• Happy holidays, folks! Thanks to each and every one of you for being part of the Tom's Hardware community!

Report: AMD's Richland Flagship Will Be A10 6800K APU

Page 2 - 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.
[citation][nom]unionoob[/nom]from my own experience budget gamers don't care if that is intel or AMD or something else, they just care is money, as it has to be cheap with good performance.[/citation]

Brand recognition. AMD has a long way to go, because a lot of people I know default on Intel simply because it was much more visible.
 
they should really implement a triple or quad channel memory controller with their next gen fusion. and, if it is ready, ddr4 memory. using a desktop, performance increases were really more dependent on system memory speed than gpu speed, when using a10-5800k.
 
The HSA Foundation (HSAF) was formed as an open industry standards body to unify the computing industry around a common approach.

The founding members of HSA are: AMD, ARM, Imagination Technologies, MediaTek, Texas Instruments, Samsung Electronics and Qualcomm.

Its good to see big power hitters like Samsung working with AMD and others. Maybe a marriage(Samsung and AMD)in the future?

I'm building a fanless HTPC with AMD A10-5700 Apu. AMD and ATI forever.
 
[citation][nom]otacon72[/nom]More APU garbage.[/citation]Who says the APU is garbage. The APU is good for gaming on a tight tight budget and HTPC. In a HTPC would you rather have a fast cpu and a slow gpu or a moderate cpu and a fast gpu. The APU gives a HTPC some decent gaming ability that no Intel cpu can't.
 


Yes :lol:

Just stuff that's leaked out over the last month or so -- Tom's had the Radeon 8000-Mobility article with the breakdown on GCN/Turks cores (I think). Not sure where I got the slide, but thanks to whoever it was. Someone at XS straightened me out on Trinity --> Trinity2 (Richlands) --> Kaveri. The AMD 2013 Roadmap is from CES.

AMD clearly feels the Richlands APU can significantly improve efficiency with 40nm Turks cores and 32nm Piledriver cores, and is not rushing to shrink any *APU* cores in response to Haswell. They've been doing 40nm since the ol' HD4770s, and 3rd-stepping 32nm Piledriver would likely be another good leap for them on a mature process (the 3rd stepping at 45nm was Denebs --> 6-core Thubans, essentially 50% more cores in a lower power envelop). It likely doesn't hurt that it's TSMC bulk, either.

The 40nm Turks cores are OpenCL/GL monsters. The 28nm Cape Verde GCN cores on the Kaveri APU will likely be furious bidness and blow past the 550ti that's been mentioned. Note how they are moving desktop HD7770 1GHz --> high-end mobile 600-700MHz (both with 640 shaders). Presumably, the *SIMD Engine Array* on Kaveri would highly resemble the Radeon Mobility 8800m.

With direct-addressing of unified memory, sez AMD :heink:

edit: from Toms

On the Kaveri APU, folks were **guessing** 512 shader cores -- like the HD7750. 640 cores, even being down-clocked, would be a blast ...

since everyone would be clocking it to 1GHz and running DDR3 2133 RAMs - LOL


battlefield%203%201920.png
 
[citation][nom]whyso[/nom]I doubt you will see 550 Ti levels. The 7660D is not quite as good as a 640 (about 10-20% less). A 550ti is about 40-60% better than the 640. I think it unlikely to see a 50-70% jump in gpu power, especially as the 640 is completely memory bandwidth limited (the 650 is essentially the same chip but with a 18% boost in core speed it is about 40-60% better than the 640 (about equal to the 550 ti)). Until AMD figures out a way around the bandwidth problem they are unlikely to exceed 640 levels.[/citation]There already are ways to address bandwidth limitations - unfortunetely they tend to conflict with other concerns like cost/power/space. One method would be to set up a sideport-esque pool of GDDR5 connected directly to the GPU side via a 128-bit bus, say 512MB-1GB. This would work in conjunction with the shared system DDR3 (2133?), for a large increase in total bandwidth.

Then there's eDRAM. It helps with certain bandwidth-intensive tasks, and you don't need a super-large amount of it to have an impact, but it does eat up a lot of space. ATI designs have already been used in conjunction with eDRAM on consoles. One advantage over sideport is that it isn't dependent on the motherboard manufacturers to supply good setups at a reasonable price.
 
[citation][nom]mazty[/nom]Budget gamers don't buy intel. That's just stupid.[/citation]

while that used to be the case you can get some decent low cost intel chips with low tdp if that is important to you, built a low budget rig for a friend with a 13 3220 dual core with hyper threading, 55w tdp and 22nm built, on black friday sale with motherboard and chip ran $200 which is pretty budget friendly imo $40 bought 8 gigs of ram and he already had a gtx 460 and reused a corsair 650 psu and case that a pretty nice budget build but he probably has less than $600 in it
 
[citation][nom]alextheblue[/nom]There already are ways to address bandwidth limitations - unfortunetely they tend to conflict with other concerns like cost/power/space. One method would be to set up a sideport-esque pool of GDDR5 connected directly to the GPU side via a 128-bit bus, say 512MB-1GB. This would work in conjunction with the shared system DDR3 (2133?), for a large increase in total bandwidth.Then there's eDRAM. It helps with certain bandwidth-intensive tasks, and you don't need a super-large amount of it to have an impact, but it does eat up a lot of space. ATI designs have already been used in conjunction with eDRAM on consoles. One advantage over sideport is that it isn't dependent on the motherboard manufacturers to supply good setups at a reasonable price.[/citation]

Very true, but if AMD were to be improving bandwidth through some manner as you described they would for sure be talking about it for marketing purposes. So I doubt they have anything there.

Plus the cost would pretty much destroy the whole deal about apu's which is cost.
 
[citation][nom]ddpruitt[/nom]How can Richland have an 8000 series Graphics core when it hasn't been announced and trinity still uses a 6000 series graphics core?I like the idea of going ARM though, low power + high performance (yes it can be done).[/citation]

It doesn't need to be GCN to be a Radeon 7000 core so long as it has the feature set of Radeon 7000. The same is true for Radeon 8000. Besides, Trinity's VLIW4 implementation is not exactly identical to that of the Radeon 6900 series anyway. It is die-shrunk with some Radeon 7000 features. If Trinity's successor is 28nm, then that's a lot less die-shrinking work, so it is possible and even likely that the successor to the GCN implementation used in the Radeon 7700/7800/7900 cards is in use in Trinity's successor.
 
[citation][nom]whyso[/nom]Very true, but if AMD were to be improving bandwidth through some manner as you described they would for sure be talking about it for marketing purposes. So I doubt they have anything there.Plus the cost would pretty much destroy the whole deal about apu's which is cost.[/citation]

It wouldn't be expensive to add another two channels to the memory interface for two or four cheap GDDR5 chips. Remember, they don't even need to be high end GDDR5 chips, just good enough to compare to that of say a GTX 650's memory or such.

I agree that it's unlikely that AMD did anything to fix the problem because, like you said, they'd probably be marketing it, but it is a plausible solution.
 
[citation][nom]g00fysmiley[/nom]while that used to be the case you can get some decent low cost intel chips with low tdp if that is important to you, built a low budget rig for a friend with a 13 3220 dual core with hyper threading, 55w tdp and 22nm built, on black friday sale with motherboard and chip ran $200 which is pretty budget friendly imo $40 bought 8 gigs of ram and he already had a gtx 460 and reused a corsair 650 psu and case that a pretty nice budget build but he probably has less than $600 in it[/citation]

I think that the point was that very low budget gamers wouldn't use an Intel IGP over an AMD IGP and wouldn't use an Intel system when a good discrete card such as the GTX 460 is not available without significantly increasing budget. An i3 is more of a mid-ranged CPU for gaming than a low-budget CPU, especially since an A6/A8/A10 APU can be had cheaper and comes with both a capable CPU and GPU.
 
Nice. Probably release just under $150 dollars, don't you think? I don't get why these things aren't moving like hotcakes. You can't beat them for the price in a midrange general purpose system. I hope AMD continues to expand it's FX line, but it would make more sense for them to focus on the APU series at the moment. In a related story, that would be a bit of an upset if PS4 features a custom A10 APU. Hasn't IBM had a stranglehold on the console market with custom PowerPC solutions for a while now?
 
@tmyk
Well of course Richland would perform better than Brazos, and I believe even more so than Kabini and Temash. Richland is replacing (or supplementing) Trinity, which I have a feeling would outperform, Brazos, Kabini, and Temash. I don't mean to offend, but you seem confused as to which codename each CPU (line) has.
I may be wrong with these, but here is what I know they stand for.
Trinity -> Richland -> Kaveri (mainstream--high-end APU's)
Brazos -> Kabini (low power APU's)
Hondo -> Temash (tablet APU's)
 
Status
Not open for further replies.