Hailo-8 M.2 and mini PCIe AI accelerator cards deliver up to 26 TOPS
When looking at the performance chart released by Israeli startup Hailo, which pits Hailo-8 against Google Edge TPU and Intel Myriad X using popular ResNet and MobileNet models, I might as well titled this article as “Hailo-8 M.2 card mops the floor with Google Edge TPU and Intel Movidius Myriad X AI accelerators”.
The amount of AI processing power packed in such a tiny M.2 card is impressive, and it’s made possible by the higher 3 TOPS per watt efficiency compared to the typical 0.5 Watt per TOPS (or 2 TOPS per Watt) advertised by competitors. So a 26 TOPS Hailo-8 card will consume around 8.6 W, while a 4 TOPS Google Coral M.2 card should consume about 2 W.
As noted by LinuxGizmos who alerted us of the new cards, Hailo-8 was previously seen integrated into Foxconn fanless “BOXiedge” AI edge server powered by SynQuacer SC2A11 24x Cortex-A53 cores SoC and capable of analyzing up to 20 streaming camera feeds in real-time.
from the we're-all-grasshoppers-at-some-level dept.
Since the SN polls only allows for 8 entries, and I really thought this poll should have more and I don't have the time to learn perl at this moment just to suggest the necessary changes in rehash, I'm throwing it up here. (and it's also here because there's a good chance any election related poll will get veto'ed by the editors... and it's probably better that way too)
For the 2020 election, I intend to stock up on:
(42) weed, beer, beans, tortillas
My recommendation is to vote as AC, responding to the top comment "vote here" using the list number and provide commentary as logged in user or AC as you would normally below the voting thread.
Something interesting I read.
PDF is here.
Abstract
This paper presents a study of the runtime, memory usageand energy consumption of twenty seven well-known soft-ware languages. We monitor the performance of such lan-guages using ten different programming problems, expressedin each of the languages. Our results show interesting find-ings, such as, slower/faster languages consuming less/moreenergy, and how memory usage influences energy consump-tion. We show how to use our results to provide softwareengineers support to decide which language to use whenenergy efficiency is a concern.
Languages are broadly divided into three groups:
* compiled (eg, C, C++, Rust, others)
* VM (Java, C#, others)
* interpreted (Python, others)
The fastest language is not necessarily the most energy efficient. The authors measure CPU energy and DRAM energy separately.
I was most interested in the results for the language I use most: Java.
It did surprisingly well. On page 7 it was interesting to look at the bar graphs for several of the benchmarks in the Compiled, VM and Interpreted groups.
For a language in the VM group, Java had low energy numbers. If you look at the bar graph scale, Java was more comparable to the compiled languages group.
The results on page 8 are also interesting. Three tables. In the Energy and Time table, Java is fifth from the very top. On the Memory Use table, Java is 6th from the bottom. My conclusion: memory only costs money. You can't buy back Time, and you want low energy use. So higher memory use for great performance while having low energy use seems to be a win.
Interesting on page 10 is: Table 5.Pareto optimal sets for different combination of objectives.
The Combinations of Objectives are:
1. Time & Memory
2. Energy & Time
3. Energy & Memory
4. Energy & Time & Memory
On number 2, my favorite, Energy & Time, I see Java is again fifth from the top (below C, Rust, C++ and Ada). With a whole lot of languages below Java. Again I would point out that Memory is only a one time capital cost. Development Time (not in article) and Energy cost are ongoing costs. Execution performance is important to provide good customer experience, and Java does well in on that. You could get better execution performance than Java by trading off other factors.
There is also the factor of development time in different languages, which is another factor. But it is an economic consideration when building big software systems.
There is a cohort of close observers of our presidential elections, scholars and lawyers and political strategists, who find themselves in the uneasy position of intelligence analysts in the months before 9/11. As November 3 approaches, their screens are blinking red, alight with warnings that the political system does not know how to absorb. They see the obvious signs that we all see, but they also know subtle things that most of us do not. Something dangerous has hove into view, and the nation is lurching into its path.
You don't say!
The danger is not merely that the 2020 election will bring discord. Those who fear something worse take turbulence and controversy for granted. The coronavirus pandemic, a reckless incumbent, a deluge of mail-in ballots, a vandalized Postal Service, a resurgent effort to suppress votes, and a trainload of lawsuits are bearing down on the nation’s creaky electoral machinery. Something has to give, and many things will, when the time comes for casting, canvassing, and certifying the ballots. Anything is possible, including a landslide that leaves no doubt on Election Night. But even if one side takes a commanding early lead, tabulation and litigation of the “overtime count”—millions of mail-in and provisional ballots—could keep the outcome unsettled for days or weeks. If we are lucky, this fraught and dysfunctional election cycle will reach a conventional stopping point in time to meet crucial deadlines in December and January. The contest will be decided with sufficient authority that the losing candidate will be forced to yield. Collectively we will have made our choice—a messy one, no doubt, but clear enough to arm the president-elect with a mandate to govern. As a nation, we have never failed to clear that bar. But in this election year of plague and recession and catastrophized politics, the mechanisms of decision are at meaningful risk of breaking down. Close students of election law and procedure are warning that conditions are ripe for a constitutional crisis that would leave the nation without an authoritative result. We have no fail-safe against that calamity. Thus the blinking red lights.
There is more to this article. I know the Eds do not want Soylentils to see this, but I thought it was important it at least be submitted, twice.
____________
Other things the Eds do not want you to see.
Rejected in under ten minutes!
A Far-Right Militant Group Has Recruited Thousands of Police, Soldiers, and Veterans
Profile of Stewart Rhodes, of Rhodes Scholar-ing fame and Kaileigh.
Rejected in less than 60 seconds!
Scott Adams to vote Biden!
Says in Twitter video, that Trump personally screwed Scott in Tueday's debate, by not repudiating white supremacy. Once you've lost the Dilbert guy, who is left besides your children?
Thanks you guys!
You sure picked a couple of winners there!
What's the ugliest
Part of your body?
(Darling, when I . . .)
What's the ugliest
Part of your body?
(Darling, when I look in your eyes . . .)
Some say your nose
Some say your toes
(My dearest, my darling, my . . . darling darling . . .)
But I think it's your MIND . . .
AMD Ryzen 9 5900X “Vermeer” 12 Core & 24 Thread Zen 3 CPU Allegedly Up To 5 GHz With 150W TDP
Rumors say:
AMD will skip over 4000 to name the new Zen 3 desktop CPUs. There were plenty of complaints about Zen+ APUs being lumped in with Zen 2 CPUs and so on.
Somewhere between 15-20% higher overall IPC, ~10% higher clocks. Higher TDP: 150 Watts for the 12-core. Single core boost to 5 GHz.
Part of the improvement will be a unified 32 MB L3 cache for an entire core complex die (CCD). Maybe applications that are sensitive to that will show the most improvement.
12-core might be the top model at launch. That's how it was with Zen 2. The 16-core 3950X launched 4½ months after the 12-core 3900X. But it used better binned chiplets and had higher power efficiency than the 3900X.
We'll see how it works out in November
I don't visit there often, but I do now and then.
I tried to log in recently, and was taken to a "change your password" form.
Seems that I can't browse any of the rest of the site until I change my password.
However, after failing to change it a couple of times due to a mis-filled form, it now won't let me change it because I've submitte too many forms.
Anyone else having this issue?
Democrats start civil wars in other countries*, not here at home. You're supposed to fight on their turf. Ya don't pee in yer own pool! DUH!
Pretty freaky (but not mysterious in any way) to see how easily the democrats convince people to support republicans over them. The spread should be 80 points, or damn near unanimous. Can't be an accident. It should be well understood that winning is not the goal, it's to keep the game running and the money flowing. 50/50 partnership. 95% reelection. The commies should be jealous, they have to force people to vote for them. Here we do it freely and vociferously. All naysayers are sent to the cornfield
*lovely propaganda piece, don't ya think?
Well, all aristarchus submissions on all and sundry are being rejected according to plan. That is to be expected. What is not is the immediate rejection of aristarchus submissions to keep them out of the view of web crawlers, lest it appear that SN was a hive of liberality and objective news! And, then, of course, when we least suspected it, the dumbest Soylentil figured out aristarchus' fiendish plan! Oh! Busted! My Bad! Mea Culpa! So now we are flooded in the IRC with submissions from some "nutherguy". Dude, if you are going to do this, at least do it under your own username!
But things have gotten testy, even on the Editor front. Now there is nothing like TMB calling me stupid, which really is no insult at all since I know he knows not of which he posts, but more the banning of editors by other editors, and where I least expected to see it. Marty B has always be the exemplar of fairness, a Rock of SoylentNews. God knows I had my differences with janrinock when he was EIC, but then, he is not all that bright. Now, however, I come across this, over the continued queue stuffing by Runaway, and his threat to shoot our esteemed Editor in Chief:
[01:23:59] tell that to the guy who said he'd shoot you. even in jest that shit's not exactly civil
[01:24:10] I wartnedf you
[01:24:14] .kick AzumaHazuki
[01:24:14] -!- AzumaHazuki was kicked from #soylent by Aphrodite!Aphrodite@dodekatheon.olympus.gr [(Bytram (martyb)) No reason given]
[01:24:31] ++ Bytram
[01:24:41] nutherguy: watch it.
[01:25:03] *sigh* I got it backwards, don't I?
So, Bytram "wartnedf" Azuma Hazuki, and then kicked another ed from the #soylent channel on IRC. Maybe Marty is having a bad day, or perhaps the influence of a certain Boomer Trump supporting Soylentil has become toxic to the point of causing dissension within the ranks of our dear volunteer editors?
Now, I am just a regular soylentil, besides being 2400 years old, and I have seen this many times before. I suggest that the cause of the friction the removed. It really is time we ban Runaway1956. I say this with great trepidation and regret, as it is an extreme action to take, but the survival of SoylentNews as a viable place for discussion is at stake. "Nutherguy" poses an existential threat to SoylentNews.
Discuss.