• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Richland, Kaveri, Kabini & Temash; AMD’s 2013 APU Lineup Examined

d_brennen said:
Holy wall of text. BD was wrong for the desktop from the get go. AMD bet on parallelism, which hasn't materialised, yet...
Click to expand...

The thing is though,unlike Intel AMD essential shares the same CPUs for both server and desktop,which Intel does not,so they are compromise. Companies like IBM and Sun Microsystems make CPUs which are decent at throughput but have weaker single thread performance,probably worse than what AMD has ATM. The thing is though the BD derived cores have their issues,the whole architecture seems more modular than the previous ones,and AMD probably might even use more automated design tools than Intel(not surprising as Intel has 9 to 10 times the number of personal),but OTH we are starting to see much quicker updates and refreshes than with the Phenom and Phenom II. Even the console wins are probably down to how much easier the architecture can be tailored for different uses,ie,the whole SOC thing AMD was going on about a while back.

Intel might be able to optimise designs better than AMD,but they have the personal and can spend more,and this is allowed by the fact they simply produce more CPUs to spread out the costs. So AMD is limited by costs. You might have the best CPU in the world but if R and D costs are too high you will never make the sunk costs if you are a smaller company.

Regarding BD and its derivatives,the focus on integer performance, and not FP performance does indicate an architecture which probably is more suited to having FP operations done by a graphics card,which is funnily what you find in a lot of supercomputers.

This is why Kaveri will be important as both the CPU and IGP share the same memory space, and why HSA Foundation is important,so that the relevant tools can developed and used. The thing is I can see just like with 64 bit extensions for the consumer market,integrated memory controllers,etc, I expect AMD will be the first to do this.

However,Intel just has more money to spend,so I can see them catching up at some point and better implementing the ideas AMD has!! :(

humbug said:
It seems completely insane to me that a small core low power mobile chip like Bobcat can have 85% of Bulldozers IPC, i don't know what to make of that... (thats how good Bobcat is, or, thats how bad Bulldozer is) probably a bit of both.
Click to expand...

The thing is though,Jaguar might not scale well to higher clockspeeds.

Intel had the same issue with the P3. The P3 could not simply scale high enough in clockspeed, and only after many years and newer process technology did we get the Core and Core2 CPUs which are evolved from them. Hence they made the P4.
 
Last edited:
CAT-THE-FIFTH said:
Some more details about Richland:

http://translate.google.com/transla...spezifikationen-erster-richland-apus-genannt/

http://translate.googleusercontent....42013/&usg=ALkJrhhTy9SR1D_lO3ItPgWXeBtka-v08w

It looks like the 65W TDP A10 6700 should be faster overall than a 100W TDP A10 5800K.
Click to expand...

I think there will be a small improvement on the CPU performance and perhaps a slightly larger improvement on the GPU.

It does not look like its GCN, i think this will be a stop gap until kaveri to steal Intel's Haswell thunder.
 
humbug said:
I think there will be a small improvement on the CPU performance and perhaps a slightly larger improvement on the GPU.

It does not look like its GCN, i think this will be a stop gap until kaveri to steal Intel's Haswell thunder.
Click to expand...

I expect Intel to push benchmarks of the special GT3 IGP with embedded RAM in reviews. Then expect more AMD doom and gloom predictions.

Of course in reality you will find that most of the Intel desktop CPUs will have much slower GT1 and GT2 IGPs, probably without any embedded RAM, especially those at a comparible price to Richland,but expect that to be forgotten. Even with laptops most probably won't use a GT2 or GT3 IGP with embedded RAM anyway,because Intel charges a decent amount for it from what has been rumoured.
 
Last edited:
so its looking like steamroller in 2014 will be the next desktop replacement? looks like my 1090t will be at work a bit longer then.

are they looking to stick with AM3+ or move to a new socket? any idea anybody?
 
Nutella33 said:
so its looking like steamroller in 2014 will be the next desktop replacement? looks like my 1090t will be at work a bit longer then.

are they looking to stick with AM3+ or move to a new socket? any idea anybody?
Click to expand...

AMD said they are sticking with AM3+ for at least one more CPU. that would be Steamroller, at worst it would be AM4 and backwards compatible to AM3+.
 
excellent. thanks for the knowledge bomb humbug.

will have to find something else to scratch the upgrade itch. almost a year without upgrades feels wrong.
 
humbug said:
AMD said they are sticking with AM3+ for at least one more CPU. that would be Steamroller, at worst it would be AM4 and backwards compatible to AM3+.
Click to expand...

But this could change, regardless of what they've said.
I hope it doesn't, but it wouldn't even be the first or even second time AMD have left people on older sockets.
 
Martini1991 said:
But this could change, regardless of what they've said.
I hope it doesn't, but it wouldn't even be the first or even second time AMD have left people on older sockets.
Click to expand...

To my knowledge AMD sockets going as far back as AM2 have always been backwards and forwards compatible, My Phenom II x6 has a DRR2 IMC in it along side the DDR3, its an AM3 socket CPU that fits an AM2+ and is currently running in an AM3+ socket.
 
humbug said:
To my knowledge AMD sockets going as far back as AM2 have always been backwards and forwards compatible, My Phenom II x6 has a DRR2 IMC in it along side the DDR3, its an AM3 socket CPU that fits an AM2+ and is currently running in an AM3+ socket.
Click to expand...

You got the wrong end of the stick.
Was more on about compatibility of SR, if it came on a new socket, there's no guarantees it'd work AM3+.

AM2 was "The ONE!" but that got changed for AM2+, only a handful of AM2 boards ever actually had support for Phenom II.

Then AM3 was "THE ONE!!" Which got dropped for AM3+.

Only takes AMD to change the pin count for SR and a new socket/FM2 integration to break the cycle.

If AM4 did come out, same pin configuration, I'd expect AM3/AM3+ CPU's to work on it (Obviously, I can't see an FM4 existing just yet, I want AMD to move on from PGA)

The sockets themselves might be somewhat backwards/forward compatible, CPU's aren't always though.
 
Last edited:
Martini1991 said:
You got the wrong end of the stick.
Was more on about compatibility of SR, if it came on a new socket, there's no guarantees it'd work AM3+.

AM2 was "The ONE!" but that got changed for AM2+, only a handful of AM2 boards ever actually had support for Phenom II.

Then AM3 was "THE ONE!!" Which got dropped for AM3+.

Only takes AMD to change the pin count for SR and a new socket/FM2 integration to break the cycle.

If AM4 did come out, same pin configuration, I'd expect AM3/AM3+ CPU's to work on it (Obviously, I can't see an FM4 existing just yet, I want AMD to move on from PGA)

The sockets themselves might be somewhat backwards/forward compatible, CPU's aren't always though.
Click to expand...

AM2 is not AM2+, backwards and forwards one step, not two.... AM2 DDR2 >< AM2+ DDR2 >< AM3 DDR3 + DDR2 >< AM3+ DDR3 ?? AM4 DDR4 + ? No one can expect them to keep compatibility indefinitely.

I agree that we can not know what will happen when it come to it, but from my perspective.

AMD know and have acknowledged changing compatibility hurts sales, they sell CPU's, not Motherboards. They will look out for themselves.

Steamroller is not all that different from Piledriver, if it has DDR4 they can still keep the DRR3 IMC, in just the same way my CPU has a DDR2 IMC in it as well as the DDR3 and keep the pins the same.

There is no reason for AMD to change the compatibility from Piledriver to Steamroller, to do so would just hurt their business, i doubt Steamroller will see LGA/PGA, no reason for it.

Whatever comes after that, perhaps.
 
Last edited:
There was no reason to introduce AM3+ in the first place, as the CH IV is an AM3 board and runs Bulldozer, but is AM3 Bulldozer compatible on the whole? Nope, that's my point, you don't know what they will do, they could very well stop SR on AM3+ being compatible, or introduce a new socket, as they have done in the past.

AM3+ to AM4 would be one step, and it could very well lose compatibility if DDR4 was introduced, again that's just speculative talking.
AM3 to AM3+ was one step and the AM3+ CPU's don't work in AM3 (Bar the odd board here or there)
AM2 to AM2+ was one step, Phenom II's on AM2 weren't common or standard in fact, they were AM2+ CPU's.
AM3 was a filler socket (As seen now) and only introduced AM3, it didn't actually get any exclusive CPU's.
AM2+ CPU's to AM3 don't work (Such as the AM2+ Phenom II's)

Phenom II works in boards from AM2 (A handful) /AM2+/AM3/AM3+, that's a really positive thing, but I'm sure also came at a disadvantage of progression, so for Phenom II, there was a lot of backwards compatibility, but that's just really positives for their CPU they launched in 2009, as they had stagnant CPU performance for years. Sure, you could put the CPU in a ton of boards, but that's both an advantage and disadvantage.

See, you'll see the sockets in a Pro AMD light, I'll see them as both bad and good, because I can see both the advantages, and the negative things they've done.

Although, AMD do sell chipsets, it's in their interest to push new chipsets too (And ultimately new boards)

I can't imagine SR being LGA either, rather their next socket being LGA.
Although, your edit has PGA as no reason for it? Do you know what PGA is :p?
 
Last edited:
Martini1991 said:
There was no reason to introduce AM3+ in the first place, as the CH IV is an AM3 board and runs Bulldozer, but is AM3 Bulldozer compatible on the whole? Nope, that's my point, you don't know what they will do, they could very well stop SR on AM3+ being compatible, or introduce a new socket, as they have done in the past.

AM3+ to AM4 would be one step, and it could very well lose compatibility if DDR4 was introduced, again that's just speculative talking.
AM3 to AM3+ was one step and the AM3+ CPU's don't work in AM3 (Bar the odd board here or there)
AM2 to AM2+ was one step, Phenom II's on AM2 weren't common or standard in fact, they were AM2+ CPU's.
AM3 was a filler socket (As seen now) and only introduced AM3, it didn't actually get any exclusive CPU's.
AM2+ CPU's to AM3 don't work (Such as the AM2+ Phenom II's)

Phenom II works in boards from AM2 (A handful) /AM2+/AM3/AM3+, that's a really positive thing, but I'm sure also came at a disadvantage of progression, so for Phenom II, there was a lot of backwards compatibility, but that's just really positives for their CPU they launched in 2009, as they had stagnant CPU performance for years. Sure, you could put the CPU in a ton of boards, but that's both an advantage and disadvantage.

See, you'll see the sockets in a Pro AMD light, I'll see them as both bad and good, because I can see both the advantages, and the negative things they've done.

Although, AMD do sell chipsets, it's in their interest to push new chipsets too (And ultimately new boards)

I can't imagine SR being LGA either, rather their next socket being LGA.
Although, your edit has PGA as no reason for it? Do you know what PGA is :p?
Click to expand...


It was the AM3 / AM3+ debacle that's prompted them to think again.

No one knows for sure, but Bulldozer (originally AM3) was delayed for so long; Motherboard Manufactures pushed AMD to change the socket so they could get fresh sales.
The only reason there are so few AM3 Motherboards that support Bulldozer is because Motherboard manufacturers refuse to release Bulldozer BIOS for them.

A similar thing happened with Piledriver, no new socket, Still on AM3+... just before Piledriver was released Motherboard manufacturers started releasing revision Motherboards and called them Piledriver compatible.

To give you one example, Asus reintroduced their entire line of AM3+ Motherboards as Rev-2 and added Piledriver to the compatibility list.

As you know Piledriver requires a BIOS update on all Rev-1 Asus Motherboards for it to work properly, if at all.
This is because Asus did not support Piledriver on Rev-1 Motherboards, and even told customers who contacted them about Piledriver compatibility on Rev-1 that they would need to buy Rev-2 Motherboards for Piledriver.

AMD stepped in and reiterated that Piledriver is AM3+ compatible full stop.

Soon after Asus changed their tune and released a Piledriver BIOS for all Rev-1 Motherboards.

And Asus were far from the only ones who did this.
 
In my opinion, what you've replied with is a different topic, while a problem, isn't quite what I was on about (AIB's lying)
Although, I've seen no AIB have any issues updating any of their boards on Intel (H61 boards/P67 boards, etc) all supporting Ivy, so it's definitely a fishy scenario.

So we'll just leave it there.
 
As always, not sure what to make of them.
3DMark 11 I assume is an IGP test, but the 6800k shows minimal gains over the 5800k. Thought it was meant to be like uber IGP power?
 
CAT-THE-FIFTH said:
Some A10-6800K results:

http://wccftech.com/amd-apu-perform...utm_campaign=Feed:+Wccftechcom+(WCCFtech.com)

However,they are from OBR.
Click to expand...

3DMark11: 5800K = 1489 / 6800K = 1667, + 12% < not bad.

This is more intresting though.

PC Mark Prodoctivity.

P-II x6 1100T = 4021
FX-8350 = 4412
6800K = 4176

Better than a (6 real cores) x6 1100T and not far behind the FX-8350 which has twice as many moduals.

Having said that PC Mark is not exactly a good baromitor.
It would have been useful to have the 5800K tested in that aswell.

Martini1991 said:
As always, not sure what to make of them.
3DMark 11 I assume is an IGP test, but the 6800k shows minimal gains over the 5800k. Thought it was meant to be like uber IGP power?
Click to expand...

No, the 6800K is, and was always going to be just a 5800K refresh. Kaveri is the one thats a new archticture.

Edit, is that a new FX-8### for 2013 in that road map?
 
Last edited:
humbug said:
No, the 6800K is, and was always going to be just a 5800K refresh. Kaveri is the one thats a new archticture.
Click to expand...

When's Kaveri out :p?

A guy can get confused with all these names lol.

Richland according to that is the one with GCN, I always could have swore you mentioned the GCN one as being pretty game changing.
humbug said:
3DMark11: 5800K = 1489 / 6800K = 1667, + 12% < not bad.

This is more intresting though.

PC Mark Prodoctivity.

P-II x6 1100T = 4021
FX-8350 = 4412
6800K = 4176

Better than a (6 real cores) x6 1100T and not far behind the FX-8350 which has twice as many moduals.

Having said that PC Mark is not exactly a good baromitor.
It would have been useful to have the 5800K tested in that aswell.
Click to expand...


Look at the Phenom II X4 980.
Not so interesting now is it? Doesn't look that multithreaded.
 
Last edited:
Martini1991 said:
When's Kaveri out :p?

A guy can get confused with all these names lol.




Look at the Phenom II X4 980.
Not so interesting now is it? Doesn't look that multithreaded.
Click to expand...

Very end of the year I think they will be showing it off, shipping to OEMs for the summer next year. Expect mobile variants first again. I could well be posting about my new Kaveri based laptop in November before it exists :p
 
You must log in or register to reply here.
Back
Top Bottom