IT'S HAPPENING (possibly)
AMD has revoked its claim that Zen would offer 40% IPC change, and has instead begun saying Zen has HIGHER IPC than their original flat 40% prediction. That would be an odd move for them if Zen was actually worse than their original 40% prediction - which hints at the possibility that Zen is actually getting better as 14nm silicon from Samsung matures.
Kellar the Stellar is saving AMD. All that server and datacenter revenue will surely float AMD enough to accept thin margins for us desktop and laptop users.
AMD is going to go out of business before then. LOL at AMD lying to everyone.
Nobody can beat Intel. AMD never has and AMD never will. If they do, it's going to be because they copied Intel's chips and they'll be extremely buggy and burn themselves out in a week.
I will wait till I can get a chip in my hand. AMD pulled some dumb shit with bulldozer.
Quads AND a dank meme?
Today is your lucky day, goy.
Thanks m9, waiting till AMD gets a nice hevc decoder not the half assed one that does not do 10bit on there gpus too and hopefully then can put a full size hdmi 2.0 port in the back.
I highly doubt that. Even so, a 40% uplift over Excavator is still amazing, though. The 8350 was known for its low single core performance as well as its overall chip performance. When properly utilized, it could beat an i5/i7.
If they can retain their 8 cores, lower the power/heat with 14nm and get 40% IPC, that would really be something.
The 8350 even today isn't all too terrible despite being a few generations old. With Vulkan and proper core scaling, it actually comes back into i5 territory.
If Zen does all of these things, there's no reason that any Intel fan could be upset with what it's doing.
Intel does everything to make good processors, wastes billions every quarter and AMD just comes and makes a new processor with pocket change that's better? That's what you want? That's mighty unfair to Intel and just outright scummy from AMD
anon get your selective memory checked
HORY SHITTE AMD MY DICK AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
yeah I think I'm stuck with this processor forever, not that I'm complaining it runs since 5years now overclocked at 4.2ghz at a idle temp of 24c on my XBOXHUEG cooler, it does the job for everything now
AMD doesn't have to you cuck, no shit they'll never be able to beat the ultra-highend Intel chips (at least not for a while). All AMD has to do is offer a superior processor at a budget, doesn't have to be as efficient or fancy, just better. That's what AMD has done in the past, and that's what it's doing now.
People on /g/ love to joke about the bankruptcy of AMD, while in reality they are not that bad off. They're financials are improving, they are using higher quality manufacturing, they have plenty of massive contracts lined up in the future (ps5 etc.), and for the first time in a while they are expanding their market reach and branching off into the enterprise market more seriously by producing server chips.
They're pretty well protected from bankruptcy thanks to anti-monopoly laws, and licencing. It's likely Intel themselves would invest in them similar to what happened with Microshill and Apple.
One massive area of improvement is marketing, they desperately need to improve that, I'm sure most people here doesn't even know the 380x came out.
Granted I wouldn't put your college trust in AMD, but its definitely a stock worth considering, even if Zen fails the hype leading up to the release is sure to bump them up.
> Falling for the bait
You dump fucking cuck, do 3 minutes of research before opening your fat ugly edgy gentoo gob. AMD not playing fair fucking nek yourself you Intel shill faggot.
>That's mighty unfair to Intel
fucking underages man, you can't make this shit up
AMD gets absolutely ass fucked by both nvidia and Intel.
> Google 'gameworks', nvidia forces devs to use their shithouse framework that cuts %10 performance from AMD cards.
> Intel's aggressive marketing and bs lawsuit's almost killed AMD
I would go but I'm spoon feeding an autistic brick wall and am ashamed.
Would that be the same AMD that claims every thing the are about to release is the dogs bollocks irrespective of whether the product is a peice of shit or not?
The same "over-clockers dream" AMD?
The same AMD that name their CPUs shit like steamroller (because they'll steamroll over the competition) and then turn out to be outdated low performing fucking garbage cheap shit CPUs?
Call me sceptical if you want but I don't believe a single thing those AMD arsholes say.
Chizow, pls. go back to shitposting on currynigger tech.
He may very well be a currynigger cuck but he speaks the truth.
Steamroller wasn't shit because it was outdated or anything. It was shit because AMD gambled on the direction that software would utilize hardware and was totally wrong. If the
approach would have taken with consumer software, history would have been different. So no, he's totally wrong on that.
That "overclockers dream" shit was indeed true though. I'll give the filthy currynigger that though.
it's not just
> moar coars
because of the shared front ends, paired cores really have to be running identical threads, ideally with tight processing loops.
this is achievable in a lot of server farm situations, but virtually never in most workstation loads.
being able to really take advantage of highly tuned software and workload allocations is great, but you shouldn't bet the farm that all potential customers are both willing and competent enough to do so correctly.
Nah white as white blue eyed blonde northen European thank you very much.
You can't deny that AMD seriously over hype anything the are about to launch. It's ridiculous. No one seems to learn. btw I really wish everything they say were true. God knows the industry needs it.
why can't everybody be happy about this? amd making actual advancements will forced intel to compete with others, and in the end everyone gets faster hardware
>tfw didn't buy at $1.65
I wonder where the anons are who bought more than a thousand shares.
A 40% IPC would definitely put them above Sandy, but I don't know about Haswell. If Keller did a fucking miracle it would be at Haswell's level. I would buy the shit out of an 8 core behemoth with SMT that beats Devil's Canyon, but that isn't happening. Broadwell-E will do it, but I'd love AMD to do it as well.
>a full AMD in 2016-2017 might not be so far fetched
I actually don't think something going head to head with Haswell is too far fetched. Their CPUs aren't as far behind as most people seem to think. Going forward to SMT instead of CMT is really going to put them back on track along with the fact that it's on a lower lithographic process to Haswell. If Keller-magic is real, then it's not too far fetched.
If Skylake was actually good and people were saying Zen was going to beat Skylake, then I'd say they were retarded.
Your meme is only partially correct, good sir.
BD had slower single core perf, but it beat the Phenom II overall, assuming the entire chip (all 8 cores) were being properly utilized.
AMD built a radically different chip and birthed it into a market where it would have less than 10% market share.
As a result, most software still only uses 12.5% of the chip (one core). Not just that, but "core 0". The first core.
Look up any gaming benchmark out there, they tend to load all their shit onto core 0 and leave random trash to the other 3, 5, or 7 cores. Vulkan can't come fast enough.
Can't wait until it comes out and it turns out yet AGAIN to be another amd dud.
You guys keep hyping everything they do, you did it with every single little thing they released for fucking years already ever since the pre-bulldozer era and it always turned out to be a massive piece of shit in the end. And once the benchmarks come out showing it's shit you guys start backpedalling to damage control and saying "n-nobody ever s-said that it was g-gonna be b-better! Anyone s-saying so was c-clearly s-stupid". The most recent one being the Fury gpus which /g/ overhyped to hell and beyond for MONTHS and then pulled a complete 180 saying that nobody did so when it came out and it was shit. Same for the 390/390x which turned out to be yet another flock of rebrands when you guys were saying it was gonna make leave nvidia in the dust and blablabla.
It's like you people just simply never learn.
>why can't everybody be happy about this?
Because anyone with a normal functioning brain knows better than to fall for the latest "it will get better next year!" AMD meme that you guys keep pushing around and that has at full force for over half a decade already. It's just obnoxious at this point.
More like Bulldozer-Excavator
AMD gambled on the software industry shifting to a superior-if-utilized way of doing things. It wasn't utilized yet. Vulkan will change this, though... at least for gaming.
It's not a marketshare thing. They were beyond retarded to play on cores like this when everyone knows that dx and ogl don't allow interacting with a device out of the mainthread without notable overhead and significant work.
>All AMD has to do is offer a superior processor at a budget, doesn't have to be as efficient or fancy, just better. That's what AMD has done in the past, and that's what it's doing now.
I dunno, those FX processors were pretty pricey when they had the performance nailed down.
>I'm in the middle of disproving Logan and his Zen benchmarks
Ah, Logan, the guy that did a video comparison between ivy bridge vs bulldozer in fucking Arma 2, aka one of the worst optimized games of all time, where the bulldozer was magically running the game way beyond 60 fps and ivy was struggling.
For a short while.
AMD has been known to lower the price of a chip a lot as yields get better. Slowly they ramp down production until there's none left, and allow the successor to replace it.
>With Bulldozer, they slowly lowered their claims.
Bull-fucking-shit. My goddamn ass they did.
They kept saying that the shitty results were because windows 7 didn't properly utilize the CPU and released two placebo hotfixes that changed fuckall. Then they started saying that windows 8 was gonna fix it and unsurprisingly enough it remained the same piece of shit. They also claimed some shit about bios updates that would have to be made and blablalbala.
I got an fx 8120 in early 2012 and I know damn well what I went through. AMD can kiss my fucking ass.
so pretty much the kind of work loads normal people will never see.
AMD is better in only certain circumstances, if you are building your rig around that circumstance then yes go with AMD, otherwise intel is the over all winner between the two.
Its AMD employees jobs to hype up Zen as much as possible, cherry picked data ? who cares! its technically true!
AMD as been a long string of disappointments, so lets all just wait until proper benchmarks come out from trusted sites about Zen.
>it remained the same piece of shit
>so pretty much the kind of work loads normal people will never see.
The only FPU-heavy jobs most people do are gaming and video playback, and there's dedicated silicon for that. Name one thing other than a synthetic benchmark that leans heavily on FPUs, cannot be GPU accelerated, and normal people are likely to use.
Yep. Isn't it amazing how time flies? Mid 2006 was when the first Core 2 processors were released and Intel recovered from their Pentium 4 moment. Now in late 2016 AMD is poised to do the same thing with Zen.
>the Opteron, which itself was kind of lame too
Perhaps you missed the competition at that time. AMD had Opteron with HT, onboard mem controllers and Athlon64 arch, while Intel still had frontside buses and Netburst arch. Opteron stomped all over Intel.
Piledriver (your 8350, 6300, etc.) is about on par with C2D per core.
Assuming a raw 40% increase over EXCAVATOR, which is the marketing claim (which itself is 12ish% faster than Piledriver):
Nehalem was about 15% faster than C2D
Sandy was about 12.5% faster than Nehalem.
Ivy was about 5% faster than Sandy.
Haswell was about 10% faster than Ivy
(skipping Broadwell) Skylake is about 8% faster than Haswell
15+12.5+5+10+8 = 45.5
40+12.5 (piledriver + excavator) = 52.5
Ergo, Zen is faster than Skylake.
I think you're supposed to multiply those percentages, not add them.
110 * 1.15 * 1.125 * 1.05 * 1.1 * 1.08
That's like 62% faster than C2D.
1.4 * 1.125
That's like 57%.
I wouldn't expect Zen to BEAT Intel per-core, but if it's close and they have 8 cores, they'll be competing just fine. Especially if they make some budget chips available between the i3 and i5.
Anon it doesn't work like that.
Let the C2D be the initial value.
Then Nehalem is 1.15C2D.
Sandy is 1.125Nehalem or 1.125(1.15C2D) == 1.29375C2D.
See where it's going?
Assuming your data is correct, we can conclude that Skylake is about 1.6138C2D or 61% faster than the C2D.
Going with Zen and C2D as the initial value also including the extra from Piledriver you get that Zen is 1.568C2D or 57% faster than the C2D.
Pretty damn impressive if it holds. I won't believe it until I see it though.
All they had before the Opteron was the Athlon MP, which was a piece of half-assed garbage barely even worthy for workstations thanks in part to its lack of RDRAM support as well as few OEMs worth a shit shipping them (of course thanks to Intel paying them off, but a con's a con no matter what caused it)
Even when the Opteron came out, Intel had their own licensed 64-bit implementation running a year later to close the gap. I'm curious how Nocona/Irwindale and Cranford/Potomac actually performed in comparison though, now I'll have to go see if there are any worthwhile reviews.
A few specs assuming C2D is 1.
Nehalem = 1.15
Sandy = 1.29
Ivy = 1.36
Haswell = 1.49
Skylake = 1.61
Excavator = 1.12
Zen = 1.57
I'm on Ivy Bridge myself, I'd love to upgrade to an 8 core with multiple threads. For the love of Stallman, at least match Haswell.
Not talking about the Athlon MP, or the amd64 instruction set. Just that the core design (athlon64 vs netburst) and memory access differences (hypertransport and integrated controller vs frontside bus) were enough for Opteron to beat the Xeons of the time handily.
The pre-Opteron chips were originally what I was talking about. Anyway, here's an article I found on the early first-generation Opterons, comparing them to Intel's rather dated Gallatin-1M Xeons as well as their higher-end Pentium 4s:
I only really gave the results a cursory glance, but they looked quite close, at least closer than I expected.
AMD's core design was far more efficient clock for clock than NetBurst of course but Intel, retarded as their architecture was, could keep on pushing for more megahurts while their server chips featured large caches to make up for the difference in performance. I'm still looking for some good comparisons between the Opteron 8xx and the Xeon MP, which was what I was particularly interested in in the first place. I'd like to see if those big L3 caches were as tough as they looked.
That article misses something really important though, the actual reason Opteron's sold so well at the time and why Intel is leading now.
Netburst, especially the Xeon Netburst CONSUMED electricity on a rate not before seen outside of a Cray, (they ate more juice than the competing POWER and Sparc chips too...) where the Opterons were quite frugal.
This was at a time where we where blades started to become popular - density was up and still on the rise.
Datacenters literally were starving for electricity, I know a few DCs that were less than 1/3rd full because they simply could pull up enough to fill out.
Opteron smashed that shit out the park, low power consumption on equivalent (if not better) performance.
Even up to about ~3 years ago, AMD was still competitive in the datacenter simply because they offered 'moar coars' at a cheaper price - this is great for VM setups (but not so good for single thread workloads)
Once Intel ditched netburst and moved back to P6, x86 killed everything else in DC space - POWER and Sparc, dead. For the moment AMD is also dead, Intel won the DC fair and square.
>well 'fair' being that they didn't play their dirty tricks in DC space, they did that in corporate desktop and consumer space.
Lessee... The Opterons win on memory and overall performance while the Pentiums try and make up the gap with sheer clockspeed and bigger cache. Looks about right. The gap definitely is closer than I expected as well. But the telling thing to me is how the Xeons are generally getting BTFO, with the desktop P4 line doing notably better than them.
Well, it's not like Sun, Fujitsu and IBM weren't doing exactly the same thing before the rise of x86.
We're at a crossroads again though, ARM is going to be though competition in the future, considering that anyone can buy a license.
A cursory glance at the TOP500 lists around 2004-2006 seems to be an even mix of mostly exotic hardware like Blue Genes, Itaniums and then a small population of x86-based systems with an even share between Xeon and Opteron systems, the latter of which were mostly built by Cray. Could be wrong though, I didn't check it out exhaustively.
You're right, but looking at the TDPs of the Gallatin chips, the difference was on an order of around 5 watts more towards Intel. Absolutely pathetic, but not necessarily a huge deal breaker. The 64-bit Nocona chips that replaced them ran at a TDP of around 103-110 watts however, which AMD beat out by up to 20 watts depending on what chip you're comparing, some of their late 2004 high-end offerings hit 104W themselves though.
>But the telling thing to me is how the Xeons are generally getting BTFO, with the desktop P4 line doing notably better than them.
They comment on that in the article as well, as did I, the Xeons used in the comparison were getting long in the teeth, running on a 533 MHz bus and DDR 266 memory controllers. Intel released updated Nocona chips with a more competitive 800 MHz bus and faster memory support the month after that article was written, don't really know how how much better they would have fared with it.
Intel touted their higher-end P4s as workstation-grade parts, which I guess is feasible for a single-socket setup.
Shit, do you think POWER and SPARC were any cheaper in their prime? My primary slowlaris workstation, a Blade 150, cost in excess of $4,000 in 2004 despite being basically a crude PC clone with a long dated 650 MHz UltraSPARC II chip grafted on.
>u didn't buy amd? must be jew, neckbeard or both
It's not a bad deal. He gets allowance money from intel, enough to sustain an unhealthy diet and rent of an apartment, so he will have no need for other jobs but a lot of free time to shitpost all over the internet. It's beneficial to both of them.
After a while though, it can go into his head, he develops some sort of father-son complex in his head, and might even become homosexual, but that's the small price he pays to have a sugar daddy take care of his living costs for the rest of his life.
It'll work this time guys! Intel is FINISHED!
Intel has had many failures. Pretty much every 4004-8086 derivative was successful but after 8086 Intel pretty much couldn't invent anything better. Except the i960 ISA which was used in microcontrollers.