or actually anyone could make them really. The have ones designed for embedded use and plenty of choices of TDP between 4w and 25w.
>>
>>52867714
I was actually looking closer at the 725 vs 820 honestly just because I like smaller laptops, I got a bit mixed up there when starting comparisons.

The closest prebuilt config I saw (that I would use at least) was the $1200 models with 1080p screen. I guess the price does come out fairly similar because the Intel one had a touchscreen (+$85 to Intel) the AMD did not but the AMD one has as a 256GB self encrypting drive (+171 to AMD) the Intel does not. Of course the AMD is configured wrong with a single 8GB stick at the wrong frequency that is costing it close to 40% of its graphics performance. Price difference between a A12-8800B and a i5-6200U should be more than the $86 seen here but I guess that isn't too bad.

Not sure I believe it without reviews but HP says the battery life is the same on both with a 44wh battery in each.

Here are the two I was comparing:
http://store.hp.com/us/en/pdp/business-solutions/hp-elitebook-725-g3-notebook-pc-%28energy-star%29-t1c17ut-aba
http://store.hp.com/us/en/pdp/business-solutions/hp-elitebook-820-g3-notebook-pc-%28energy-star%29-v1g99ut-aba
>>
>>52866134
Floating point unit != floating point pipeline
>>
>>52867931
They kind of had a product like that with the cat cores in the socket AM1 stuff, but they need to go even further for Zen.
>>
>>52868375
Well ALU != integer pipeline so something was horribly wrong with >>52866011 anyways.
>>
>people here unironicially trust manipulated benchmarks
Well to be honest some end user applications do use intel's compiler.
>>
I'm getting the suspicion that it won't be fast enought to make switch away from a 2500k, because fuck me, that thing is fast!


Just got it recently and it's literally overpowered for a Samsung 850 Pro and a HD 7870.
>>
>>52868499

>getting a 2500k now

Good lord anon, you'd be better off with an 8350 at this point.
>>
>>52868499
Eh, if I can overclock it and match my per core speed of my [email protected] but have twice as many cores it will probably be worth it just to upgrade off this aging Z68 mobo.
>>
>>52868431
An ALU is a type of integer pipeline, but both are a subset of an integer unit
>>
The first amd cpu that my house will have is gonna be an Athlon 5350, did i do good?
>>
>>52868941
At least you will not have to worry about housefires with that one.
>>
>>52868556
OC 2500K is better than stock 8350, which isn't worth overclocking. My 6350 is good enough though
>>
The only thing that would redeem Zen's pitiful single threaded performance would be how many PCIe3.0 lanes the chipset and CPU has. If it's more than x16+x8, then all hail Zen.
If it's the same or less, then why even bother.
>>
>>52869268
>Zen's pitiful single threaded performance
Citation needed
>>
>>52869300
Use maths
7% over Steamroller + 15% over Piledriver + 40% over Excavator < Kaby Lake's IPC
>>
fuck intel, I'm just going to buy POWER cpus
>>
>>52869331
>pitiful
>less IPC than an unreleased CPU
Pick one anon
>>
>>52869407
>Unreleased CPU has supposedly a lower IPC than another unreleased CPU
ftfy
>>
>>52869407
Everyone and their mother already knows that Zen will not match Broadwell's IPC. Skylake is a bump over Broadwell and Kaby Lake is a bigger bump over Skylake
>>
File: 1446930114072.jpg (92 KB, 523x440) Image search: [iqdb] [SauceNao] [Google]
1446930114072.jpg
92 KB, 523x440
>>52869260

The chart disagrees, an overclocked 8350 (which is what a 9370 and 9590 are) has quite a lead on the 2500k. You'd need to push that 2500k to the limit to get that sort of performance in vidya (and even then anything that can into cores will always have the 8350 ahead of the 2500k).
>>
>>52869432
Kaby is a Tick. Not a Tock.
It will be focussed on lowering TDP and overall power usage as opposed to increasing performance.
Just like Devil's Canyon vs Haswell, it will be 3% better outside of synthetics vs Skylake.
>>
Steamroller was still 20% below fucking Sandy Bridge.

http://www.cpubenchmark.net/compare.php?cmp[]=2362&cmp[]=790
>>
File: AMDfag.jpg (62 KB, 576x1024) Image search: [iqdb] [SauceNao] [Google]
AMDfag.jpg
62 KB, 576x1024
>this thread
>>
>>52869473
TDP is too high on an overclocked 8350. Also you're unlikely to get one that will overclock significantly
>>
>>52869622

>TDP is too high on an overclocked 8350

High? yes. Too high? Well depends which metric we are going by as many mobos and coolers (even air ones) can tame such a beast.

>Also you're unlikely to get one that will overclock significantly

This however is complete bullshit - 4.7ghz is easily done on the majority of 8350 chips (especially ones built in the last 2 years or so), it is just a matter of having a mobo and cooler capable of feeding and cooling it.

Pic semi-related, my 8320e running at 4.7ghz.
>>
>>52869473
>2500k has a TDP of 95W
>oc'd 8350 has a TDP of 300W
enjoy your obscene electric bill
>>
>>52869744
Is your PC the only thing you can afford to run in your place? Do you have to go down to the 7-eleven to microwave your cup noodles because you can't afford to do it at home?
>>
>>52869744

The 2500k won't have a TDP of 95w to compete against the 220w TDP of the 9590. The 2500k running at 4ghz (or perhaps higher!) will have a TDP closer to 150w (or perhaps more depending on the voltage it wants).

>>52869773

Even then how many people run their cpu at 100% 24/7? Sure some do but they are in the minority for desktop users.
>>
File: 2500k.png (55 KB, 417x402) Image search: [iqdb] [SauceNao] [Google]
2500k.png
55 KB, 417x402
>>52869773
>hurr if u dun want to waste more electricity to get less performance u must be poor
top kek

pic related, even my old oc'd 2500k would have consumed less power while giving more fps in GAYMING than your overclocked housefire
>>
File: Screenshot (247).png (29 KB, 510x545) Image search: [iqdb] [SauceNao] [Google]
Screenshot (247).png
29 KB, 510x545
>>52869808
[email protected] reporting in.
>>
>>52870025

>1.396v

Jesus thats got to be near the limit of what those chips can take without some exotic cooling.
>>
File: Screenshot (249).png (50 KB, 636x924) Image search: [iqdb] [SauceNao] [Google]
Screenshot (249).png
50 KB, 636x924
>>52870103
I'm not sure HWMonitor shows that correctly, it always seemed to me CPU-Z is more accurate for that (and this could be true for the power figures as well).

Compare the VID there to the VCore above in this pic. CPU-Z normally shows 1.312v-1.32v so I think its more accurate there.

If that was really running at almost 1.4v then it would be doing 150-160w at that clock speed.
>>
>>52860729
Forgotten
>>
>>52860729
IT WILL FUCKING REK INTEL
ITS MID END CHIPS ARE 3 TIMES THE PERFORMANCE OF INTELS TOP END
CHIPS
MARK MY WORDS
>>
Skylake is only slightly better than Excavator see

http://www.cpubenchmark.net/compare.php?cmp[]=2648&cmp[]=2654

An extra 40% IPC on top of that will kill Intel.
>>
>>52870491
so when's it coming out? 2026?
>>
>>52864309
3570k here, I'd like Zen to at least beat mine at 4.4Ghz. Going by expected numbers alone, we should at least be seeing Haslel performance, but I'm definitely expecting less than that. An 8 core behemoth would hopefully be the last silicon CPU I ever buy.
>>
>>52866899
>>52866833
>In computer architecture, instructions per clock (instruction per cycle or IPC) is one aspect of a processor's performance

Literally line one of the wiki entry on instructions per cycle. It's hardly some bizarro term nobody has ever heard of. They're completely interchangeable.
>>
i for one put my trust in Jim "certified shit-wrecker" Keller
>>
>>52861024

>it's for your own security, you just have to trust!!!

NSA needs to step up their shilling game
>>
Hey guys I'm tired of all this intel/AMD bullshit so I decided to build my own microprocessor. Where do I start? Any YouTube vids?
>>
>>52872509
you should get started on a logo
>>
>>52872550

All right, what about a snake rolled up a m4a1? That would be the gaming line.
>>
People would still buy intel even if it was good since AMD fuck everything up that that was something great. You'll all be disappointed. There's a reason why AMD barely exists compared to its competitors.
>>
>>52872509
Do you have your soldering iron handy?
Thread replies: 193
Thread images: 29
Thread DB ID: 515537



[Boards: 3 / a / aco / adv / an / asp / b / biz / c / cgl / ck / cm / co / d / diy / e / fa / fit / g / gd / gif / h / hc / his / hm / hr / i / ic / int / jp / k / lgbt / lit / m / mlp / mu / n / news / o / out / p / po / pol / qa / qst / r / r9k / s / s4s / sci / soc / sp / t / tg / toy / trash / trv / tv / u / v / vg / vip /vp / vr / w / wg / wsg / wsr / x / y] [Search | Home]

[Boards: 3 / a / aco / adv / an / asp / b / biz / c / cgl / ck / cm / co / d / diy / e / fa / fit / g / gd / gif / h / hc / his / hm / hr / i / ic / int / jp / k / lgbt / lit / m / mlp / mu / n / news / o / out / p / po / pol / qa / qst / r / r9k / s / s4s / sci / soc / sp / t / tg / toy / trash / trv / tv / u / v / vg / vip /vp / vr / w / wg / wsg / wsr / x / y] [Search | Home]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.
This is a 4chan archive - all of the shown content originated from that site. This means that 4Archive shows their content, archived. If you need information for a Poster - contact them.
If a post contains personal/copyrighted/illegal content, then use the post's [Report] link! If a post is not removed within 24h contact me at [email protected] with the post's information.