D
Deleted member 13524
Guest
As far as anandtech's assiduousness in their reviews... well, I'm still waiting for the Part 2 of the Galaxy S4's review, so...
NVIDIA Tegra Architecture
- Thread starter french toast
- Start date
-
- Tags
- nvidia
Nebuchadnezzar
Legend
Oh lol.
The N9 article got delayed by other reasons unrelated to the device itself. I hope we'll get it up in next few days or next week at worst case. Immediately following that will be a piece where you'd probably piece together why Nvidia might be wise to go to vanilla A57 cores.
=)
I've been waiting for the N9 review mostly because Denver is an interesting CPU, and I want to learn more about it.
Anandtech has a reputation as the most technically insightful review shop, so I look forward to what they have to say.
On the other hand, SGS4 wasn't super interesting to me from a technological perspective...
Well, if we are discussing Anandtech.. there was a time when you could just read Anandtech and trust that the article writer was thorough. The conclusions, explanations or speculations were sometimes wrong, but those were honest mistakes (i.e. they did the best anyone possibly could have).
Now, with the new writers, there are some that are excellent, and there are some which you simply cannot trust, as they make unsubstantiated claims even a layman can spot mistakes in. And the problem is, Anandtech was the one site where you didn't need to "filter" – you could just read it and trust it.
Thankfully, with the N9 review, it will be among those that are really worth waiting for and reading
Well, if we are discussing Anandtech.. there was a time when you could just read Anandtech and trust that the article writer was thorough. The conclusions, explanations or speculations were sometimes wrong, but those were honest mistakes (i.e. they did the best anyone possibly could have).
Now, with the new writers, there are some that are excellent, and there are some which you simply cannot trust, as they make unsubstantiated claims even a layman can spot mistakes in. And the problem is, Anandtech was the one site where you didn't need to "filter" – you could just read it and trust it.
Thankfully, with the N9 review, it will be among those that are really worth waiting for and reading
I had contacted Anand himself more than once in the past to try and correct more than one mistakes over several articles I had read. And I intend to continue doing it whenever I pick up other mistakes in the future. I'd say cut them some slack; no master ever fell out of sky
ltcommander.data
Regular
http://purch.com/purch-acquires-anandtech-dominates-tech-expert-and-enthusiast-market/
Purch who owns Tom's Hardware has now bought Anandtech. The increased resources should be good for Anandtech, but it would have been nice to have 2 fully independent sources for reviews. It's interesting that other than the names and titles of people being quoted, the other major text that is bolded in that press release is the number of sellers and marketers and sales potential.
Purch who owns Tom's Hardware has now bought Anandtech. The increased resources should be good for Anandtech, but it would have been nice to have 2 fully independent sources for reviews. It's interesting that other than the names and titles of people being quoted, the other major text that is bolded in that press release is the number of sellers and marketers and sales potential.
Nobody cares about the masters falling out of the sky. If they didn't ever, they probably wouldn't be masters. Carefully qualified guesses is what gives Anandtech value.
The problem is the writers whose work simply isn't diligent in general.
A1xLLcqAgt0qc2RyMz0y
Veteran
All this discussion on the next Tegra using standard cores vs Denver cores seems to miss the point that Nvidia currently has two Tegra K1 variants: The Tegra K1 with Cortex-A15 cores and the Tegra K1 with Denver cores.
I expect that this will continue with the next gen Tegra with Maxwell GPUs.
I expect that this will continue with the next gen Tegra with Maxwell GPUs.
Nobody cares about armchair philosophies either.
....and everyone else's is? I for one am willing to give them a chance to see if they can improve. If not there aren't all that many serious tech sites left on the internet either. At least I'm confident that - on topic - I'll gain more insight on Denver from their upcoming two articles about it then any whitepaper copy/paste job discussed above. And that at least for me is value enough.
Last edited:
The A57/A53 variant has taped out quite some time ago under 20SoC. If a second Erista variant exists it can't hide, we'll hear of its tape out fairly soon.
A1xLLcqAgt0qc2RyMz0y
Veteran
Both Tegra K1's (32bit & 64bit) were announced at CES 2014 but the delay between them being available was around 6+ months.
So if the same holds true then the Tegra M1 (A57/A53 variant) and the Tegra M1 (Denver) can be announced together at CES 2015 with the Tegra M1 (A57/A53 variant) being available much sooner than the Tegra M1 (Denver).
It may be that to hit the release patterns of tablet/auto/phone makers that Nvidia is going with a staggered release.
Also with the latter release of the Tegra M1 (Denver) would the 16nm FF be available for it.
D
Deleted member 13524
Guest
Wasn't Denver's original purpose to be put inside GPUs for high-end discrete graphics cards in order to assist some tasks and offload the main CPU?
A1xLLcqAgt0qc2RyMz0y
Veteran
No, that was pure speculation especially for the general consumer market.
In HPC (Denver and GPU on same die) it may have some use (in getting rid of the Intel CPUs) but with the close relationship now with IBM this probably won't happen.
Erm..the point has not been missed at all..we've discussed that in detail already.
~6 months back if my info was correct.
Yes but we dont know when both chips taped out..only that the A15 based K1 taped out in Q3'13. I suspect the gap between the tapeout dates was less than the actual time between product availability.
For the tablet market possibly..but I dont see them getting any traction in the phone market. We've seen that with Tegra 4 and K1. And automotive is not that time sensitive. AFAIK Tegra 4 isn't even shipping in the automotive segment yet (and not sure if it ever will as I remember reading something about it not having the required interface support)
I doubt they are going to go for a different process node for the Denver variant. The design time and cost would be likely too expensive to justify it. Not to mention 16 FF readiness and availability in that timeframe.
Wasn't Nvidia's original purpose to try to get into the x86 market with it?
While they are partnering with IBM for the supercomputer market..I do see a market for Denver CPUs to be integrated with GPUs. ARM based CPUs will probably be more power efficient that IBM's POWER architecture and could be better for some applications. And the die size for an ARM core on a sub 20nm process will be so small that it should be very feasible to do.
The whole ARM cores on die story started when the first Exascale research data appeared on the internet years ago projected then for 2017 which was mentioning 16 ARM cores on die. I think but am not sure that more recent material speaks of 8 cores, which is understandable considering how problematic manufacturing processes have become. In any case to be honest any of the above material didn't ever mention Denver from what I recall, it's probably just that everyone assumed that since NV is developing a custom ARM core that they'll be using it there also.
KNL can boot by itself according to Intel.
Look above; Erinyes just said that the ARM Erista variant taped out about 6 months ago. I said above that if a Denver variant exists we should hear about its tape out anytime soon and yes they'll probably announce both. I still don't see any compelling reason as to why they would want two variants with 64bit CPUs but that's just me. As I said a long time with K1 the primary reason of having two variants is that the first was 32bit only.
KNL can boot by itself according to Intel.
Look above; Erinyes just said that the ARM Erista variant taped out about 6 months ago. I said above that if a Denver variant exists we should hear about its tape out anytime soon and yes they'll probably announce both. I still don't see any compelling reason as to why they would want two variants with 64bit CPUs but that's just me. As I said a long time with K1 the primary reason of having two variants is that the first was 32bit only.
Last edited:
A1xLLcqAgt0qc2RyMz0y
Veteran
Project Ara Gains Support For Nvidia Tegra K1 And 64-Bit Marvell Chip Modules
Looks like Tegra will still be phones.
Looks like Tegra will still be phones.
A1xLLcqAgt0qc2RyMz0y
Veteran
NVIDIA Announces 2015 CES Press Conference, Live Webcast Coverage
http://finance.yahoo.com/news/nvidia-announces-2015-ces-press-140000013.html
http://finance.yahoo.com/news/nvidia-announces-2015-ces-press-140000013.html
Nice. However, do they always write that long paragraph about the statements in the press conference being just forward looking statements, which realization depends on several factors? Is that a new thing?
Similar threads
