Leaked AMD Trinity APU Slides?

Think of Bulldozer and Llano as shots in the dark to find out what works and what doesn't ... then Trinity is what they designed from scratched from what they learned. Fusion wasn't the easy task they initially believed ... hope they got it right this time.
 
Taldren said:
Think of Bulldozer and Llano as shots in the dark to find out what works and what doesn't ... then Trinity is what they designed from scratched from what they learned. Fusion wasn't the easy task they initially believed ... hope they got it right this time.
Click to expand...

Get it right this time? Fusion laptops have been selling like hotcakes and their performance is impressive. I just got a 500$ MSI laptop with a Brazos chip, clocked low, and it last 10 hours on battery. No fancy ULV chips from Intel for double the price, and it does just fine outputting to multiple displays including simultaneous HDMI and VGA. I'm really impressed by the size and value of the platform and say they did just fine out of the gate.
 
I still believe they should be putting the video RAM on the APU itself for designs such as this.

Maybe then we'd finally see the performance potential of integrated media processors, rather than just boring cost, space and power savings.
 
Are the IPC improvements enough to compete with Ivy Bridge? Because that's all that matters right now.
 
niconx said:
Are the IPC improvements enough to compete with Ivy Bridge? Because that's all that matters right now.
Click to expand...

Except, this Processor isn't supposed to compete with Ivy Bridge for CPU performance. It's supposed to compete with the Ivy Bridge for CPU+GPU performance, which I'm sure it does, and if it can do it at a lower price point along with higher performance, OEMs will buy it and use it.
 
Really interested in how this might do in a laptop form-factor. Not really interested in iGPs for desktops, but for a laptop, it's actually pretty compelling.
 
Dolor said:
Really interested in how this might do in a laptop form-factor. Not really interested in iGPs for desktops, but for a laptop, it's actually pretty compelling.
Click to expand...

That's it's main intended market. Laptops/notebooks are outselling desktops, and low-to-mid end models are the bulk of notebook sales.
 
clarkkent57 said:
That's it's main intended market. Laptops/notebooks are outselling desktops, and low-to-mid end models are the bulk of notebook sales.
Click to expand...

Yep, I'd love to have a $450 portable laptop for the couple of times a year I can't be at my desktop that could play recent games at 30+ fps at a decent resolution.
 
JosiahBradley said:
Get it right this time? Fusion laptops have been selling like hotcakes and their performance is impressive. I just got a 500$ MSI laptop with a Brazos chip, clocked low, and it last 10 hours on battery. No fancy ULV chips from Intel for double the price, and it does just fine outputting to multiple displays including simultaneous HDMI and VGA. I'm really impressed by the size and value of the platform and say they did just fine out of the gate.
Click to expand...

You do know that AMD themselves weren't even impressed with their first two fusion processors they released. I mean, its nice they have a niche in the market ... but going full fanboi doesn't change the fact that from an engineering standpoint it wasn't hitting on all cylinders.
 
VIC-20 said:
I still believe they should be putting the video RAM on the APU itself for designs such as this.
Click to expand...
you really don't. this would do really bad things to the cost as it would likely quadruple the die size (or more) leading to fewer usable dies per wafer.

increasing the bit width and bandwidth of qpi/ht has system wide advantages as well as igpu performance increases. granted, no way around the latency but if you really want performance for graphics then select the right hardware for it.
 
niconx said:
Are the IPC improvements enough to compete with Ivy Bridge? Because that's all that matters right now.
Click to expand...
IB will span from dual core Pentium chips up to 4c/8t versions on the mainstream LGA1155 platform. It's very likely that 2m/4c Trinity will match some model(s) in that range. Will it be competitive with IB at the high end? Of course not, and it's unlikely to be priced much higher than chips it can compete against in CPU performance. That's been the pricing strategy for the last 6 years.

Lacking a L3 cache, even with core improvements, Trinity is probably not going to be much faster (if at all) than current BD chips core for core and clock for clock. AMD expects to higher CPU clock speeds. With an improved GPU, it should be a good upgrade from Llano if AMD hits its CPU frequency targets.
 
Taldren said:
You do know that AMD themselves weren't even impressed with their first two fusion processors they released. I mean, its nice they have a niche in the market ... but going full fanboi doesn't change the fact that from an engineering standpoint it wasn't hitting on all cylinders.
Click to expand...

AMD was very happy with the A6/A8's, they weren't counting on intel being able to compete on integrated so now they can't share the memory (and limited at that) bus.

Lets see if they can seperate the two.
 
madrebel said:
you really don't. this would do really bad things to the cost as it would likely quadruple the die size (or more) leading to fewer usable dies per wafer.

increasing the bit width and bandwidth of qpi/ht has system wide advantages as well as igpu performance increases. granted, no way around the latency but if you really want performance for graphics then select the right hardware for it.
Click to expand...

Nothing stopping them from stacking the chips like they do in smart phones. Just stack some high speed video RAM on top.
 
Dolor said:
Really interested in how this might do in a laptop form-factor. Not really interested in iGPs for desktops, but for a laptop, it's actually pretty compelling.
Click to expand...

My current Llano high end HP laptop runs most games very well at medium settings.
 
niconx said:
Are the IPC improvements enough to compete with Ivy Bridge? Because that's all that matters right now.
Click to expand...

26% faster

so therotically a FX6100 no. of core / clocked piledriver CPU will edge out an i5 3570K by 3-5% in passmark

but srsly im still waiting for benches... :D
 
ktk_ace said:
26% faster

so therotically a FX6100 no. of core / clocked piledriver CPU will edge out an i5 3570K by 3-5% in passmark

but srsly im still waiting for benches... :D
Click to expand...

um, yeah ...and the bulldozer was supposed to be fast too. (will have to see benches first as well).
 
If these are valid, I like the fact they improved the following with Piledriver:
- Branch prediction
- L1 Inst./Data cache sizes
- L2 speeds
- Prefetcher
- FP and INT scheduler

Hopefully the above improvements net better IPC over Bulldozer modules.
 
ktk_ace said:
26% faster
Click to expand...
That wasn't an IPC improvement number, but a vague "26% better desktop system performance". Many of these types of multi-tasking benchmarks AMD uses for APUs also rely on GPU performance.

Trinity is an upgrade from Llano APUs for low end and mainstream systems, and with low clock speeds on those models it's not surprising that there may be room for higher system performance with Trinity-based processors.
 
pxc said:
That wasn't an IPC improvement number, but a vague "26% better desktop system performance". Many of these types of multi-tasking benchmarks AMD uses for APUs also rely on GPU performance.

Trinity is an upgrade from Llano APUs for low end and mainstream systems, and with low clock speeds on those models it's not surprising that there may be room for higher system performance with Trinity-based processors.
Click to expand...

I hope they increased GPU performance
 
You must log in or register to reply here.
Back
Top