Intel's Future Chips: News, Rumours & Reviews

Page 11 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Status
Not open for further replies.


The other APU affect is tier based systems. AMD are effectively removing "Entry Level" components with Graphichs on Chip, if Kaveri is best case scenario then it will likely mean that in accordance with rumours HD8700 family may be scraped with just the 8600's for hybrid crossfire support and the 8800 family being mainstream while the 8900 family high end. That will mean less SKU's and less process.

I think AMD will use BGA for ALL-IN -ONE's which imho are not very market conducive but then again not many people will upgrade HTPC/Media systems.

And yes Intel got caught very cold and it makes you wonder whether APU was on their mind for Haswell at the expense of node space, power savings and obvious process cost.

 


That was my reaction as well. It's nice to see, but like anything else that has more features (like cars for example) the higher the likelihood of something to break. In this case Intel can blame the end user and force them to repurchase a product or go elsewhere.

 
Its Intel's GT3 iGPU the thing that seems to only be an enigma. Its been talked about for over a year and nothing. What we have heard is that it plays Skyrim at 1080 ultra settings and is faster than a GTX650M on mobility, both I have my severe doubts about after the HD4600 revealed itself to be more of a HD4000 side grade on a new socket.

It is times like this you realize how dependant Intel is on AMD and Nvidia, if the latter decided to sabotage GPU drivers to not work on Intel platforms, everyone will abandon ship.
 

hd4600 is a sidegrade to hd4000. are you under some kind of delusion that just 4 more eus significantly increase performance?

that's plain dumb. nvidia would be shooting itself in the foot if they try to lock out intel. same for amd, even though they make x86 cpus, they will lose huge amount of revenue as the majority (by marketshare) of x86 systems are powered by intel. it'd be different if amd offered strong competition (they don't) or if intel didn't have near-monopolistic hold on x86 market (they do).
 


Yes the latter won't happen but how many people will keep their Intel's if AMD and Nvidia locked them out, that means absolutely no gaming as Intel graphics is nowhere near mainstream ready. So that will mean you would have to sell your Intel's and buy AMD x86 processors then you can use AMD and Nvidia GPU's and the joys that come with that. In short the market will change fast because the very large majority still require the GPU more than CPU brands.

As to the first part, the article last week with the i5 put through 3D mark benches failed to even outscore HD4000 at 3ghz and I don't really thing 200-300mhz makes a huge difference. GT3 is the be all and end all, if it fails to impress it puts Intel into a hard place, to scale performance they need to congest the node with GPU components, Intel want smaller nodes and thats going to make staking up a lot harder, then there is the power element. I guess making a GPU was a little harder than anticipated. GT3 has to blow Trinity away to be a success, already 60% higher clocks and integrated DRAM which is not cheap, so on that basis CW is very expensive and very resource dense which is not sustainable if you want to lower power. I do expect GT3 to be within 10% or so of Devastator but the costs are unfeasible and its not a accurate reflection. You saw my LN2 Trinity score at 1300mhz, it will blow GT3 to smithereens at those clocks.

 

no.
investors will likely fire the entire management if either company does that, then re-establish relations with intel after paying hefty 'compensations'(read: intel actually ends up making money from the lockout).
assuming both amd and nvidia entirely lock out intel: amd goes bankrupt the same quarter they announce the lockout (because intel will retaliate by rescinding x86 license, anti-competiiton goes both ways 😀), nvidia follows several quarters later after everyone skips tegra for exynos or snapdragon socs.
right now money is in mobile and intel has a significant share in laptops, most of which doesn't even have discreet solution. apus are stuck in niches, one of which is dying fast(netbooks) and the other is being increasingly threatened by intel's improving mobile solutions.
neither amd nor nvidia have strong presence in mobile platform. nvidia doesn't have cpus and almost no one knows about amd's.
mobile(laptop) gaming is taking off and both intel and nvidia has carved out their own presence there. amd's mobile cpus/apus severely bottleneck their own discreet gpus - that comes much later in the long line of drawbacks like poor power management, immature dynamic gfx switching and poor cfx/dual gfx support, mobile gpu driver support, power hungry mobile discreet gpus, lack of mobile apu skus and so on.

rofl. and what kind of laptop is gonna house your ln2 powered trinity rig? desktop market is shrinking and unfortunately, apus don't have a place in there unless they offer 'true' discreet class igpu performance. intel and amd don't offer that. rising ddr3 prices is another reason (one of many).
intel's gt3e sku has 65w tdp on desktop while trinity and richland will have 100w at least. i'd say intel has perf/watt down, and that's part of the reason why they'll charge high prices. less heat - easier to build sff pcs. that's the reason why ivb+kepler combo has been so successful. for example, you won't see an fx6300+radeon 7950 in an sff pc anywhere.
 


1) GT3 will be made available for desktop.

2) AMD and Nvidia would not operate anti competitively so the point is moot.

3) Richland mobile is out, touted synthetic rumors of at least 50% faster than HD4000 mobility, that coincides with GT3 and around 2.5x faster than the fastest current Trinity mobile part. The bigger talking point is the touted 20% power efficiency bake ons AMD has incorporated, talk of 1-2 hours of peak load time improvements. Up until recent AMD hasn't had the mobile part to be competitive but these Kabini parts may well be very feasible and affordable alternatives.

4) Intel and mobility in terms of Cellular market is still a pipe dream, this to go with the fact that Broadwell may only be out in mid 2015 according to toms not that long ago article so Intel and cellular process is just talk, then the second issue, Qualcomm, Nvidia, Samsung, ARM etc are all major players and just because Intel is there doesn't mean the option is good.

5) Well there are existing Trinity SFF systems, Mobo, Chip and RAM is all you need for a system. Please tell me how difficult that is? As we speak I am building a prodigy with a A85 ITX board and waiting on my Richland based part and a $60 kit of DDR3 2133, this is competent enough to play BF3 at 1680x1050, Max out F1 2012 not to mention a few other things it does exceptionally well. So you take a 65w TPD GT3 which will probably cost around $500-600 and it will get its rear end handed to it by a 65w Trinity part. Superior x86, A monster of a iGPU to sort of compete with Trinity which will drop in price. Richland we know has new power features, probably a more active RCM, 22-40% iGPU gain as was touted and delivered by benches and x86 improvements all within that $120-130 mark. To me it makes it hard to justify a GT3 BGA and GT2 is pants.

6) Doesnt' have a place, you are sounding like Chris now. Here is me with my APU setup, put about every GPU combination including a Titan with it, running a CF 6850 combo I picked up for $175 for both cards, plays BF3 on Ultra around 90 FPS, don't seem to run into this microstutter deluxe crapstorm that has been cooked up, sure a little here an there but not that much different to what my 560ti's did. Anyways longs story short, a APU with a 650ti boost or 7790 is competent enough to play crysis 3 on high settings, most other games maxed and cost next to nothing in the process... Balance maximum performance and lowest cost with a lot of other features cheaping out doesn't offer.



 
I dont think your numbers are quite right for Richland or gt3e.
And Richland is nothing more than a filler, not the future.

To reach its potential, or to be competitive, gt3e cn be oceed heavily, and wont stay within certain specs doing so, and even then will it beat out the non oceed AMD alternative? It will bunch perf/watt closer, and its going against a non long term solution to do so.
 


GT3 is unknown so speculative, some say its iGPU clocks are locked to 1300mhz and 1350mhz boost, you know that even a Intel part at those clocks becomes very thirsty not to mention the heat it must be passing off. its also BGA for DT which is probably locked parts so no overclocking. When I say overclocking the 5800K to 1300mhz it is really to show that intels gains are unsustainable clock bumps not architectural. The kicker is the price, those deluded enough to think this will retail at $200 a pop need to times that another 2x and at $600 discrete based systems are better options, so if anything doesn't have a place in DT its this.

Richland has benches out, the benches were future mark benches which represent worst case scenario. AMD did expressly tout 22-40%, going on the benches the non clockable 6700 65w part was on average 25% faster than the 5800K, that is in line with worst case and the specified lower limits. Richland is by no means a filler, it represent more than just x86 and iGPU gains, it features new AMD power control features so while its probably a low key relative to Kaveri it is a lot faster than Trinity, the best part is the iGPU is only clocked 44mhz higher which rules out iGPU overclocking.



 
What I am saying is, its lifetime wont be very long lived.
If it had come in the roadmap timing, it would have been different, now, it resembles more an E series Intel, towards the end, with new and shiny coming soon.
This was AMDs fault it will be looked at this way, and is why I said as much
 

i just said the same thing in my earlier post. that's the one i was using for comparing other dt apus.

:lol: anti-competition is what you've been proposing all along. you refuted your own 'logic'.
you forgot that intel owns x86 license and can effectively leverage that against both amd and nvidia. that's the one reasons why amd and nvidia still do business with intel even after being repeatedly screwed over by them.
rumored numbers are not worth discussing about even though they seem way off.
amd themselves have stated that they will use kabini for high end tablet space. high end isn't exactly affordable.

weren't we discussing (before you hilariously contradicted yourself, again :lol:) amd and nvidia locking intel out? anywho, you don't have to go to the future or boradwell, intel already sells x86 socs for smartphones. those are nowhere near arm marketshares but significantly higher than amd's, which is zero. same with tablets and tablet pcs (where amd is very near zero).
also keep in mind that microsot's windows 8 has failed so far, so amd doesn't have a strong, high-selling os platform to build on...
it's not that difficult to build. more difficult to live with the performance and power use in that form factor, especially with unlocked apus.
i assume(guesstimating using current prices as reference, could be wrong) desktop core i5+gt3 will cost around $200-230 or more and around $350+ for ultrabooks. $500+ range is usually reserved for high perf mobile quad core i7 cpus. kinda looks like you're fearmongering.
i am not gonna discuss gt3 vs richland without seeing benches. i am only aware of tdp and bga packaging, nothing else.
i can't find a $60 ddr3 2133 kit on newegg. i checked. the cheapest costs $74. ddr3 prices will likely keep rising. price increase is bad for cheap, entry level pcs.
we don't know anything about rcm. amd never confirmed anything.
bga is normal for sff pcs, even amd has had bga solutions for years.
if you're gonna guess performance - intel will sell core i5 and i7 gt3 skus for desktops which have partially unlocked turbo multipliers and (may be) unlocked igpus. vendors like asus and gigabyte (asrock too) can take advantage of that and use multicore enhancement to max out turbo oc while making igpu oc available. combo that with a low-profile down-draft cooler from noctua or silverstone, you have a capable pc, that can at least wipe the floor with similar tdp amd apus/cpus in terms of cpu performance and can handle powerful discreet gfx card (most likely kepler or kepler 1.1) in sff And put out less heat than trinity/richland's 100w/65w apus.

amd is already suffering from lowered revenue because of shrinking dt market, windows 8 made it even worse (weren't amd fanboys the ones who kept praying for windows 8 to salvage bd performance? win 8 at least has that disaster factor common with bd. 😛)
6850 cfx actually microstutters much more than gcn cards, toms' testing showed that long before nvidia released fcat for public. since then drivers have improved but driver performance doesnt benefit older, lower midrange cards that much. it's bad when crude tools like fraps can catch it... i am not amazed that you don't notice anything since you have claimed repeatedly in the past that you are incapable of noticing fps differences and stuttering.
as for your claim about apu playing crysis 3 on high - that seems to be incorrect as well. toms' own benches show how miserably apus perform in crysis in cpu benches and high settings @1080p needs at least a 7870 2gb or gtx660.
 
Lets make this clearer, one company has billions of assets, with billions tied exclusively into R&D into those assets, where those assets need to be used to pay for themselves as well as future R&D thru selling DT, so yes, AMD has been hurt, but they arent the only one
 
The time THG did the 6850 review was prior to catalyst 12.1 was out dated subsequent drivers had improved the stuttering, I did say it stutters and glitches but so does SLI, SLI is just more consistent.

Crysis can be played on high, not ultra which is system taxing, like Metro 2033 high is a breeze ultra brings Titans and a Intel Extreme to its knees.

Windows 8 like all new OS will take time to feel in, already it is better with updates and it does have a nice friendly GUI.

AMD hasn't had players in the Tablet and Ultrabook market because of the higher TDP's from initial releases, now they have adapted it they now have players, its like saying company x is new to making product y and saying they have poor sales in that area. You were the one that harped on synthetics and synthetics of mobile and DT Richland are there, they co-coincided what AMD stipulated earlier this year.

Intel's pricing was a report listed in THG article section that Intel raised SKU prices $30-50. There is also no pricing on GT3 but because of the complexity in its node it will be far more expensive than the lower speced GT2 parts, I speculate but I will expect top end GT3 vs GT2 parts to show around $100 difference in pricing or Intel lose out selling off higher end for cheap prices. And Ultrabooks at $500 is entry level, entry level Ultrabooks according to the chart have GT2. Intel want GT3 to replace the need for discrete AMD and Nvidia cards though most OEM's will strongly incorporate discrete. AMD does have a little advantage here in that they can offer up Dual Graphics before mid to high end discrete cards and on mobile resolutions you will get pretty amazing performance at a very low cost, and if you go by the chart they used the term "50% faster than the more expensive i7" implying lower cost so Kabini may have high end parts but that is not a tacit inference that they will only be making high end mobility parts.

I will agree on waiting for something on GT3 to come out, that is as before and enigma. It was touted so much yet nothing but delays and speculation, makes you wonder about the true nature of its performance, going on the costs involved here, I do expect it to at very least be decent.

As to not telling difference in FPS, well 24FPS constant is better than 50GPS spiking, the APU in basically all THG and TR reviews seemed to have latency under tight control, some instances it shows Intel HD to be spiking a lot, this is very noticeable in BF3 where FPS bounces min and max with a degree of regularity, the APU holds 40+ and so the experience is far smoother. Its quite ironic that here we talk about the "Weak APU" cores yet it more than handsomely pulverizes the 3770/HD4000 used in yesterday's drivers review, I have my suspicion that Intel HD reviews don't isolate its GPU contribution to the x86 and that the results reflect more x86 influence masking the real nature of HD. Vantage and 3DM 11 do isolate results but as before its worst case scenario.

DRAM and VRAM prices are escalating now due to stock preference shifts, I got this kit prior to increase.

Any ways nobody is really expecting you to like AMD products, I know you don't like the GPU's. For me I am still feeling my way around CCC as its the first time ive used AMD since the ATI X850 days and that was some time ago, I find Nvidia's control panel to be easier to navigate. I am awaiting the new AMD drivers, lots of talk on CFX improvements. I am also keen to get me a HD7750 and test that Dual Graphics fud with the APU, if it works out true thats a nice little freebie I hope AMD intended it and not just a glitch 😀
 
Thing is, are margins comparable here?
Is Intels VRMs the same cost overall, is the dram same costs?
Margins have to be made here, and unless theyre close in price, where the old way had competition, Intel wont, but still must maintain those margins.
Pricing will be interesting
 

not true, lol. amd did make tablet apus (z01 and z60 iirc), they didn't catch on. they also have ultrabook-class trinity apus, with appropriate tdp. they couldn't make ultrabook-class llano apus because stars cores didn't scale that low or something, i've forgotten. iirc intel's ulv cpus actually beat amd counterparts because amd's arch didn't scale that low. it was the opposite with brazos vs atom in netbooks.

this is a total lie. i don't post synthetic benches. ironically, you do that yourself rather frequently and point finger at others. i'll prove that later in this post. i understand what synthetic benches are and do not use them unless they serve a very specific purpose on rare occasions (e.g. 3dmark cpu physics).

those were pre-order prices.
if intel does try to lock out 3rd party d-gfx, won't start until mainstream desktop broadwell-based platform. won't happen with haswell iterations. intel(and laptop oems) already sell huge numbers of laptops with no discreet gpu. amd sells a sizable number of laptops with no discreet gpu. hell, they even ask customers not to buy discreet gfx cards by claiming their desktop apus already provide 'discreet-class' performance. i think amd has a headstart over intel in this regard, due to their powerful igpus. 😀
amd themselves neglect dual gfx support. i've pointed out the glaring drawbacks of dual gfx (dt and mobile), not gonna repeat here. like everything apu-related, dual gfx has a lot of potential, it's amd's fault for not supporting the way they should support.
amd can't help going for high end tablets, cheap ones won't make them enough money due to fierce competition from low end and cheap whitebox ones. kabini's biggest competition isn't anything from intel, rather tablets powered by apple, qualcomm and nvidia socs.

you know very wrong. i like amd gpus (and mobile apus) a lot. i just don't like what they do to those.
btw, which benchmark did you use to brag about your ln2 cooled apu rig? was it something real world, such as image/audio/video converting, archiving or gaming? which one was it again? :lol:

afaik, one of the biggest reasons for gt3's outworldly price would be the edram. rumor is intel is disappointed with how the fivrm turned out, so i'm guessing that they didn't improve power consumption for mobile parts and have to depend on more mature node and marginal perf improvement instead. it'd be worse if fivrm affected yields and intel end up selling cpus with defective vrm lasered off, lol.... wait, it could explain the external power delivery on the haswell mobos.... or does it?!?!
they're already running their fabs at nearly half capacity, then there's pc market's nosedive and early usb glitch (could make potential customers hold out for c2 rev.). pricing will be interesting indeed. i am expecting more stagnation this time around.

edit:
Intel revenue down, but forecasts hold
http://techreport.com/news/24670/intel-revenue-down-but-forecasts-hold
 
Futuremark, I posted it to isolate x86 from GPU performance that what futuremark does very well. But as before like all synthetics it gives worst case scenario over what a person may experience on a day to day, so for that I rely on Cinebench, Aida64, Sisoft Sandra suite, Futuremark, Heaven and gaming to find performance in perspective.

That was our local team participation in the ASRock FM2 challenge, set the continental record, finished 13th overall. Vantage was the bench used as it best represents synthetic values. The impressive part is that at 1.3ghz is the APU's score rivals some mid level graphics cards.
 
Someone said Intel could become similar to IBM, and that link doesnt exactly disprove this either.
ARM, its potential reminds some of the old riscy business in todays markets.

If I had to, I would say HSW is one of the most important chips Intel has made , and certainly since C2D
 


Ironically AMD and Intel are on important releases. AMD with SR arch which basically no questions asked has to be good, and Intel for future mobile.

 
Status
Not open for further replies.