AMD Readies Radeon HD 8000M GPUs for Notebooks

16

Comments

+ Add a Comment
avatar

beta212

This would be nice, if amd fixed up the naming scheme which is slightly confusing, and the sucky drivers... But I have no doubt that they can pwn the hd 4000, I've tried gaming with a hd 4000, and with most of the eye candy off I was able to achieve a glorious 9 fps!!!

avatar

vrmlbasic

Perhaps with this new mobile GPU hitting the market we'll finally see PC laptops across the board move up in screen resolution, making 1920*1080 the minimum (finally phasing out the awful 1366*768) and something closer to the glorious MBP retina resolution the new maximum.

Here's hoping.

avatar

Hey.That_Dude

That's great AMD now put some energy into making your APUs have the CURRENT generation's graphics core instead of LAST generation's.

avatar

vrmlbasic

I'd like to see AMD accomplish its mission of allowing the Radeon on the APU to lose the memory-access-middleman that it still has to see performance gains. I'd also like to see software take further advantage of the APU Radeon's superior FP-crunching abilities, a goal aided by fixing the memory issue. I'd like to see this done still using the FM2 socket too ;)

Though if those things did happen then the APU could stand to outperform my CPU on floating point calculations, in which case the only point of pride I'd have left would be that my discreet GPU is still more powerful than the APU's Radeon, which is the only reason I can think of for the APU to not have the latest in graphics technology. :D

avatar

whr4usa

hahaha bragging rights fail indeed eh

can you clarify for my benefit what you mean by "memory access middleman" please sir? I *think* I might know to what you're referring but AMD's new internal architecture isn't a strength of my knowledgebase!

avatar

whr4usa

that would eat into discrete Radeon marketshare which is the only thing keeping them financially afloat

it'd also require modular internal processor design and well-engineered manufacturing processes neither of which AMD has ever been able to do successfully and both of which require R&D $$'s they can't afford in places where Intel has had them hopelessly beat for a decade now...

without same modular design and superior manufacturing processes how do you propose they design an x86 serial processing unit around a graphical parallel processing unit which itself hasn't been designed yet? huh? (:

avatar

Hey.That_Dude

It's not been made? So they're going to go from 0 to built in 3 months? Yeah like hell. Think a second before you go blabbering.
Their next gen CPU is supposed to be build on 28 nm solving the die scaling issue with the last bunch. They're supposed to add direct memory addressability to the GPU section removing that memory bottle neck. The APU would hardly even have the power to compete with discrete chips(not until they hit DDR4 3700 Mhz or around and about that).
HOWEVER, it would solve issues concerning Dual Graphics capabilities and make their Dynamic Switchable Graphics simpler with similar architectures running on the combined chips. THOSE two things would make this a nice chip to gamers who are pissed at the shoddy sata III implimentation on Intel chips. (namely me, I want a natively supported Raid 01 on SATA 3 INTEL!)

avatar

whr4usa

I thought for many minutes; if you thought for a couple seconds before your original post or this re-reply then you'd have realized what I was trying to explain is unavoidable and true

let's just say for example you start designing a new GPU in month 00 and then by month 03 you're designing a new CPU/APU/chipset/socket/whatever

well even if your GPU is finished by month 18 and your processor design isn't complete until month 27 then you can't possibly incorporate your new GPU deisgn into your new CPU design because you didn't have the final spx. of that GPU when your were designing the package, pinouts, lithography etc. of that new APU so instead you'd have had to deisgn that APU with your new CPU architecture and a previous GPU with known specifications from month '-18' without soem of the abovedescribed changes

avatar

Hey.That_Dude

Which would make sense if that was the case but it's NOT. Kaveri's Steamroller CPU architecture is already in the testing phase as is the new 8000 series. If you're telling me that it has a 7000 in it because it was being developed when the 7000 was released then you don't understand how development works. This is the kind of stuff where if AMD wanted to, which they should since it would increase profit, they could have the CPU design team and GPU design team talking a bit more on their parts as they're going to have to talk together on the silicon either way you slice it. I understand that there is a new GPU arch every 6 months or so and a new CPU arch every year. But that would mean by your logic that the CPU team would have to design based on what they had at the time of start which would be two generations ago! Which isn't the case! Therefore they are capable, just not willing, to integrate the new GNC 2.0 arch into the Kaveri APU.

avatar

whr4usa

yes, that's exactly what I'm telling you because their architecture isn't very modular compared to intel's and in "organizational years" they just bought ATI yesterday whereas Intel has had Larrabee for awhile and been doing basic display stuff internally forever (yes I know their chipset display controllers were horrible that's besides the point)

I agree they *should* have their cpu and gpu teams talking more if they wish to continue to exist as a wholistic business the next few years and not continue to shed business units and human resources and become the next Novel, Palm or BFG . . . I see a perfect analogy between Novel and AMD in that sense; mostly inferior yet effective and still competitive technology and a valuable global intellectual property portfolio but horrible management and finances ...the results; Novel and most of its neterprise disappeared 'overnight' in economic terms, while its 2 business units with the most gorwth potential have lived as independent wholly-externally-owned entities (ZenWork? and Suse) which were mostly saved by some of its former direct competitors (Microsoft, Red Hat, VMware and Citrix)

your last couple statements are completely wrong, my logic isn't, Intel's architecture is more modular and their product release TOCKtick cadence very well established internally AND they do their own manufacturing . . . AMD simply doesn't have these advantages or that level of organization and has a high rate of turnover whether because of layoffs, lack of available talent etc.
I am not saying it is impossible I'm just saying it can't happen with their current structure both organizationally and technically (...and yes, I do understand "how developement works", though I won't pretend to be an expert)

also let's for the sake of argument say I was wrong; you still ignored my alternative reasoning and incorrectly stated it'd increase their profits
most of their marketshare comes from the low-end discrete market especially mobile... if their integrated Gfx become equivalent with their lesser discrete cards, they just destroyed their primary market and thus their Gfx unit's profitability which they can't risk since 't's the only thing keeping their company afloat...and since Intel will also have the crown for energy-efficiency compared to AMD for x86 (until Via's newer processors are more widely avaialble) if they lose the mobile dsicrete low-ends they're just helping intel! so there's no economic or business reason for them to do it and no technical or organizational possibility of it happening, not for a couple years atleast

avatar

vrmlbasic

To which post are you actually responding?

If more software were designed to take advantage of OpenCL/GL for its relevant calculations in the unique scenario of the APU then the APU could stand to outperform a traditional CPU + discreet GPU setup, especially when AMD allows its APU's gpu DMA (as you mention). The integrated Radeon can blast through floating point calculations faster than most any CPU can and the data & instructions can be sent to & fro the APU's Radeon much faster than they can be shipped to the discreet GPU.

Moving data from its location in system memory to a region of system memory accessible by the discreet GPU then moving the data into the GPU's memory, performing the calculations and then having to move all the data back to system memory is SLOW, so slow that I believe that the APU will have an edge in this aspect of computing.

avatar

dawnripper

wow! hope not too skyhigh prices. Still using 6470M playing farcry3, 3missions to go...

avatar

kixofmyg0t

And here I sit completely satisfied with my 7770 lol.

I would like a 17" laptop with a A10 and at least 7870M.

avatar

DR_JDUBZ

mobile baaaaaaaaaaaaaaaaaaaaaaaaaahumbug, desktops are where its at, laptops are as fragile as a ferarri!!!

avatar

Andrew.Hodge

You may be right, but right now, the personal computer world is all about mobility. AMD needs to make money, and desktops just don't sell like they used to.

avatar

Refuge88

Exciting!

Log in to MaximumPC directly or log in using Facebook

Forgot your username or password?
Click here for help.

Login with Facebook
Log in using Facebook to share comments and articles easily with your Facebook feed.