>skylake 10% increase at best in performance
>kaby lake just new bullshit for the chipset
>up to 24 pci-e slots
>up to 10 usb 3.0 ports
time to buy haswell before they shoot up to 400$
What's the news?
Kaby was always just another 14nm-step with minor additions, 10bit HEVC hw-decoding and native USB3.1-support being two of them.
>>52135147
>rebrand
I believe it's just a tock, you fuck.
>inb4 muh ddr4
enjoy getting goyed for 3% performance
>>52135147
>>52135221
it's not a tock, skylake was a tock. the next tick will be cannonlake.
they're realeasing kaby lake between these two as a skylake refresh, just like with haswell (devil's canyon)
>>52135245
Wasn't Skylake the 14nm tick? Or was that the mobile Broadwell?
>>52135288
broadwell was the 14nm tick (haswell shrink)
>>52135147
I don't get it. Why would this increase the price of haswell?
>Enthusiast 95W Options
hah wtf
>>52135288
Broadwell Core M were the first 14nm parts, they launched in late 2014. Performance uplift over Haswell was really tiny.
Skylake, still 14nm, had a really small uplift over Broadwell.
Intel clearly isn't doing tick-tock any more.
They're sticking with 14nm for Kaby because the node still has piss poor yields, and they won't have their 10nm sorted out any time soon. Desktop Broadwell totally disappearing coupled with really low supply of high end desktop Skylake has been something of a scandal for industry analysts and process technicians. Everyone saw trouble back in 2014 when the Core M lineup was launched, the differentiation among the SKUs was a complete mess.
>>52135376
Plot Twist, high TDP SKUs just have more IGP. Intel is increasing their eDRAM size from 128MB to 256MB.
Now or never, AMD, now or never.
I have got a i5-4670 I was wondering if there was any reason at all to upgrade to a skylake but from what I've gathered it's the same shit huh?
How about I go for a i7-4790k instead?
What do you guys think
>>52135147
why do you fucking idiots keep using wccftech as a source?
literally 95% of their articles are based on rumors and speculation and use laughable sources
Anyone here knows when intel blocks the non-k skylake overclocking? i want to buy one but currently i just have no money
>>52135816
because it's great as a bait
you'll see 300 post threads from op linking in a wccftech/currynigger tech post
>>52135863
why would they tell us
>>52135635
not worth it for gaming you'll get like 6 extra fps.
>wccftech
These guys just publish any old crap. Putting in the effort and actually looking at their old predictions will tell you it's all a load of shit.
>>52135918
Sure but all in all responsiveness of my os should go up right? Also I could watch 1080p videos on my second screen while playing vidya without losing fps with the 8 threads no ?
>>52135918
He wouldn't gain jackshit. If anything he would actually lose FPS because hyperthreading is shit for games.
>>52136085
Dude, your 4670 is more than enough for any you use it for. Seriously, needing a fucking i7 to watch a video while playing something? Where the fuck did you get this from?
>>52135147
>There are people who still buy intel products
Good goyim, all of you.
>>52136094
You sound like you get triggered by the word "i-7"
Whats it to you if I have the dosh, I just wanna know if it would make my os more responsive
>>52138733
Not at that level. Get an SSD if you don't have one, that actually will help.
Kaby Lake is not a rebrand when it adds full fixed function HEVC Main10/10bit & VP9 10bit hardware decoding, HDMI 2.0/HDCP 2.2 support
OP IS A DUMB FAGGOT READING SHITTY CURRYTECH SITE
>>52138860
But what about performance or power consumption? Who cares about hardware decoders that won't see much use for at least 2-3 years.
>>52138879
What are you expecting? It's on the same Intel 14nm process, there's no changes
>>52135450
>Intel is increasing their eDRAM size from 128MB to 256MB.
800usd 4cores in coming.
>>52135523
This. God I hope and pray they pull out a win with zen.
>>52138902
What I expect? What I expect is the faggots at Intel to start actually giving us processors not tablet-sized dies sold for $300
>people willingly support the jewtel monopoly
Kaby lake is a tock not a refresh, they had to break the cycle because 10nm fabs are not ready so there will be two tocks in a row. Broadwell-E will be the refresh of next year.
>>52138879
>Who cares about hardware decoders that won't see much use for at least 2-3 years.
watch out, your neetpoorfagism is showing.
4K HEVC 10bit is being used right now, fella.
>>52139582
No it isn't.
And it's not gonna be used until available HEVC encoders can output nonshit.
>>52135147
>cuck lake
>>52139713
You're a moron
Netflix uses it
Now STFU and get lost
>>52141131
>Netflix
You planning on giving me an example of nonshit output yet?
Netflix, haha oh man some people are really something.
>>52139380
>2012+3
>Not knowing what 'monopoly' means
i shiggidy digg
>>52135523
The problem is that Intel could sell their quad core i7's for $100 and still turn profit. They just don't have to because there is no competition at the moment but if Zen is even nearly as good as Intel CPUs Intel can just outcompete AMD with price.
>>52141314
Intel DOES have a de facto monopoly on high-end desktop and laptop CPUs.
>>52141324
>The problem is that Intel could sell their quad core i7's for $100 and still turn profit.
Maybe if they completely gutted their R&D. Intel is profitable, but not THAT profitable.
>>52135147
>What is Tick Tock
>>52141636
kaby lake is outside of tick-tock. it's a refresh.
>>52135224
DDR4 is cheap now nigger
>>52135523
>>52139280
Zen might, MIGHT, get up to 1st gen Haswell performance as a best case scenario
Just like Fury was supposed to be the 980ti killer and the doorway to the next generation in graphics power, or when Bulldozer was supposed to be good
But we all know how that panned out
>>52139509
But tock implies new architecture
Isn't Kaby Lake just going to be the Devil's Canyon for Skylake?
>>52141901
>Just like Fury was supposed to be the 980ti killer and the doorway to the next generation in graphics power
Fury X was supposed to be competing against Titan X at significantly cheaper price. Nvidia just got 980ti out before AMD.
>>52135147
Jesus fuck AMD, please pull out a win with Zen. Intel isn't even hiding that they aren't going to increase performance.
>>52141979
But Fury has much more raw horsepower than the Titan and yet AMD still couldn't harness it for any real gains
>>52141922
>Isn't Kaby Lake just going to be the Devil's Canyon for Skylake?
Kaby Lake is virtually the Haswell Refresh of Skylake. It will be virtually the same as Skylake except for chipset update (standard USB 3.1 - which you can already get with Skylake on Thunderbolt 3 boards) and some IGPU improvements.
Cannonlake is after that, and it probably will be nothing more than a die shrink of Kaby Lake with yet again some more minor improvements to the chipset and IGPU.
There are no performance improvements coming out on the Intel front for a long, long time.
This is literally the perfect time for AMD to come romping home with Zen, assuming Zen is actually a decent chip, as Intel have left the market wide open until 2018.
>>52142045
64 ROPs is really killing the Fury series in some games where they fail to perform any better than 390X.
>people actually believe AMD will get their act together in time for 2016
>despite the fact that AMD's R&D budget has been shrinking by millions of dollars every year AND has to share that budget with its CPU, GPU, embedded solutions, and enterprise departments
Delusional, all of you
Skip Skylake, Skip Kaby Lake, Skip Cannonlake, and Skip Zen
We're never going to see leaps and bounds in performance gains like we did in the early 2000s until a significant shift in the ways CPUs work comes to market.
Kaby Lake and Cannonlake will be the usual Intel Jewery, while AMD will go full Jew and feature-locking with Zen because they need to make more money.
The future's so bright, we don't need our eyes to see anymore.
>>52142323
Intel has said it would no longer be feasible to use silicon after Cannonlake 10nm and that they would use a different material like a graphene composite or something
Real change will come with quantum computing, but that's still a long ways out
>>52142401
Then wait until graphene CPUs come out to replace your Sandy Vaginas
>>52142323
Considering AMD's R&D budget was going almost all to advertising that means nothing.
>>52142464
>Considering AMD's R&D budget was going almost all to advertising that means nothing.
That's even worse. How much money is even going into developing Zen and the Artic Islands?
>>52135147
And here I am still using a 2500k.
>>52135147
>tfw your supervisor went home for christmas and he lived beside microcenter
>give him $220 to get a xeon 1231
>they were sold out so got a 4790k
>told me not to worry about the rest
Guess it makes up for the dead fx6300 I got a year ago
>>52142478
Zen is already done with development for quite some time now. It's just in testing now.
>>52142603
What about its successors, the so-called Zen+?
What about a die-shrink?
What about improved and added features that AMD has had no real experience in and has to rely on outside vendors to accomplish?
What about actual innovation that can help bring AMD back to a competitive level with Intel?
You can't do that with a shrinking R&D budget.
>>52142654
Why are you believing their advertising team that clearly has no real clue what's going on as seen in the past?
>>52142514
>and here i am still being poor
>>52142514
>>52142884
And here I am with one of the best processors ever made
>he cant afford an i7 6700k
Haswell and Ivy Bridge-E is the last processors worth buying.
>>52142323
>Skip Zen
considering how shit AMD's current offerings are, if it stands up to marketing material (never ever) it might just be leaps and bounds ahead compared to their own
>>52142943
>buying 6700k instead of haswell-e
>>52143265
>buying Haswell E if you aren't specifically rendering or doing other multicore intensive tasks
It's like you want lower IPC
>>52142943
>he enjoys getting cucked
>>52143265
>buying haswell-e when broadwell-e is so close
>>52135635
the non K? i hav the 4670K @ 4.4ghz, no reason what so ever to upgrade.
Do you think there will be a time in the distant future where it will be possible to unlock locked haswell processors?
>>52135816
most of the shit they post are legit articles.
the cpu/gpu rumors tend to be true (havent seen any false one yet)
the only rumors that are laughable are the ones regarding phones, but who gives a shit about phones anyway
>>52143298
If you aren't doing any of that, why would you get an i7 in the first place? If it's just for gaymin you might as well get an i5.
I'm still running an i7 920 @ 4.4ghz and see literally no reason to upgrade.
>>52143406
you can do it now but its a shitty overclock that could ruin your entire system not worth it.
>>52143476
More games like GTA V and Witcher are using hyperthreading, and it can still be useful if you do light rendering or photoshop or compiling
I'm talking if you are serious about content creation vs just having it as a hobby
>>52138816
This
>>52135635
Man I have an i5-2500k from forever ago and I dunno what to upgrade to; I'm seeing all sorts of shit talking about Memelake; Haswell decimating stuff. I just have no clue.
>used 4th gens are going for higher prices than new skylakes
lmao anything past haswell is shit and not worth it
video games are going gpu dependent past 2016.
>intel stops producing the greatest cpu ever to charge people for meme's
why cant we have a 150$ 4770k right now?
fuck intel
>>52143495
I'm not going to bother with it then I guess.
It's just I wasn't informed enough about intel processors and looking at my budget back when I bought my current PC I got a 4460. And maybe in the near future I will regret it (since I don't know how much of a bottleneck would it be for something like a rebranded 390 in the future or whatever) but whatevs.
>>52143545
At the moment 5820k actually costs LESS than 6700k at least at newegg. Even if the motherboard costs a little extra, there's still no point at all in getting 6700k unless you absolutely MUST have the best possible single-threaded performance as well as 4 cores+hyperthreading. Anything that benefits from hyperthreading will benefit even more from having more real cores and the IPC difference is not worth worrying over.
>>52143767
>literally buying inferior obsolete hardware with 20% less IPC because more coars meme
AMD users
>what should I upgrade to
If you are asking yourself this question you probably don't need the extra power and you are a moron for supporting consumerism.
>>52143767
>buying 6700K
I seriously hope you don't do this. Get a 6700. You can overclock them now on Z170 chipsets. K = scam now (scam always, but especially now)
>>52143825
>20%
Try 8%, if that. And again, if you don't need more cores, you should be getting an i5.
>>52143909
You can only do some BCLK tweaks, multiplier overclocking is still locked
>>52143955
>multiplier overclocking is still locked
You don't need it. On Non-K Bios bumping up BCLK will take you anywhere you want to go.
If you are fully in control of BCLK, the CPU multiplier is absolutely useless.
>>52142915
>tfw I'm still using dual xeons from lga1366
>>52143997
But how far can you take the BCLK and still be stable?
I heard Z170 uncoupled a lot of things from it in order to give more flexibility
>>52142514
>>52144016
And I'm still using a Q6600.
>>52144038
Either it's really time for a upgrade or you're still completely fine with core2 performance
>>52144038
Q6600 is second to Sandyvagina
>>52139582
if your CPU is pulling most of the weight for decoding, then mentally, you're definitely a candidate for buying yet another Intel rebrand
>>52139582
and my 4770K can decode it just fine
>>52144062
Performance is still alright for anything. Even GTA 5 runs fine with it, my HDD is currently the bottleneck for it
>>52144037
http://www.techspot.com/review/1108-intel-locked-skylake-cpu-bclk-overclocking/
They pushed an i3 to 4.7 GHz.
>>52142413
If graphene was the CPU game-changer it was promised to be back in 2010, then we'd have had graphene CPU's a lot sooner than now. There's a reason only Buzzfeed-tier tech sites proclaimed the MIRACLE MATERIAL GRAPHENE as the second coming of Jesus
>>52144037
>But how far can you take the BCLK and still be stable?
Allegedly a locked 6700 can overclock to about 4.6, which isn't that bad considering that most people get around 4.8 on core multipliers with base clock overclocking
>>52144037
>But how far can you take the BCLK and still be stable?
You can take it a really long way (the limit will be on the CPU stability, not motherboard stability). Basically the Z170 hardware uncoupled PCI-E and SATA from BCLK.
On regular bios, it forces PCI-E and SATA timings to follow BCLK even though the hardware doesn't mandate it (Intel spec does though). However, on the special Non-K bioses that manufactures have been gradually rolling out for the last few months the BCLK setting will only affect CPU and RAM, while PCI-E, SATA, etc will stay at 100 always.
If the bios lets you safely up the BCLK with freedom then the CPU multiplier doesn't matter at all. Even if your RAM sucks you can just lower the RAM multiplier as you bump up BCLK.
>>52144161
5 years may be a long time in tech, but it's certainly not a long time for something to be turned from theoretical curiousity to a practical application.
>>52138816
oh okay thanks bruh
>>52144161
Graphene is just hard to manufacture and manage at this stage, since the technology is still in its infancy. It needs to be more refined, but they promised the consumers that we'd be getting graphene-based products to mass market before 2020. I doubt that now at its current progress, since the impetus on graphene development seems to have leveled off with several graphene R&D companies going bankrupt since 2010.
dumb question: what is the technical reason we dont/cant seem to be going over the 4ghz speed?
i do remember people OCing to 5ghz+ and shit
why dont we ship cpus with a 6ghz stock core PLUS 6 cores or whatever
>>52143716
i got a 4670 without k
FUCK
>>52144204
>>52144223
Last I checked, graphene has no band gap, and as such, can't be switched into a off state, which kinda kills any hope of it being made into a desktop CPU
>>52144171
Interesting, thanks anon
>>52144274
You're retarded, fu?? off
>>52144274
The heat and power draw for CPU's like that are an insane sell from a marketing standpoint
http://www.anandtech.com/show/8316/amds-5-ghz-turbo-cpu-in-retail-the-fx9590-and-asrock-990fx-extreme9-review
And if Intel were to do it, you'd be looking at $2000 CPU's
>>52144274
Heat, power and quantum physics
>>52143997
>>52143909
>the CPU multiplier is absolutely useless
No it isn't. Changing the base clock still affects your memory, which means you have to spend time stabilizing your memory, and then dialing in your CAS latencies, otherwise you're just trading CPU overclock for RAM underclock. The CPU multiplier is still the best way to overclock the CPU. It always has been. It's not like raising the BCLK is a new idea.
>>52142401
Intel said they would be exploring several things, GAA transistors were one, SiGe was another. They mentioned they were exploring graphene as well, but thats a meaningless statement since everyone and their mother is. They're going to still be using a silicon substrate no matter what. A SiGe channel is still sitting on silicon, it won't be replaced anytime within this decade.
>>52144274
Its a matter of current leakage.
When a chip is fabbed the designer working with the foundry chooses a specific set of transistor libraries. These libraries are tailored for a sweet spot within a narrow range of frequencies and drive voltages in which the trade off between frequency and leakage is optimal. Example
Chip vendor A takes their design to the foundry
The foundry asks what level of performance the IC is targeting
Chip vendor says they want 3ghz within 50w
Foundry says RVT is the right choice for them, but notes low end LVT could hit that target in lower volume, but would have more OC headroom
Chip vendor must decide if they want the cheaper but more limited RVT run or the more expensive and versatile LVT
RVT could hit that 3ghz at 1.1v while consuming 50w, but trying to overclock it to even 3.5ghz might see power consumption spike to 80w.
LVT might hit 3ghz at 55-60w nominally with 1.15v, but could overclock to 4ghz while only consuming 65w.
Transistor libraries only give you so much wiggle room on a given process, so your ultimate limits are still confined, but this is the reason why we see clock speeds sitting around where they are. intel absolutely could ship a factor 5ghz chip if they wanted to. The process and transistor libraries would have to be tweaked to reduce leakage as much as possible at that high clock speed. They don't do it because its a lot of added cost.
There are also other issues that can arise with the design of the IC itself. Certain units on die may not function correctly outside of a certain range of frequencies.
>>52144663
>you have to spend time stabilizing your memory, and then dialing in your CAS latencies, otherwise you're just trading CPU overclock for RAM underclock
You can adjust the RAM multiplier accordingly with your higher BCLK.
> It's not like raising the BCLK is a new idea.
Sure, but this is the first time you can actually raise it by more than 2 to 5% since otherwise it absolutely messes the hell out of your PCI-E and SATA.
>>52145091
Nehalem had uncoupled BCLK, you could either go for high multiplier, low BCLK, low multipler, high BCLK or a mix of the two.
This is pretty much the same as FSB/multi OC'ing before FSB was phased out.
In most cases, a mix of increased multi/BCLK did better on voltages than just raising one, this was the same during the FSB era, but it still depends on the silicon.
>>52145091
>adjust RAM multiplier
Yes, that stabilizes the RAM, but then you're running the RAM at lower speeds with the same latency set. Your effective memory speed takes a huge hit from this. So, like I said, unless you take the time to dial in your CAS latencies, which takes just as much time as overclocking your CPU, and is trickier to stress test, you're trading CPU speed for memory speed.
>the first time you can actually raise it by more than 2 to 5%
Lol... people have been raising the base clock to overclock for a very long time, kiddo.
-e processor update when?
>>52145352
Broadwell-e is supposed to be coming out sometime next year. I'm not holding my breath.
>>52145359
I thought we were getting haswell-e q1 2016?
>>52145383
Haswell-e was released a long time ago. They launched Q3 2014.
Face it: Moore's law is kill.
10nm is years away.
7nm is a futuristic dream.
>>52145542
Maybe if we fixed x86 and stopped adding proprietary bload to CPUs
Who /sandy bridge/ here? i5 2500k at 4.5Ghz going strong. Literally no reason to even consider upgrading till cannonlake comes out in 2017.
>>52145542
We need InGaAs or some miracle with graphene.
>>52145596
>Literally no reason to even consider upgrading till cannonlake
unless AMD zen is really good this will be nothing more than a skylake die shrink
>>52145631
I can't wait for the same performance as Skylake, same power consumption, same amount of cores and cache, high cost but with a 90mm^2 die !
Thank you Intle
>>52145596
still rolling on a i3 sandy
actually not bad till this day
This isn't news, we knew Kaby Lake was a refresh.
Zen and Broadwell-E soon. I'd like an 8 core from AMD that beats a 4.4GHz 3570k in single core performance. Give me that and I'll buy. I wonder how AM4 motherboards will fair with all the Thunderbolt, NVMe, and soon to come out xPoint SSDs.
>>52143716
Wait are these going out of stock? Shit, I didn't build a new rig this year because I figured they would drop in price this time next year. FUCK.
>>52145739
Intel probably stopped producing 22nm parts.
>>52145063
thanks bruh pretty elaborate
>>52145703
>I wonder how AM4 motherboards will fair with all the Thunderbolt, NVMe, and soon to come out xPoint SSDs.
I think the AM4 chipsets were confirmed to have Thunderbolt 2 compatibility, but no word on if all the chipsets will have it. Hell, we don't even know if certain chipsets are necessary for FX CPUs and incompatible or feature-locked with APUs. It's far more likely that AMD will promote feature-rich, expensive chipset pairing with FX CPUs and cheaper chipsets for APUs.
>>52145739
>Intel CPUs
>Dropping in price
Intel CPUs never drop in price. Intel stuff from 3 gens ago goes for as much as the newest gen. Intel keeps the prices fixed because if they didn't no one would buy the new stuff with it's lackluster 3% performance gain over the previous gen and instead just buy the discounted old stuff.
>>52145739
Devil's Canyon are going fast too. My Micro Center has a handful of 4790ks left and they will not be getting shipments of those CPUs anymore.
>>52145739
get a 4790k before they stop producing them lol
im getting a 4690k probably on monday before they run out for real
>>52145911
One of the computers I built for a client had a really good 4790k, one of the best samples I've seen.
4.9GHz on air and minimal voltage increase. I'm pretty sure I could have gotten it to 5.0 stable with a bit of time and adjusting.
>JUST bought 16GB DDR3
>Haswell-E and above only support DDR4
Feels cucked by Intel
Come on AMD
>>52145911
1231v3 much better price/performance
>>52146151
Socket AM4 only supports DDR4.
Why would you expect anything else?
>>52146151
GET A LOAD OF THIS KEKED FAGGOT
AMD ALSO GOING DDR4 WITH AM4
>>52146192
>>52146317
I did NOT expect double penetration, did expected it though
Doesn't matter, going to grab a second hand ivy bridge-e or haswell
>>52146151
how do you even make that mistake senpai. my vengeance LPX's have 2400 and 2666 xmp profiles so i'm so i'm set till DDR5.
>>52145091
>Sure, but this is the first time you can actually raise it by more than 2 to 5%
I realize you're a child, but people have been overclocking via bclck for longer than you've been alive, and certainly more than 2-5%. My first experience was with a Pentium II 300 MHz. Intel literally just took the same exact bin of 450 MHz Pentium II's and set the multiplier lower. So everyone just pumped up the bclk and got a free 50% CPU overclock with absolutely zero percent chance of getting a "dud" CPU that couldn't do it.
For many years this was the only way to overclock anyway.
And the enthusiast community has been doing bclk overclocking before even that.
bclk overclocking is as old as overclocking itself.
There's a reason why multiplier overclocking became so popular once there was hardware to support it: you just get better results, all the while being easier to do.
