If we actually do get 40% better IPC over excavator then not only will it save AMD but it will make intel sweat bullets. The only reason intel is the popular choice by OEMs is because their chips use less electricity and expel less heat. Both qualities sought after for use in laptops and tablets.
Now with zen it won't make much sense to use intel anymore especially since the iGPUs are always better on AMD chips.
>>52455039 Not really, skylake laptops are already hitting the market and some of Lenovo's new offerings this year get 16 hours of battery life and will blow the shit out of Zen too if it ever makes it into any laptops that people actually buy.
>>52455264 Well if Zen does deliver then theoretically a dual-core Zen will compete with a haswell i3. We don't know how energy efficient Zen chips will be but looking at carrizo laptop APUs, I'm willing to be they will be pretty fucking energy efficient especially since they will be on 14nm. Maybe not as energy efficient as skylake but pretty dam close.
Anyway what will attract OEMs more will be the igpu anyway. Zen will have hardware acceleration for 10-bit H264 and HEVC I think. We might get laptop Zen APUs with HBM so gaymen laptop OEMs will take great interest in Zen as well.
>>52455264 >some of Lenovo's new offerings this year get 16 hours of battery life That's because 75% of the laptop itself is a battery now lel. It would still get around 16 hours of battery life even with a zen chip.
>>52451481 You know, a few months ago I would have said Yes. But with the market heading down some treacherous waters, I don't think Zen will have the impact on AMD's sales that it needs to. PC sales have declined more than ever last year as tablets and smartphones also began to stagnate. PC game sales haven't been going up at the rate that the market needs it to in order to keep the desktop market alive, indicating that the number of PC gamers are not growing at a rate necessary to counteract the steadily falling consumer desktop sales.
AMD, Intel, Nvidia, and countless of other manufacturers were gambling on the PC gamer market expanding, which is why Skylake launched with such a heavy emphasis towards gamers even in segments targeted for business markets (Xeon E3 C236 chipset gamer boards? It's already happened). But at this current trend, AMD is unlikely to secure enough sales to counter their cash burn merely because there are not enough people getting into PC gaming as it was once thought. VR might not be the silver bullet either, since they're launching at absurdly high prices with equally absurd requirements (+$300 for a peripheral that requires a $1000 minimum PC to use?). 2016 looks like it's going to be a bleak year for everyone. When Zen fails to save AMD, it won't be from a lack of effort.
It will do more than save AMD, it's going to rape intel without lube. Not only will people start choosing AMD Zen over intel due to the lower prices but OEMs will too. AMD already showed us how based their integrated graphics are with kaveri chips are and the ones in Zen chips will only be better.
Zen is going to fucking rape intel not only in the desktop market but the laptop and server market as well. Rumour is AMD plans to btfo intel on the server market with a 32-core zen processor with a big ass igpu.
Remember how everyone was saying all the exact same shit about bulldozer? Remember how it was outperformed by their own phenom 2 chips and came nowhere close to sandy bridge? Of course you don't because you were probably 12 at the time.
>>52455719 I remember bulldozer/piledriver performing about on par with the mid range sandy bridge chips but with somewhat worse power usage on average. Thing is bulldozer/piledriver come with all instruction set extensions while Intel limits their more useful extensions to their mid to high end chips. Some chips lack useful features completely regardless of what tier you buy like IOMMU on K series chips.
>>52456408 Stock prices do not reflect the overall performance of a company. That's like saying that having a high temperature is a disease rather than a sign of one. The clear indicator is that AMD is still not operating off a net profit and their sales in nearly all markets has either stagnated or shrunk in 2015. The stock prices merely reflect that (as well as general panic over tech stock prices thanks to fucking Apple)
>>52456543 There is no single test or benchmark that is a clear indicator of IPC alone. You'd need to run several different programs and benchmarks aimed at taking CPU performance into account and draw your conclusion from the overall performance. Anandtech is one of the few sites that do this for CPUs, but they don't have all the CPUs out there tested.
>>52455264 >!BE/4wes0mE How much do Intel and Nvidia pay you for your shilling? I'm genuinely interested, because _every_ hardware-related post I see with your tripcode attached is basically just shitposting against AMD. At least you had some cherrypicked benchmarks with the GPU drivers earlier, now it's like you're not even trying.
>>52456600 AMD is going to sink with or without Zen. They won't have enough cash on hand to survive the crumbling market. They've burned far too much of it already when the going was good, but now that the global market is in the shitters, they won't be able to survive for much longer without cold, hard cash to run on.
>Going forward, as new silicon generations are introduced, they will require the latest Windows platform at that time for support. This enables us to focus on deep integration between Windows and the silicon, while maintaining maximum reliability and compatibility with previous generations of platform and silicon. For example, Windows 10 will be the only supported Windows platform on Intel’s upcoming “Kaby Lake” silicon, Qualcomm’s upcoming “8996” silicon, and AMD’s upcoming “Bristol Ridge” silicon.
>Zen comes after Bristol Ridge & only supported in BOTNET 10
>>52456601 Surprise surprise Different processors process different instruction sets at different rates. IPC is not a single number for everything. You might get 8 instructions per clock in some niche workload, and get 2 per clock in others.
>>52455839 >they will have to stop manufacturing CPUs to not go bankrupt.
Its hilarious seeing you shit eating morons continually regurgitate this line over and over.
>>52456706 Your post really makes you sound like a 10 year old kid, thinking that companies fight each other like pokemon where one has to "win."
Zen based Summit Ridge, Raven Ridge, and Opterons bringing a positive ROI will allow AMD to pay down their long term debt and maintain operating cash in hand. With that $2,000,000 burden gone they can reinvest all profits into the company, and likely start paying dividends to investors in a couple years. The specifics of how the Zen core arch compares against intel's core per clock doesn't matter at all, its a significant uplift from their current offerings which alone is a selling point. There is no failing when a design is profitable.
>>52456795 >bringing a positive ROI will allow AMD to pay down their long term debt and maintain operating cash in hand It might be too late. Word is that AMD is burning through their last reserves at a faster rate than Q1 2015. That is a serious concern, especially since the global economy is on the verge of tanking and AMD doesn't have a hope of turning a profit off of Zen until Q1 2017. There is a strong possibility that they will run out of cash and have to take on more debt, if they can even do that after 2016.
>>52456662 Right, this is the canned response I get repeatedly. So what set of applications are best used to test the different workloads to get the whole picture? I've never seen anyone offer this up, just vagaries.
>>52451481 I hope so. I plan to invest the equivalent of 5k USD into AMD stocks prior to the launch of Zen. If it goes well, I'll stand to make a pretty penny. If it goes poorly, well, it'll be with money I've borrowed from a bank so I'll either pay off the loan as I can or just kill myself depending on how badly it turns out.
>>52456893 Cinebench is one, but Handbrake with GPU acceleration disabled is another. The problem with most of these programs is that they're also dependent on other factors, like RAM, storage, and even the motherboard.
Be sure to measure power usage while you do all of those things.
>>52456893 There is no single test for the "whole picture." You processor works with dozens of different instruction sets. Synthetic benches try to focus on more narrow ranges of a processor, while others attempt to simulate real world workloads which are more varied and far less intensive than synthetics.
>>52456962 I built 196USD, before rebates, PC for my dad, can play a bunch of games better than my 4670k, though his is running linux, so not a focus. They're pretty good bang for buck, if you aren't a hardcore gamer or whatever.
>>52456962 I meant for AMD. Most people with prebuilts don't really care about graphics cards so they would be fine with pretty much any APU. People like us who need discrete GPUs would be screwed though.
This is why - as a rule - /g/ is completely ignorant of how cpus work and what specific chips are good/not good at. If there isn't an easy chart with a blue bar longer than a red one to point and and shout memes about /g/ doesn't understand.
>>52457469 also desktop computing is coming to an end. there is no need for powerfull processors or gpu's anymore because nothing is being made to use the hardware. If you do rendering or heaavy computation there is a need but that is a very niche market now.
>>52457031 They are pretty good bang for the buck, when you might want to play some not very demanding games slightly better than other integrated graphics, but not demanding games. That's actually a niche.
The "eh I might want to play some game but not properly" mentality is stupid, and thankfully not common around me.
>>52457105 >Most people with prebuilts don't really care about graphics cards so they would be fine with pretty much any APU Exactly. They'd be fine with a weaker igpu too, like intels. >People like us who need discrete GPUs would be screwed though. exactly.
>>52457761 nope. Consoles have always been under powered its just recently that the poewer of consoles is close enough to pc that devs said fuck off to the pc market, and all we get is console ports. You think they are just going to jump over to pc for no reason all of a sudden. Its over man PC is in its death throws.
You appear to have it backwards amigo. This is the first gen where you can build a pc thats morep woerful for similar money. To wit on the 360's launch you couldn't get that class of hardware for less than 1 thousand shekels.
>>52457745 A quad socket rack with huge APUs would have better perf/watt than a quad socket rack with CPUs and discrete GPUs of equal performance. The APUs would also have all the same benefits in socket to socket and rack to rack communication from their Freedom Fabric IP which is substantial. No one needs to write code specifically for HSA, any OpenCL2.0 code will benefit due to feature parity. JAVA 8 also incorporates HSA, whil JAVA 9 takes it a step further and will make it complete ubiquitous.
>>52458962 as far as I'm concerned a multiplat is a console port. You take away the multiplats and theres like 2 devs that are AAA making games for pc unless they are making some f2p garbage or indi games.
>>52460327 Lawsuit over what? New hardware not being supported on a OS that has officially passed its 'no new features' date? Its time to face facts, if you don't like the direction Microsoft is heading with Windows, its time to take the alternatives seriously and stop living in the past on a OS that's four years from its total end of all support.
>>52460704 It may be artificial, but as with the restriction of DX12 to 10 Microsoft is well within their rights. Like users who clung to their Be Boxes and Amigas a bit too long its time to face the fact that Windows 7 isn't a current platform anymore and its time to move on.
The sad thing is if AMD dies there won't be any other company out there like them. Whoever acquires them or their assets are just going to lock them away.
I think the only other company like AMD was Sun but they were acquired by an absolutely vile company that immediately told the open source community to fuck off and ruined basically everything Sun had built up.
>>52460977 No, the most valuable part of AMD is Radeon.
Nobody in their right mind would launch a new x86 manufacturer, even if AMD's x86 license was transferable. Just look at the long, long graveyard of 'em - Transmeta, VIA (life support from netbook money), Cyrix and Centaur (gobbled up by VIA)
IIRC even for all the failings their cpu division is more profitable than their gpu division. Admittedly with Lisa "fucking schlomo's shit up" Su, Jim " certified shit wrecker" Keller and rajeet being in charge of gpus AMD has built itself over the last 4-5 years to suddenly eco-boom waaaaaaaaaaaay up to Intel and Nvidia's level by playing the long game.
Dominanting the HSA market (as niche as iti s) and getting the console contract puts AMD ion the unique position of dictating the direction of a lot of hardware despite being a tiny player. Their entire gameplan is based around "the enemy of my enemy is my friend" and from initial viewing it appears to be working.
Now that I'm fully aware of the botnet nature of modern AMD/Intel processors I give much less of a fuck about Zen than I used to. I just want it to succeed for economic reasons because fuck monopolies. IBM is the horse I really want to back with OpenPOWER.
>>52451481 Not if what MS said is true, about it not working with W7 and W8.1. It's still x86, so I don't see how that is even possible. Outside of MS releasing an update to specifically make W7 and 8.1 refuse to boot on Zen/Kaby Lel, I have no idea how it could be true, but if it's an update you can just install from an older ISO and disable that update from installing in the first place. Or disable updates entirely.
>>52462056 That isn't what the Microsoft rep said, stop buying into clickbait. You can install Win7 on anything the same as you can with Wn95 or XP. Being unsupported and not working are two entirely different things.
>Will Zen save AMD? It'll be just like the R9 300 series.
Zen will basically be 1-2% worse than intel's equivalent offerings in overall performance. This will cause an uproar of people saying that AMD is shit for not meeting their excessive expectations, and to never recommend AMD to people because they think they'll benefit from that marginal performance difference.
>>52456647 unless your daddy is an intel investor you are fucking retarded if you want intel's only competitor to go out of business. From the consumer's point of view its shit, no competition means monopoly and then they will start charging you double for cpus 2% better than last years
>>52462115 Being unsupported can also mean you could get various system peripherals that don't work with any drivers W7 currently has and that no drivers for it will ever be made (by MS, Intel or AMD). I don't doubt that W7 will be able to boot, but even so the experience may be shitty and lacking (or downright unusable) if you're going to be missing official drivers for various things. I don't mean the CPU itself, but shit like the new chipsets for Kaby Lake and Zen.
>>52462300 >Zen will basically be 1-2% worse than intel's equivalent offerings in overall performance. What exactly do you mean by that? In terms of IPC or what? I kind of doubt a 4C Zen CPU is going to be that close to 4C Kaby Lake, even if Intel does their typical 5% improvement shit.
Zen and polaris will boost market shares. Nvidia is having issues with new gpus, and could potentially fall behind. Intel on the other hand, has been holding back for some time. And zen needs to live up to the hype, or else Intel can just drop amd out of the market. It's a scary time for AMD, and I really hope they give me a reason to switch over
>Multiplayer-centric AAA titles with only token single player support >F2P-Hybrid AAA games where you buy half the game and pay for the rest through DLC microtransactions >Offering discount games through several license management services like Steam >Slowly adapting all these things to the console market
>>52461578 To be fair though, many developers are pants on head retarded. They have zero fundamental soundness in designing their game/render path, made up for by frantically shoehorning in optimizations for each console version to meet a 25-30fps target, and it shows. This leaves it to the driver developers to basically re-engineer everything in the driver. This is literally what happened for say, Assassin's Creed Unity.
Luckily there's still games like Grand Theft Auto V and Witcher 3 coming out to show that there are still good programmers left in video games.
>>52456348 >>52455476 good thing amd doesn't even plan for the zen to be sold in volume to consumers, it's their way back into the server market, which they plan on seeing most of their gains from.
as for the consumer, if you have a core 2 quad or even a phenom II x4, you have VERY little reason to upgrade, especially if you don't play at above 60fps, so there goes the gamer aspect of "got to upgrade my 4770 to 6770 or else i cant play the newest games" angel, and if you just go on a software side, ever since dual core chip there has been little reason for normal users to upgrade. the only reason you buy a new desktop now is
1) you can use the power to its fullest 2) your current hardware is lacking 3) your current hardware broke 4) you are actually retarded and think this is the early 90s where every 4-6 months computer hardware made serious gains and you must upgrade
>>52462097 >>52457516 not really a joke sure games can push a gpu, but that is a niche in itself, most people don't even need more than a dual core system with how they use them, its why cellphones and tablets are so popular and used.
hell, if no one gimped performance at all with updates you think anyone would have the newest apple thing?
AMD won't die for a long time as long as consoles are popular. They only danger is someone buying them and cannibalizing their product lines and structure to the point that they can't innovate outside of specific parameters.
>>52464216 >most people don't even need more than a dual core system with how they use them
This is true in the sense that web browsing/gaming technically works on old hardware... but it takes a special kind of retard to honestly believe even the most average normie isn't getting a huge benefit from faster cpus. Everyone, no matter how light their usage, perform tons of cpu-heavy operations in quick bursts behind the scenes.
If you literally only browse forums or fb, then yeah... maybe an upgrade isn't the most effective use of $400. It is an appreciable difference though, and anyone who uses their computer a lot will notice.
>>52465317 A bit slower than Haswell but offering 6 (or even 8) cores at the typical unlocked "mainstream" i7 price point could be pretty nice. If it OC's decently too, that could be a pretty sweet desktop CPU.
>>52455476 >MUH PC market share decline sales projection.
It doesn't matter if PC grows or decline. AMD's problem is that nvidia or intel infintatley more aggressive with their proprietary software/hostile business practice. Nvidia/Intell have secured their market before the pc market was even on decline.
>>52456898 You say that but the profit margins on APUs is not very good. Most APUs sold are in the bargin-bin shitbooks that populate best buy's shelves that people simply buy because they're the cheapest laptops.
>>52466803 What I mean is APUs integrate several things into one slab of silicon. Having one die fabbed vs several is more efficient. There's pretty much no way AMD would ever go back to having separate dies for everything it's just unneeded costs.
>>52467042 Bulldozer was faster in cases where threads made the difference and it has newer instruction set extensions which can give it a significant advantage in newer applications. When comparing older applications bulldozer was often within 10% of Phenom II. The main problem with bulldozer was the power usage and I'm starting to believe that was mainly due to the leakage from the node technology. Kaveri and Carrizo are very power efficient
>>52467108 >The main problem with bulldozer was the power usage and I'm starting to believe that was mainly due to the leakage from the node technology. it was ENTIRELY due to GloFo's leaky shitty 32nm node and the clock rates they were trying to hit.
Now that it's been refined over 4 years later, they can make 8-core piledriver chips that end up 95w at 3.7-4Ghz.
I dunno, like, literally if AMD would simply release a flagship Dual core CPU with performance that beats intel in Single-Threaded applications and had modern instruction sets for encryption and such things, I'd totally go with AMD.
Them single-threaded photoshop filters and render things on High Resolution images make you cry when you have 8 cores and can only use 12.5% of them. Also saving high res images as PNGs, fuck AMD's whole MOAR COARZ mindset
>>52467463 Seriously though during the Athlon AMD had almost 50% of the market and they were running ads to demonstrate the superiority of CPUs vs Intel's. Even the names of their CPUs were made to show how superior they were. Athlons came with a vague numbers at the end like 3000+ which was to show that even though they were 2GHz they were comparable to an Intel processor running at 3GHz. They were wrecking Intel's shit.
>>52467454 >>52467384 I have a 6700K. Funny thing is, 90% of the time, if I'm not encoding or rendering video content I don't use more than 2 of the cores unless I'm running multiple games at the same time. A lot of the time I'm slowed down to single-core speeds.
I'd rather have two fast cores than 8 slow as shit ones. knowwhatimsayin?
>>52468530 >I've never seen a single review of any of these chips ever
Steamroller isless than 10% faster than Piledriver at equal clocks. Cache latency took a big hit which hampered all of the performance gains, that second decoder does almost nothing unless you're running some oddball integer synthetic on all threads. Excavator brought about the same uplift. By AMD's own admission the average IPC uplift was only 5%, though it reaches 14% in a few niche workloads.
They never gained 20% IPC between generations, even Piledriver was a 10% uplift, much of the performance over Bulldozer came from higher clocks.
Thread replies: 225 Thread images: 25
Thread DB ID: 435471
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@example.com with the post's information.