AMD Piledriver rumours ... and expert conjecture

[Reynod]

We have had several requests for a sticky on AMD's yet to be released Piledriver architecture ... so here it is.

I want to make a few things clear though.

Post a question relevant to the topic, or information about the topic, or it will be deleted.

Post any negative personal comments about another user ... and they will be deleted.

Post flame baiting comments about the blue, red and green team and they will be deleted.

Enjoy ...

 

[triny]

- Enjoy brilliant visuals in an HD tablet powered by an AMD APU..."

http://www.amd.com/us/products/notebook/apu/Pages/tablet.aspx

You can't miss the video.

AMD z high tablets with more graphic cores than IB hd4000

 

[nforce4max]

I caught the same article and got Crysis2 for $5.99...
so off days are me in the back of the house using 'nano-vision'...

:lol:

 

[jimmysmitty]

AMD z high tablets with more graphic cores than IB hd4000

More doesn't mean better. ATI has more cores than nVidia but nVidias cores are normally stronger. And with 80 cores, that means HD5450 performance, meaning on par with SB. IB will be more powerfull than a HD5450.

 

[Cazalan]

the way AMD is using new ideas you never know
what they'll roll out next .They have a third party dock on board
I doubt they went to the trouble of devising the on chip
3rd party dock for nothing.
could be a good power saving benefit for ulv solutions

That's not the purpose of the 3rd party IP block, practically the exact opposite.

It's more for networking IP, 4G/LTE, DSP's, to offer an alternative for companies already using ARM.

 

[triny]

More doesn't mean better. ATI has more cores than nVidia but nVidias cores are normally stronger. And with 80 cores, that means HD5450 performance, meaning on par with SB. IB will be more powerfull than a HD5450.

if amd z high chips equal or beat IB laptop igp
what will people say when they find out their new Intel ultra book can't perform
better than a 150$ tablet

mobile is tilting towards AMD coffers today even high end intel platforms ship with AMD
vcards more often than not.
I see a shift in the market favoring AMD
When I bought my i5 I used 2 hd6950 resulting in AMD making more money than Intel
a common scenario today,one that is likely to continue to be played out.
AMD caught Intel flat footed ,they have capitalized on Intel's weakness.
Should AMD be able to leverage this advantage into an opportunity to increase cpu power I'd say it's a bleak 2006 shaping up for Intel again in 2012 at least until Haswell.

 

[triny]

You idiot, that's what AMD's aiming for !!!!!

I mean........Bulldozer.........PileDriver.....Excavator.
They're all related to making holes in one way or another

steamroller flattens
it's like crescendo

Bulldozer rubble mover
Piledriver to put down roots
steameroller flattens evrything
Excavator to change the scenery
Tonka mania

 

[-Fran-]

The process advantage is a PITA for AMD, that's a given. Intel can put in their die a lot more than AMD can and keep a respectable TDP.

But I really want to see the HD4K tested against Trinity when they are both out there. Plus the QuickSync 2 it will sport (it's quality mostly) and the horrible driver support for that Intel is currently giving (quality I mean, again).

For HTPC duty, the HD3K and even nVidia's low end fell short in the last round against AMD (see the in-depth form Anandtech's). If it weren't for CUDA for nVidia and Intel's quality sacrifice for speed, they'd be really hopeless.

In the game arena, quality again might become important. With previous reviews, I've seen a couple of rants about the games running "funny", so let's see how Intel has improved their drivers with DX11 support.

Cheers!

 

[jimmysmitty]

if amd z high chips equal or beat IB laptop igp
what will people say when they find out their new Intel ultra book can't perform
better than a 150$ tablet

mobile is tilting towards AMD coffers today even high end intel platforms ship with AMD
vcards more often than not.
I see a shift in the market favoring AMD
When I bought my i5 I used 2 hd6950 resulting in AMD making more money than Intel
a common scenario today,one that is likely to continue to be played out.
AMD caught Intel flat footed ,they have capitalized on Intel's weakness.
Should AMD be able to leverage this advantage into an opportunity to increase cpu power I'd say it's a bleak 2006 shaping up for Intel again in 2012 at least until Haswell.

Thats a big "if". As I said, it will probably be equal to, at best considering tablets are only at LPDDR2 single channel, a HD5450.

I highly doubt AMDs tablet IGP will be better than IB IGP.

Check the IB rumors thread too. Some leaked benchmarks of HD3K vs HD4K showing average of 67% improvement, in some cases over 100%.

 

[nforce4max]

Thats a big "if". As I said, it will probably be equal to, at best considering tablets are only at LPDDR2 single channel, a HD5450.

I highly doubt AMDs tablet IGP will be better than IB IGP.

Check the IB rumors thread too. Some leaked benchmarks of HD3K vs HD4K showing average of 67% improvement, in some cases over 100%.

This is where I disagree with you when it comes to AMD IGP, sure Intel's PowerVR designed IGP will have features and will be much more power efficient than what AMD has to offer but raw performance is in AMD's favor. Unless you got some good links to convince me I will be sticking to what I know.

 

[triny]

all this talk about the AMD APU graphics part being better, OK we understand that.
how about the huge whooping Intel puts on AMD in the CPU dept.?
if the IB and it's HD 4000 can perform just decently then Intel and it's superior CPU gets my vote in notebook sector.

any 4 core x86 has more than enough cpu power for there about 90% of the people it's only hardcore gamers and power freaks that crave more cpu power.more often than not
Intel high end chips are sold in conjunction with radeon v cards .
better AMD should deliver a better package for 90% of customers.that will leave a visual experience
To most people a computer is generic .
if AMD can deliver magnitudes better igp and it scales so even with a vcard the igp is an extra bonus ( Trinity 384 core igp bonus to overall visuals )
an accumulative effect each year downsizig to 28nm in 2013 increasing igp to gcn 8860D
meanwhile AMD is improving cpu
Intel I expect will use a 6 core 14nm process cpu 2 cores igp in an effort to catch up
still won't be enough as AMD will have GCN scalable igp +HSA + improved cpu+ the option
to fab down.
It's the days of the multicore wars And the beginning of the supercomputer PC era.

 

[triny]

Thats a big "if". As I said, it will probably be equal to, at best considering tablets are only at LPDDR2 single channel, a HD5450.

I highly doubt AMDs tablet IGP will be better than IB IGP.

Check the IB rumors thread too. Some leaked benchmarks of HD3K vs HD4K showing average of 67% improvement, in some cases over 100%.

you have take that in context
67%+ mud on top of mud = mud
they are not getting discrete on IB but haswell 14nm could do it using a 6 core 2cores igp might catch lano 6550D
unless they buy up nvidia which isn't doing that well at the moment
As good as Intel is ,it still works tick tock
they can cobble up better igp than hd3000 incrementally but they can't do lano without more time at least by Haswell they can use more die space
Intel can't add 30% cpu uplift and give discrete 5400 igp with IB don't believe the Intelidiots

 

[jdwii]

so you speak for the public now...?
Intel high end is also sold with nVidia so calm down..
and again you think AMD will deliver this and deliverer that, assumption.
I see you have learned forum etiquette..
and then again with what you think converted into what you believe is fact so you speak as so.

I think you need to go talk to someone..
what realm (fantasy world) do you live in.?

I Don't know but i would like to join it a world where Amd delivers sounds nice! No but he has a point Most people Are fine with A 4 core and even more people are fine with a Dual core from Intel. But he also forgets that most people(NOT ME!) Are fine with Intel's 3000 HD graphics!

 

[triny]

so you speak for the public now...?
Intel high end is also sold with nVidia so calm down..
and again you think AMD will deliver this and deliverer that, assumption.
I see you have learned forum etiquette..
and then again with what you think converted into what you believe is fact so you speak as so.

I think you need to go talk to someone..
what realm (fantasy world) do you live in.?

it is not question of speaking for people ,it's a reality of where the biggest markets are .
USA is being eclipsed by China,India those markets are where the driving sales
will be.
Today nvida is on the ropes mabe tomorrow they will be ready to compete
AMD already delivered Lano made it's mark could of been a better start but it was enough to wake Intel up.
as for the rest of your dribble it's your dementia problem

 

[palladin9479]

This is where I disagree with you when it comes to AMD IGP, sure Intel's PowerVR designed IGP will have features and will be much more power efficient than what AMD has to offer but raw performance is in AMD's favor. Unless you got some good links to convince me I will be sticking to what I know.

Intel fan's won't ever admit it, but Intel's years behind ATI in graphics cards. It's not even a contest. This isn't Intel vs AMD for GPU's its Intel vs ATI. The graphics cores inside the APU's are just shrunk Radeon's. Thus for any Intel IGP to even remotely compete with an APU's IGP Intel would have to create a Graphics uArch that competes with ATI / Nvidia (they swap spaces every 6~8 months).

SB / IV stomps BD, we can all agree on that. Don't try to stretch that into the graphics department cause now your dealing with a separate engineering team.

 

[triny]

I posted this on the Ivy/Haswell thread but since the fanboism is out about Intel and nVidia graphics in the mobile sector,
I deem appropriate.

[flash=560,315]http://www.youtube.com/v/NekA62118Yg?version=3&hl=en_US[/flash]

I've said before, give me the Intel CPU but the AMD APU graphics part is better, but thinking it's trash or crap is unintelligent.
as for nVidia, well see for yourself..

Looks good by Haswell can't wait

 

[triny]

Intel fan's won't ever admit it, but Intel's years behind ATI in graphics cards. It's not even a contest. This isn't Intel vs AMD for GPU's its Intel vs ATI. The graphics cores inside the APU's are just shrunk Radeon's. Thus for any Intel IGP to even remotely compete with an APU's IGP Intel would have to create a Graphics uArch that competes with ATI / Nvidia (they swap spaces every 6~8 months).

SB / IV stomps BD, we can all agree on that. Don't try to stretch that into the graphics department cause now your dealing with a separate engineering team.

I just find it amusing how their testosterone gets their panties in a wad lol

 

[jimmysmitty]

This is where I disagree with you when it comes to AMD IGP, sure Intel's PowerVR designed IGP will have features and will be much more power efficient than what AMD has to offer but raw performance is in AMD's favor. Unless you got some good links to convince me I will be sticking to what I know.

He wasn't talking about Intels Atom with PowerVR vs AMDs whatever codename tablet APU. He was saying that AMDs tablet APU would be able to take on and beat Ivy Bridges HD4000 IGP. Atom is using PowerVR but Ivy bridge is using a in house Intel designed GPU. Sandy bridge already is equal to or better than a HD5450. The APU for tablets from AMD is using the same amount of SPUs as a HD5450, possibly VLIW. But we have to remember that its not equal to a HD5450 as it wont have GDDR3, it will use LPDDR2 like most tablets are right now in order to get better battery life and right now LPDDR2 is not nearly as fast as GDDR3 is not to mention its normally single channel, not even dual.

If anything that will be based on their E series APUs.

So if you take that into account, it will be weaker than a HD5450.

I don't doubt it will be faster than current PowerVR. Faster than the quad core PowerVR in the PS Vita? I don't think it will be as thats a monster, but also kills battery (the PSP Vita gets about 2 hours battery life gaming). Faster than HD4000? No I don't think it will be.

But thats just based on the specs and info we have. We will see when it comes out.

You have to remember that with Triny, anything Intel does GPU wise is crap even though Intel is showing great promise. I want Intel to push AMDs IGP division. That means better stuff for us much faster.

you have take that in context
67%+ mud on top of mud = mud
they are not getting discrete on IB but haswell 14nm could do it using a 6 core 2cores igp might catch lano 6550D
unless they buy up nvidia which isn't doing that well at the moment
As good as Intel is ,it still works tick tock
they can cobble up better igp than hd3000 incrementally but they can't do lano without more time at least by Haswell they can use more die space
Intel can't add 30% cpu uplift and give discrete 5400 igp with IB don't believe the Intelidiots

Right now a HD3000 IGP is about equal to or better than a HD5450. IB will nearly double the performance, which llano is double the performance of HD3000 right now in gaming only, not CPU.

How is nVidia doing badly again? The GTX460 is still a top selling GPU and the 560 is selling very well and has better performance than a equally priced ATI GPU. And I only buy ATI/AMD GPUs as I prefer CCC over nForce.

While its not the total, the Steam HW Survey is a good indication of where gamers are. According tho this nVidia has 57% of all GPUs:

http://store.steampowered.com/hwsurvey

If nVidia was doing poorly, I doubt they would be that high.

Haswells top end will have 40 total EUs, the GT3. If it scaled perfectly it would be 3x faster than IB. But we shall see.

Intel fan's won't ever admit it, but Intel's years behind ATI in graphics cards. It's not even a contest. This isn't Intel vs AMD for GPU's its Intel vs ATI. The graphics cores inside the APU's are just shrunk Radeon's. Thus for any Intel IGP to even remotely compete with an APU's IGP Intel would have to create a Graphics uArch that competes with ATI / Nvidia (they swap spaces every 6~8 months).

SB / IV stomps BD, we can all agree on that. Don't try to stretch that into the graphics department cause now your dealing with a separate engineering team.

No one has ever said Intel wasn't behind. Everyone knows that. I think that people are just excited to see intel is finally putting up somewhat of a fight on the IGP side.

And the APUs IGP is not just a shrunk down Radeon core. It doesn't have the same memory controller and is usually closked lower to lower thermals and power usage. They rely on the CPUs IMC and DDR3, which is much much slower than GDDR.

 

[triny]

true.

6core haswell with 2cores of igp 14nm
ought be able to fit some sort of discrete

 

[jdwii]

How is nVidia doing badly again? The GTX460 is still a top selling GPU and the 560 is selling very well and has better performance than a equally priced ATI GPU. And I only buy ATI/AMD GPUs as I prefer CCC over nForce.

??????????, The 560 is just as fast as the 6870 and the 560TI is just as fast if not a tad slower then the 6950 1GB card.

 

[triny]

You have to remember that with Triny, anything Intel does GPU wise is crap even though Intel is showing great promise. I want Intel to push AMDs IGP division. That means better stuff for us much faster.
I go byproducts that are available and frankly 67% plus hd3000 is not descrete


Right now a HD3000 IGP is about equal to or better than a HD5450. IB will nearly double the performance, which llano is double the performance of HD3000 right now in gaming only, not CPU.
Link to credible source
How is nVidia doing badly again? The GTX460 is still a top selling GPU and the 560 is selling very well and has better performance than a equally priced ATI GPU. And I only buy ATI/AMD GPUs as I prefer CCC over nForce.

While its not the total, the Steam HW Survey is a good indication of where gamers are. According tho this nVidia has 57% of all GPUs:

http://store.steampowered.com/hwsurvey

If nVidia was doing poorly, I doubt they would be that high.
design wise their behind the eightball
Haswells top end will have 40 total EUs, the GT3. If it scaled perfectly it would be 3x faster than IB. But we shall see.
yes how long before we see that ?


No one has ever said Intel wasn't behind. Everyone knows that. I think that people are just excited to see intel is finally putting up somewhat of a fight on the IGP side.
everyone is rooting for Intel's igp like they are for AMD cpu improvements

And the APUs IGP is not just a shrunk down Radeon core. It doesn't have the same memory controller and is usually closked lower to lower thermals and power usage. They rely on the CPUs IMC and DDR3, which is much much slower than
change with trinity and higher clocks
The A 10 5800 clocks at 4200 mhz the memory controller will be incorporated 2133mhz ram hd7660D igp ,say what about current developments for amd you add them when dreaming about haswell
GDDR.

 

[palladin9479]

No one has ever said Intel wasn't behind. Everyone knows that. I think that people are just excited to see intel is finally putting up somewhat of a fight on the IGP side.

Yes you have said that just not in so many words. Your saying that Intel's Haswell being able to compete with the Radeon GPU inside Trinity in the graphics department. Intel would need to jump forward several years in their development to compete with what ATI has ~now~. Much less what ATI will have in four years.

And the APUs IGP is not just a shrunk down Radeon core. It doesn't have the same memory controller and is usually closked lower to lower thermals and power usage. They rely on the CPUs IMC and DDR3, which is much much slower than GDDR.

Yes they are, period.

http://en.wikipedia.org/wiki/AMD_Fusion#.22Llano.22_.2832.C2.A0nm.29_2

Sabine used a HD5x series GPU with 400 shader units, 20 texture mapping units and 8 render units. They operate at 400~444MHZ and use 128-bit DDR3-1333/1600 memory.

This is the architecture a redwood series GPU.
http://en.wikipedia.org/wiki/Evergreen_%28GPU_family%29

The redwood's memory interface can accept DDR2/DDR3, the HD5450 was often equipped with it.

AMD combined the GPU and CPU external memory interface into a single unit. Its basically a HD5570 down-clocked to 400~444 Mhz and equipped with DDR3 memory instead of GDDR5.

You can hate on the BD uarch all day, but don't even try to say that AMD didn't use ATI's Radeon GPU.

 

[palladin9479]

And now that the local haters have decided to hate on the APU's graphics core. I'm gonna show what the entire concept behind BD and Fusion is all about.

Notice that the APU is using a R8xx core.

Here is some info about the architecture of that core.

Terascale 2 Architecture

With the release of Cypress, the Terascale graphics engine architecture has been upgraded with twice the number of stream cores, texture units and ROP units compared to the RV770. The architecture of stream cores is largely unchanged, but adds support for DirectX 11/DirectCompute 11 capabilities with new instructions.[4] Also similar to RV770, four texture units are tied to 16 stream cores (each have five processing elements, making a total of 80 processing elements). This combination of is referred to as a SIMD core.

Unlike the predecessor, as DirectX 11 mandates full developer control over interpolation so dedicated interpolators were removed, relying instead on the SIMD cores. The stream cores can handle the higher rounding precision fused multiply–add (FMA) instruction in both single and double precision which increases precision over multiply–add (MAD) and is compliant to IEEE 754-2008 standard.[5] The instruction sum of absolute differences (SAD) has been natively added to the processors. This instruction can be used to greatly improve the performance of some processes, such as video encoding and transcoding. Each SIMD core is equipped with 32 kiB local data share and 8 kiB of L1 cache,[4] while all SIMD cores share 64 kiB global data share.

Notice that the entire second part sounds a whole like like the internal components of a CPU's SIMD FPU.

That 3530MX APU happens to contain 400 of those shader cores which turns into 25 SIMD cores (16 shaders connected to four texture units is one SIMD core). Each of those is capable of a separate FMA / SIMD instruction.

That is why AMD is modularizing their CPUs. They plan to eventually remove their FPU entirely and use the Radeon Streams SIMD core as their FPU. Before they can do this they have lots of work to do, I'll give them another two to three years before they got it down.

 

[triny]

@Triny, you don;t mess with Uncle..... :lol:


but seriously, no one is doubting Trinity's ,IGP greater performance over Intel's HD4k IGP, but you'll have to accept the fact that the HD5450 is gonna be weaker than the HD4k 😛

and it doesn't matter if Trinity can support RAMs clocked at 2133mhz, it's just that they're expensive .
OEMs won't even go for it almost all the time because they buy bulk 1333mhz or 1033mhz RAM, and rarely (or never) touched RAMs clocked that high.

2133 ram is an option few would use maybe but an option none the less.

 

[triny]

And now that the local haters have decided to hate on the APU's graphics core. I'm gonna show what the entire concept behind BD and Fusion is all about.

Notice that the APU is using a R8xx core.

Here is some info about the architecture of that core.

Terascale 2 Architecture

With the release of Cypress, the Terascale graphics engine architecture has been upgraded with twice the number of stream cores, texture units and ROP units compared to the RV770. The architecture of stream cores is largely unchanged, but adds support for DirectX 11/DirectCompute 11 capabilities with new instructions.[4] Also similar to RV770, four texture units are tied to 16 stream cores (each have five processing elements, making a total of 80 processing elements). This combination of is referred to as a SIMD core.

Unlike the predecessor, as DirectX 11 mandates full developer control over interpolation so dedicated interpolators were removed, relying instead on the SIMD cores. The stream cores can handle the higher rounding precision fused multiply–add (FMA) instruction in both single and double precision which increases precision over multiply–add (MAD) and is compliant to IEEE 754-2008 standard.[5] The instruction sum of absolute differences (SAD) has been natively added to the processors. This instruction can be used to greatly improve the performance of some processes, such as video encoding and transcoding. Each SIMD core is equipped with 32 kiB local data share and 8 kiB of L1 cache,[4] while all SIMD cores share 64 kiB global data share.

Notice that the entire second part sounds a whole like like the internal components of a CPU's SIMD FPU.

That 3530MX APU happens to contain 400 of those shader cores which turns into 25 SIMD cores (16 shaders connected to four texture units is one SIMD core). Each of those is capable of a separate FMA / SIMD instruction.

That is why AMD is modularizing their CPUs. They plan to eventually remove their FPU entirely and use the Radeon Streams SIMD core as their FPU. Before they can do this they have lots of work to do, I'll give them another two to three years before they got it down.

by 2013 kaveri 28nm gcn 8000 on board their building a beast by excavator
Direct compute off the gpu gddr5
the reduced power usage of the 7000 series is a winner as well

 

[palladin9479]

Too bad trinity will most likely be a HD6 series GPU uArch. I can understand as their pretty strapped for manpower right now. I don't expect to really see anything impressive until mid 2013, but can always be pleasantly surprised.