Wishing for proper intel GPU
Why? Most of their iGPUs are dogshit. Also that's not a gpu, it's a co-processor.
>>51982010
Because they're consumer level iGPUs and as far as theyre concerned. If you qant better, you will get better.
Also in pretty sire OP knows that his pic isn't a gpu hence this thread.
>>51982010
>Most of their iGPUs are dogshit.
They are good for what they are and iris is actually a good entry level gpu
>Why?
Compretition on the market is much needed and amd alone is not enough to wrestle with nvidia. Also, intel usually works best of them all with lunix
>it's a co-processor
Yes it is
That looks gross.
>>51982169
The GPU is a coprocessor in general, intel just integrates it into the DIE of their CPUs.
>>51982201
0/10 bait
>>51981975
kek, intel invests in the future.
The future is iGPUs.
Intel has the luxury of not getting into this AMD vs Nvidia kiddy war because the only people who cares are kiddy gamers these days that nobody of significance takes seriously
I've always thought a proper integrated GPU is the way to go. AMD's APU lineup is fantastic, but lacking in power compared to Iris Pro. AMD should just jam in 256 MB of dedicated RAM on their APUs and call it a day. Costs next to nothing yet it improves their products dramatically.
The real question is why AMD isn't even trying anymore. This is such a simple solution even a 4chan shitposter like me can come up with, I'm sure the engineers working there have figured this out years ago.
I'd get an APU with dedicated RAM instead of shared memory in an instant for my mid-range and budget builds.
>>51982377
Zen APUs with dedicated HBM cache should be very interesting.
>>51982377
>Costs next to nothing yet it improves their products dramatically.
eDRAM is not cheap.
>>51982413
Didn't you hear? A random pleb on an anonymous image board says its cheap and therefor it must be!
How dare you suggest the enormous slab of eDRAM on iris pro is expensive which is why its only on cpus that cost 1000 shekels or so.
Aight /g/ - Consumer Tech, purely hypothetical question since I'm a poorfag and nowhere near buying new hardware.
Say I wanted to make a general purpose computer for Ganoo/Loonatics with the most graphical power i can get BUT without a dedicated GPU.
And since intel is the only one making open sores drivers for their igpu I guess I should pick intel for the "just werks" experience.
Which one is the CPU with the most powerful iGPU on the market?
What kind of experience can I expect (say in videogames, since it's comparable, my current hardware sucks too much to let me know what kind of power i need for those advanced mpv functionalities and filters).
Or If I'm wrong, how's AMD's APU support in Linux?
Some time ago I've heard it generally sucks dicks.
>>51982515
HD 4600 graphics i guess
>>51982377
I think a discrete GPU is fine for graphics tasks that have a one-way flow of data from the system to the card to the display, so that you don't have to have space for TMUs or ROPs on your CPU die, but an APU is definitely ideal for compute stuff.
>>51982515
An i5-5675C or i7-5775C has the Iris Pro iGPU on there, so those will deliver the best performance without a dedicated GPU.
>>51981975
Did you know that Intel bought this from Lockheed Martin?
>>51982010
>implying you need a better gpu for web browsing and streaming
>>51982515
Broadwell i5/i7 have the most powerful iGPUs out right now, BUT you'd have to stick with an H/Z97 board because Broadwells are LGA1150 only. It's about 1.25% the power of an A10-7870k, but running at a much lower TDP.
I still wouldn't use one for an HTPC because
a) it costs too much and
b) the TDP is still too high to use in a silent HTPC case
A low power 65W TDP APU is best for HTPCs.
>>51982919
Both of those things heavily rely on the CPU you dipshit. Hardware acceleration isn't as widespread as people want you to believe.
>>51982397
It would also be expensive as hell.
Would you spend $300+ on a quad-core Zen APU or $150 on a quad-core i3 with a dedicated $150 graphics card?
The price-to-performance justification just isn't there, even with an HBM APU.
>>51982980
>quad-core i3
I meant dual-core, fuck
>>51982034
>it can't run Crysis 3 at 4k120fps so it's dosgshit
Intel's integrated graphics is not meant to be used for gayman, you dumb shitter
>>51982980
>>51982995
It's 2015, you need a quad core at minimum.
So assuming Zen is reasonably priced, it's very likely to be cheaper than getting an intel quad + GPU.
>>51983048
HBM won't be available for Zen because of costs. It would easily bump the price of the APU to above Core i7 prices. Shit ain't cheap.
What would be the dedicated graphics equivalent of the current most powerful iGPU?
>>51983099
You run into the issue that gpus with the right amount of grunt are so old they lack modern api support.
>>51983048
Most i3's have 4 threads so they are really just quad "lite" as programs are having more support for hyper threading , plus much better architecture(currently) vs amd's
>>51983099
The Iris Pro 6200 on Broadwell i5/i7 is roughly as powerful as a GTX 750 or just an R7 250X
>just mindlessly claiming HBM is expensive
>when you literally do not have a single shred of proof
You're all utter dipshits.
>>51983120
Would CPU performance suffer if the iGPU is working full load, or is it able to handle both things without issue?
>>51983069
Except it's been confirmed that there's 100% going to be a HBM equipped Zen CPU in 2017.
>>51983137
It is hard to make, it heats up like toaster, yields are mediocre.
>>51983137
it's got to be expensive or I would buy it for my gaymen machine!
>>51983156
No, but theoretically it sort of kind of can throttle CPU a bit but not really, die is very small it cools pretty well.
>>51983137
It's expensive because of there is a short supply, costs roughly 3-5 times as much as GDDR5 NAND (according to SK Hynix), and the integration process has not matured enough to reduce the cost of the interposer and assembly. Plus, integrated eDRAM is cheaper and faster, if only restricted to die size.
>>51983157
No, that's a server-grade APU. It's rumored to have 16 cores too, so fuck the affordable price range.
>>51981975
OP missed it the first time around eh?
>Real 3D Starfighter, Intel i740 based video card
>One of only 2 consumer level graphics chips at the time that had a full OpenGL ICD
>3DLabs Permedia2 was the other
>mfw I had both back then.
So if seriously, how much intel would have to spend to make one proper GPU line? 150b? 200b?
>>51983186
Off the top of my head Intel lets the cpu run out of spec when you slam both the cpu and igpu. AMD's apus will aggressively clock both depending on thermal and TDP headroom.
>>51983195
It's still going to be a thing, and if it's HBM1 and the price has fallen enough they might end up taking the shittiest ones and binning them down to 8 cores, put them on the consumer market.
>>51983233
Depends on what you mean by proper.
If they took 2x of the Iris Pro 6200 equipped CPUs with fucked cores and threw them on a single board with 2gb of DDR4, clocked them up to 1200mhz, that'd be enough power for 1080p@60 on medium/high in most games.
>>51983156
At present all APUs have some form of hard throttling, and it can't be turned off by the user. If you buy a desktop Kaveri A10 7850k and overclock it to 4.4ghz it will not stay at that clock speed when CPU and GPU are stressed. Hit the IGP hard and the CPU cores throttle down, with the 7850k this is typically about 3ghz.
The chips are primarily designed as mobile and for whatever reason AMD didn't bake in any way to disable this behavior in desktop chips. This isn't happening from the chip hitting any thermal limits, its just programmed in behavior.
In a desktop there isn't a need for this to happen, but it'll likely remain this way for Raven Ridge unless AMD changes things.
>>51983157
No it hasn't. AMD hasn't confirmed anything specific about Raven Ridge in their internal roadmap aside from its projected release date.
>>51983195
>>51983168
>dipshits still just talking about of their asses
Cost is a relative thing, you shit eating children. If you want to make a clear factual statement then you need to have the command of actual facts, not assumptions that you pull from thin air because you *think* it sounds right.
HBM is the only way to continue to push on IGP performance, and AMD absolutely needs to, they know they do. The 8CU in Kaveri and Carrizo are still majorly bottlenecked to a degree that not even high speed DDR4 can remedy. Adding more CU without providing them adequate bandwidth is just a waste of die area and power consumption. They'd only need a single HBM module on package and IGP performance could jump well over 100% in a single generation. The cost to performance ratio makes it a no brainer no matter what the raw cost of that single module is. AMD will have higher margins on their up coming 14nm parts than they currently have on their 28nm parts so even mentioning cost as being prohibitive to include HBM makes you sound too fucking stupid to even be posting here. Get fucked, tech illiterate dipshits.
>>51983246
>It's still going to be a thing
Are you going to spend $1000+ for an APU just to play games at sub-1080p resolutions?
>HBM1
HBM1 is going to be completely supplanted by HBM2 because it's more expensive, has limited bandwidth, has lower capacity per die, and has lower yield than HBM2. HBM1 is a transitional product, once HBM2 comes into market they will forgo HBM1 altogether since only one product range uses HBM1 at the moment.
>>51982810
based Lockheed
>>51983315
>Adding more CU without providing them adequate bandwidth is just a waste of die area and power consumption
Funnily enough you can see the reverse of this on the Fiji gpus - despite enormous bandwidth and 45% more shaders the fury x has the same front end as hawaii based cards which was (partly) responsible for the cards horrific 1080p performance early on.
>>51983300
>If they took 2x of the Iris Pro 6200 equipped CPUs with fucked cores and threw them on a single board
Well pretty much only that would cost them a lot of money with unknown returns. If they would want to do proper R&D price will skyrocket.
Basically i'm trying to say they will never even try to make GPUs.
>>51983399
If they ever want to enter the GPU world they'll just buy something, possibly grab adreno off qualcomm. Don't know how interested they are in nvidia
>>51983315
>Cost is a relative thing
Nope, cost is and always will be a primary factor in consumer purchases. Would you rather buy an APU+HBM that's as strong as a current R7 370? Or a $150 quad-core CPU+R9 480/GTX 1060 for roughly the same price?
HBM will NOT appear in a consumer APU for some time, at least not until HBM modules become significantly cheaper to produce. The reason why AMD is only considering HBM for server-grade APU is because they are the only market willing to pay for the technology, whereas the average consumer will not fork over a premium for technology that will go unused in favor of discreet graphics cards. Plus the integration of HBM to the APU would require a whole redesign of the substrate that the die would sit on, since a dedicated interposer would be needed to connect the two. That would add development cost that might never be recovered from sales revenue over the lifespan of the APU.
Even with HBM, an APU is only as strong as the iGPU. An R5 GPU is not a cost-efficient pairing with HBM and it's unlikely that the next generation of APUs will have an iGPU stronger than an R7 370 due to die size.
What will happen is that iGPUs will get a significant boost compared to Kaveri due to faster speed DDR4 and could help performance when used in conjunction with a discreet GPU thanks to the proliferation of lower-level API combined with HSA. The transition to DDR4 memory will aid the Excavator-based APU launching this spring to maintain parity with the Iris Pro 6200 on the i5-5675C.
>>51983469
>possibly grab adreno off qualcomm
so basically they'd be buying amd assets
>>51983499
I'm going to just make things up: the post
You have to have a mental disorder to be this delusional. Making bold claims about what a company is doing behind the scenes and what they want to do when they've made absolutely no public comment.
>AMD is doing X because I said so!!!!!!!!!!!!!!!!
>>51983499
Feel free to explain why an APU with HBM costs the same as a $150 CPU + $250 GPU
>>51983502
They'd already own AMD if they didn't think they'd get fucked for being a monopoly.
>>51983556
He saw someone on the Anandtech forums claim HBM was super expensive without any source whatsoever, so hes just mindlessly regurgitating that like a good little sheep dipshit.
>I have no idea how much it actually costs
>b-but its probably a lot!
>>51982010
Here comes the fighter for justice which doesn't know what iGPU is for.
I'm expecting a rebuttal of showing us how amazing APU is on the AMD side which performs slightly better than 8800 with ricer RAM that cost more than the CPU itself.
>>51983532
Can't tell if this is bait
You clearly have no idea how difficult it can be to integrate new technologies into consumer-grade products. You can't always stick two things together and expect them to work flawlessly. It takes time and money, which AMD does not have in time for the Zen launch. We will not be seeing HBM APUs for some time as a result. Anyone who believes otherwise is setting themselves up for disappointment.
>>51983556
That's the current cost projection from SK Hynix, one of the main producers of HBM nodes. They specifically stated that HBM2 is projected to be 3-5 times the cost of GDDR5 NAND, which itself is about twice the cost of DDR3 NAND. Counting the development cost of designing and producing an interposer to go along with the HBM would add cost, but how much it would add to the price of the end product is up to AMD. I'm guessing it would retail above $300 because AMD's APUs is hinted at launching at a higher price post-Carrizo.
>>51982201
GPUs are accelerators, anon-chan :3
>>51983678
You're positing your assumptions as fact. This is the behavior of someone who is totally delusional. Certifiably mentally ill, detached from reality.
You're making blatant claims as to what AMD is doing despite AMD having made no such statements. When did AMD claim they were only pursuing HBM for enterprise grade APUs? Literally nowhere, thats where. They made so such statement at all, this is an invention of your deranged autistic mind, and you're too much of a childish dipshit to stay within the realm of facts.
Interposers are not expensive. AMD themselves have no technical hurdles to overcome in getting their fab to deliver a 14nm package qualified for use with on interposer memory. GloFo qualified a 14nm test vehicle for this last year in fact, they have a technical video covering it.
Stop spreading your mentally ill fantasies as if its confirmed fact, you broken pathetic shitheap.
>>51983616
You could:
A: pay $80 for an A8-7600
or
B: pay ~$200 for an broadwell i3 with an inferior iGPU and around the same multi-core performance.
Intel's iGPUs are dogshit because of how overpriced they are. The only good ones are in i5 broadwells which cost around $300.
If you don't play vydia gaems then AMD would be the smarter choice here.
>>51983899
Desktop Broadwell SKUs aren't in production any longer, and no one is getting new stock.
>>51983951
Amazon claims to have them albeit at batshit high prices lel
>>51982515
>Which one is the CPU with the most powerful iGPU on the market?
https://www.youtube.com/channel/UCXIuoZK11s9jC_L27KgN5TQ
>>51983899
*haswell i3
>>51984028
Prices are going up as existing stock dwindles. Once they're out of stock they're going to remain out. Newegg has had them perma listed as out of stock for a very long time which has been the case with a lot of other retailers as well.
>>51981975
That's my literal nightmare.
>>51984136
So much for beating the a8 apus. Now all the igpus intel has are dogshit since that a8 apu beats them it for under 100 bucks.
>>51983899
Again you didn't mention the price of the RAM.
AMD iGPU are better but the question is for what really?
They are insufficient enough for gayman (high end mind you nobody cares for console like experience) but for standard multimedia they are not excelling because bar has been set pretty low.
I like APU but most consumers who do office work etc. are more likely not even utilize the power of the APU with some shitty RAM that is insufficient for the task and price to performance ration doesn't scale that good.
>>51984347
>Again you didn't mention the price of the RAM.
He doesn't need to. 2400mhz 8GB kits are cheap. A Patriot Viper III kit is $65.
>>51984347
>Again you didn't mention the price of the RAM.
I really wanted to save you the embarrassment of showing you pic related.
>AMD iGPU are better but the question is for what really?
Faster hardware acceleration for a cheaper price.
>They are insufficient enough for gayman (high end mind you nobody cares for console like experience) but for standard multimedia they are not excelling because bar has been set pretty low.
wut...
>I like APU but most consumers who do office work etc. are more likely not even utilize the power of the APU with some shitty RAM that is insufficient for the task and price to performance ration doesn't scale that good.
An AMD APU would still be a better choice because it offers the same computing performance for less. Also there are many applications outside of gayming that can utilize the iGPU.
The point is price. An A8 APU has better iGPU performance and the same multi-core performance as a ~$200 i3 intel CPU.
>>51983773
>When did AMD claim they were only pursuing HBM for enterprise grade APUs?
This August during an interview with Lisa Suu about AMD's product pipeline, which confirmed earlier rumors from May that HBM would not be included in consumer APUs. When did AMD say that they were pursuing HBM for their consumer APUs? Nowhere.
>Interposers are not expensive
AMD could recycle the design of the interposer from Fiji, but they would still need to change a significant portion of the design to make it work with an APU. Plus, it would be the first design to include both a GPU and CPU together on the same die. This could introduce new problems or possibilities for the interposer design, which will take time and money to finalize. It's more expensive and time consuming than you think, my naive friend. AMD can not afford a delay in their Zen APU launch, so it's 99.99999% likely that they will not bother with HBM integration before 2018, if at all. There is no indication from SK Hynix or Samsung that their HBM production costs will go down before 2017, even by their own optimistic estimates.
>mentally ill fantasies
You're the one living in a fantasy land where everything new is cheap and plentiful because it's new. You're forgetting that new technologies always carry price premiums if they are not based off of existing technologies.
>>51984522
>just making more shit up and blatantly lying
Pathetic, mentally broken, autistic child faggot.
>>51984347
>They are insufficient enough for gayman
They're decent at the most popular games in the world, MOBAs and CS. You could play those game at 1080p at 60+ FPS on an A8. That's important for most gamers on a budget because they don't need to pay more for a discreet card.
>but for standard multimedia they are not excelling because bar has been set pretty low.
Nigga you jest. A Pentium iGPU is good enough for smooth 1080p playback and ample enough for encoding. An R5 iGPU is much stronger and much more adept than that.
You don't need a 2400MHz RAM to benefit from the iGPU, a 1600MHz RAM kit will be enough for decent, higher-than-Intel-HD-iGPU performance. Faster RAM only helps with performance.
>>51983218
>intel launches a video card to show off the new AGP slot
>PCI version performs better
If you didn't have 3dfx at that time you were best off running everything in software
>>51984550
All you can do is throw childish insults and make up your own detached-from-reality optimistic claims with absolutely zero backing or evidence.
Tell me, when did AMD say that HBM APUs were coming to consumers before 2018?
>>51982169
>iris is a good entry level gpu
Have fun paying +400 for a cpu with it
Meanwhile you can get a apu that performs MMMUUUUCCCHHH better for around $120
>>51984618
>the mentally ill child faggot is still going
You're quite literally posting complete fabrications as if they're factual now. Lisa Su made no such claims as you're stating. I made no positive claim to the opposite, I don't have to defend any position whatsoever. I've done nothing but call out your factually devoid bullshit that you're dedicated to spewing endlessly. You're mentally ill, and I'm not responding to any of your utter fantasy bullshit further.
Do the world a favor and slash your wrists open instead of shitting up this board with nonsense pulled from your ass, or at least stick to some shitty forum where it belongs. You'd fit right in with the LinusTechTips crowd.
>>51984732
>I'm not responding to any of your utter fantasy bullshit further
Same here
Enjoy your permanent HBM blue-balls, you cum-addicted faggot
>>51982810
Such a piece of shit GPU, here's what it looked like running Unreal Tournament in Direct3D mode
https://www.youtube.com/watch?v=XDxjuXikgs8
>>51984732
Good post, faggot.
>>51984729
Not really, Iris 6200 Pro outperforms all APUs and even some relatively usable dedi-GPUs, such as HD*770s
>>51984852
Nostalgiagasm
>>51982201
Oh fuck off and get a basic introduction into HPC
>>51984879
You know broadwell CPUs are either sold out or have an inflated pricetag right?
>>51984347
RAM prices dropped significantly brah
>>51984924
Yes, I didn't mention current retail prices.
>>51984729
my entire point of this thread was best possible linux support really, so yes I'm willing to pay moar.
>Meanwhile you can get a apu that performs MMMUUUUCCCHHH better for around $120
I do respect AMD, even moar than intlel, but intel is better in cpu department and better in gpu department (tho it come at the price, but i already stated that im willing to pay for it)
>pic related
>>51984604
intel has shady business-tactics, but their hardware-fuckups are always hilarious.
>>51982267
Yes, anon, Intel is smart enough to care about the microscopic Linux desktop community over millions of PC gamers. Wow, what vision!
>>51984879
At that price point, its better just to buy a gpu and a cpu separately
>>51984984
>my entire point of this thread was best possible linux support really, so yes I'm willing to pay moar.
So you're a dumbass? Got it. (not same anon)
>>51984984
>playing bioshock
>urinating on your ass
Okay. -_-
>>51984984
So this is what being a freeturd is like? *shudders*
>>51985034
No, the question is about functionality.
If you weren't a gamer kiddie, you'd understand.
>>51984816
>>51984732
This is what I call shill convergence; the argument is long and complicated enough that no one can tell who is the AMD shill and who is the Intel shill.
>>51984994
When you consider that their big customers with data centers to build tend to use lunix for whatever reason (as lunix is garbage) then it starts to make some sort of sense actually
>>51985022
>graph to back my point with data
>>51985088
It has nothing to do with AMD v intel.
I made it explicitly clear that AMD's Raven Ridge APUs hadn't been confirmed to be using HBM. The other anon simply started a tirade of shitposting and pretending like all of his stupid assumptions were actually truthful facts. He blatantly lied about something Lisa Su never said.
The signal to noise ration on this board is just fucking terrible any more.
>>51985069
>runs open sores operating system
>pays out the ass for a proprietary CPU
You sure showed those them anon
>>51985128
Doesn't matter, I'd choose Linux even if it was closed source.
>>51984984
>Not pictured: 7870k.
Remember when you could browse /g/ and every discussion would not center around "FPS per game"? Good times.
>>51985177
But that's literally a 7850k with a slight clock speed bump. There's no way it'd catch up with the Iris Pro because of that.
>>51985128
Actually, you can make 200nm sparc open source CPU at home with proper tools(not that expensive, just about two used cars), it's arch. is open source since 2008.
I have no idea how would you make a proper package for it though.
>>51985177
There wouldn't be any real difference. Godavari is just a minor clock bump over Kaveri, and both of them have the same throttling issues.
Anandtech and a few other places actually showed Godavari having a clear performance regression which was caused by VRM throttling from vcore defaulting way too high for the chip in their test boards.
http://www.anandtech.com/bench/product/1200?vs=1497
>>51984984
>Cheapest Intel on that list is $300
>Most expensive AMD is $130
That's not exactly an apples to apples comparison is it? Because if it was, you'd spare the AMD's a GPU worth at least $170, which should cover for something like a R9 380.
>>51985314
the $170 covers powering the amd cpu for a year or two.
>>51985177
>7870k
Yes, it's not THAT far off and half the price, but still it's all about functionality (read drivers)
>>51985191
no one with total synapses count greater than 10 cares about fps ITT. The graphs are here only to picture performance differences, sadly with games
>>51985351
The age old housefires meme, eh?
Seriously thou, the price difference there is pretty huge so I really don't think the comparison really holds up seeing how the cheapest Intel there costs more than twice as much as the most expensive AMD and the i7 is $430 on Amazon.
Gr8 b8 m8 I r8 8/8
>>51984347
Broseph, there are a lot of games that will only do 30-35FPS on an Intel HD on low 1280x720 that will run at 45-50FPS off an AMD APU.
I'm not talking 'gaming' games here, nobody gives a shit. I'm talking fun indie things things, sims 4, even browser games. All things normies need and which i expect my fucking $500 laptop in 2015 to be capable of doing
>>51985452
Those Broadwell chips were 65w parts, top end Kaveri is 95. Those 30 watts wouldn't add up to anything more than chump change even after half a decade.
The kicker is that is 14nm Trigate vs 28nm planar, Kaveri is memory bottlenecked and would perform better with fast enough memory, but Broadwell with its eDRAM still has a bigger combined die size.
AMD's latest chips are insanely efficient for the node they're on, and the Carrizo chips are even more so, and they're made on a cheaper generic 28nm HPP instead of the custom SHP.
>>51985124
>can't stand that I'm right and he's wrong
top kek
And it goes on
Google, nigger
>>51985235
is the chipset also open? the memory controller and tech? likely not. lost cause
