NVIDIA APX 2500 (AP15) Pre-Launch Thread

Status
Not open for further replies.

Arun

Unknown.
Moderator
Legend
After an absolutely insane amount of hyping up by Jen-Hsun Huang to investors in the last 12+ months, going from claims of being a completely new product category to being the first single-chip handheld computer and passing by apparent expectations of high market penetration... It'll finally be there in the next few days. Here's the proof. To Jen-Hsun's credit, he *is* very passioate about what he does so I'll forgive him even if it was overhyped to investors (which we'll know soon enough).

Odds are that it'll be good, but perhaps not worth Jen-Hsun's enthusiasm. I'm also very skeptical their definition of single-chip integration is the same as the industry's definition of that notion. We'll see. Unless they integrate power management (that'd be aggressive!) or some form of wireless (unlikely), it's not very clear what they could do to have higher integration than TI's OMAP3 (except touchscreen analogue stuff).

Regarding features, here's what I'm expecting:
- 65nm eDRAM @ TSMC
- ARM Cortex-A8
- OpenGL ES 2.0 GPU
- Excellent camera/imaging pipeline
- Integrated Audio DAC & Touchscreen Controller
- 720p H.264 Baseline decode, 720p encode (MPEG2? H.264?)

In terms of internal arch, I'm expecting a Tensilica core for audio, but a non-Tensilica DSP core along with fixed-function units for video (ala what PortalPlayer did for the 6100). No idea about graphics. All in all, nothing revolutionary if those highly speculative specs are right, but still a very solid offering in terms of features and performance. What really matters in the end is the power efficiency (especially for multimedia, but also for the Cortex-A8 core) compared to OMAP3 and Broadcom, though, and we'll see how much (if any) data will be available about that.

This is obviously the first joint project between NVIDIA and PortalPlayer engineers (the latter which had very nice power-efficiency numbers for the 130nm(!) GoForce 6100), but interestingly I suspect it's also the first one with Stexar engineers contributing. I doubt we'll get much if any info on that, but who knows!
 
Just saw that, intriguing. Not a lot of details yet though, and a high-clocked ARM11 is a strange but potentially interesting design decision. The biggest question is how they achieved that frequency and what the trade-offs were involved. I suspect it's not just performance-oriented synthesis; I think they might have picked a process variant/gate engineering trade-off higher on the performance/leakage curve.

If that's the case and they have the capability to shutdown everything on the chip depending on the workload, and they can shutdown the ARM11 even during semi-bursty workloads like web browsing, that architecture could definitely make a fair bit of sense (although it's hard to judge properly without more data) because dynamic power at a given frequency might be lower than the competition.

Either way, I'm writing a list of questions that I'm going to field to NVIDIA soon - we'll see how much they'll be willing to answer, as well as how much data they'll give publicly on their website (since it looks like EETimes published pre-NDA expiry?)
 
Status
Not open for further replies.
Back
Top