Zambezi, Vishera - AM3+
Trinity - FM2
Llano - FM1

From my understanding... FM1 will co-exist for a while together with FM2 to cater the lower mainstream level (A6, A4 APUs) and still retaining the Llano architecture. Trinity FM2 will be on the upper mainstream with updated A8 as well as the new A10.

FX will still use AM3+ at least until Q1 od 2013 to serve the enthusiast/high end consumer market.

This post has been edited by soulfly: Mar 13 2012, 11:08 AM

nobody is really sure yet. it's not even released, no engineering sample test result to see.

Llano should be great for HTPC setup and normal office work, internet browsing and those general use. Real world power consumption should also less than Phenom/Athlon + cheap discrete, measurable but not highly significant.

Llano performance also scales better with higher DDR3 speeds compare to cheap Athlon II which hardly benefit higher DDR3 speeds.

You may get cheaper build cost with older Athlon/Sempron AM3 but cheap motherboards also have outdated and limited features. Most cheaper/older AM3 mobos does not have SATA3 or USB3.0. Even built-in HD4290 (890GX) has lower performance than HD6530D in A6-series. Gaming at low resolution... Can't get any wrong with Llano.

Let say we're comparing Athlon/Sempron based AM3 with built in graphics HD4290 (890GX) vs cheapest Llano A4 with built in graphics HD6410D, the latter is like 5-times better GPU performance than the former. At non-overclocked setting, AM3 Athlon II probably max out at ddr3-1333 or ddr3-1600 while Llano can still benefit ddr3-1866 greatly.

starcraft, dota, cs:s ... Llano should be no problem.
i'm expecting that diablo 3 will run without a hitch as well since the requirement looks like decently low.

Until then best to stick with AM3+ then

A55/A75/A85 chipsets can be used on both FM1 and FM2 sockets. The chipsets aren't exclusive for each socket generations.

Production cost.

You see, AMD is a much much smaller company than Intel. So they can't afford to lose much. I can see that they're just rolling out whatever they currently have while preparing to roll out new gen of processors (Piledriver).

This post has been edited by soulfly: Mar 18 2012, 01:51 PM

One thing that I really wonder about '29 percent faster', was that referring to clock-for-clock comparison, or it's faster by 29% clock advantage?

If Trinity is really faster 19% over Llano clock for clock, the performance improvement might be very significant for Kaveri over Zambezi too.... since clock for clock, Llano is faster than Zambezi in both single or multi-threaded performance.

okay... it seems that Trinity's better performance is because of the faster clockspeed. There's like 900mhz advantage. iGPU performance increased significantly but I'm more interested in the CPU performance. I'm seeing slight performance hit when running multi-threaded apps considering the L1 inst cache is shared for each module.

I wonder how would they fare if both are at the same speed

This post has been edited by soulfly: Mar 22 2012, 05:00 PM

per-clock, Piledriver was claimed to have IPC improvement over Bulldozer, but nothing concrete over Stars... and we know Bulldozer is weak per-clock against Stars

Server/Workstation optimized design may not really work for for desktops after all

A8-3850 @ 2900mhz @ CPU score 3814 = 1.315/mhz
A10-5800K @ 3800mhz @ CPU score 4304 = 1.132/mhz

A downgrade in performance per clock, really.

One way they could improve the performance over the same architecture is by utilizing lower latency cache memory.

However it may be really interesting if Trinity can perform better than Zambezi despite the lack of L3 cache.

This post has been edited by soulfly: Mar 24 2012, 11:32 AM

the issue with the weak bulldozer architecture was because the system design was more server-oriented... that is concentrating more on integer processing than FPU processing. the shared FPU units for two integer units on each module simply gave a performance hit under load.

a quad core bulldozer in reality is no different than a dual core Intel with hyperthreading in a normal end-user client system. it's just that AMD implemented its own 'hyperthreading' on the hardware rather than using instructions.

in my opinion, AMD should compete with Intel in terms of efficiency and price point. if they cannot compete in the high end segment so be it. High end segments are just a niche market anyway.

This post has been edited by soulfly: May 25 2012, 10:11 AM

I understand what you mean... totally agree regarding the crossfire thing. A real IGP+discrete crossfire would be amazing compare to the hybrid crossfire setup. I imagine something like Lucidlogix's Virtu system. IGP on normal usage, and crossfire when loading hardware accelerated apps like gaming, encoding, OpenCL supported apps. That's what we call hardcore heterogeneous computing.

The funny thing is that the GPU in Trinity APU was actually a heavily crippled Radeon HD 6900-series.

I'm putting hope in Vishera as they claimed that it will show full potential of Piledriver compare to Trinity. Hopefully the L3 cache and quad-channel memory system will give significant advantage on Piledriver performance.

the review was comparing trinity to llano. there will be a continuation though... it's not completed yet.

If Trinity can outperform FX-4100 clock for clock then it should be the real deal indeed.

Trinity can be an advantage for those seeking for lower cost encoding machine, htpc and light gaming.

i think the writer was referring to the same ES Bulldozer C0 chip, which coolaler claimed as a Piledriver ES.

The problem is that chip doesn't even replicate the performance of Trinity, so I don't think it is a real Vishera (Piledriver) at all, instead is a minor improved Zambezi (Bulldozer).

AMD would probably skip the real Piledriver altogether and just prepare for Steamroller in APU form.

This post has been edited by soulfly: Aug 24 2012, 10:44 PM

Piledriver might be the last AMD CPU for consumer as the current trend is moving towards APU. Steamroller CPU might be reserved for the server segment only.

A6-3650 to A10-5700 or A10-5800K is a significant upgrade

both cpu and gpu

For example A10-5800K definitely has the clockspeed and graphics performance advantage at the same 100W TDP. A10-5700 even has lower TDP at 65W. A lot of apps takes advantage of higher clockspeed like encoding, light gaming, etc.

You could check tomshardware early Trinity preview where they compare A10 to A8 Llano.

This post has been edited by soulfly: Sep 5 2012, 09:52 PM

1100T definitely

gaming normally does not benefit from more than 4 cores.
by average 1100T performs better than 8120, in certain cases better than 8150.

If Vishera can perform better than Thuban at the same price point and TDP, then it can be considered as improvement.

some ppl on XS forum got their hands on Zambezi C0 a.k.a FX-8350 already