Nvidia details 64-bit Denver Tegra K1, claims Haswell-class performance for first 64-bit Android chip
on August 12, 2014 at 7:38 am
This site may earn affiliate commissions from the links on this page. .
Nvidia has finally taken the wraps off its 64-bit Denver CPU core — and, well, let’s just say that the 64-bit Android race just got a lot more interesting. The Denver CPU will be available later this year in the Tegra K1 SoC, where you’ll get two Denver cores and a 192-core Kepler GPU. Nvidia says this will be the first 64-bit ARM chip for Android, and that it “completely outpaces” other ARM SoCs (including Apple’s monstrous ). The Denver CPU is so beastly that it even beats out some PC-class chips (the low-end Haswell Celeron), while consuming much less power. The secret behind this performance boost is a novel technology called Dynamic Code Optimization.First, let’s run through the Denver/K1 hardware specs. Denver is a 7-way superscalar in-order CPU core (it can execute seven instructions per clock cycle, which is a lot). Each core has 128KB of L1 instruction cache and 64KB of L1 data cache (both very large compared to the usual 32KB/32KB for a 32-bit ARM chip). In the 64-bit Tegra K1, there will be two Denver CPU cores sharing 2MB of L2 cache. The Denver CPU is expected to ship at 2.5GHz, which is rather fast. The Denver Tegra K1 will be pin-compatible with , making it easier for hardware makers to embrace the new chip when it arrives later this year.Nvidia Denver CPU core, block diagramDenver cores are very large, and there won’t be a quad-core variant (at least not in the short term). While the lack of a quad-core part might make for an uphill marketing battle — people still love their higher core counts — in reality a dual-core CPU should be more than capable of handling most workloads (most apps and games are still bottlenecked by single-threaded performance).Nvidia Denver CPU core, Dynamic Code OptimizationDynamic Code Optimization Denver is explicitly designed to have “the highest single-core CPU throughput” — and while its 7-way superscalar design definitely accounts for a lot of that performance, the other trick up its sleeve is Dynamic Code Optimization.Back in January we speculated that . As it turns out, Nvidia did try something rather weird and funky with the Denver core — but it has very little to do with x86. While Denver is an in-order chip, most modern CPU cores use out-of-order execution (OoOE) — where code is executed whenever it’s ready, rather than in the exact order laid out by the programmer — to increase the number of instructions executed per cycle. This works well, but it requires a significant chunk of silicon to actually perform the OoOE magic — and that silicon takes up die space and power. Instead, Denver uses Dynamic Code Optimization, which stores the most regularly executed code in a special 128MB main-memory Optimization Cache. This way, instead of having to fetch and decode instructions and data multiple times, the pre-digested and ready-to-go code can be run directly from the Optimization Cache. It’s a similar approach to how
work.Nvidia 64-bit Denver CPU core, internal benchmarks – versus some other ARM and x86 coresWhile the actual process of storing code in the cache requires some additional processing power, Nvidia says the long-term performance gains are huge. In the case of Android tablets, where you probably only run a handful of apps, you could be running almost entirely from this optimized cache — in which case, the performance gains will be really quite significant.In theory…Android LOn paper, and in Nvidia’s internal benchmarks, the 64-bit Denver CPU core sounds rather exciting. The CPU handily beats every other ARM CPU on the market by some margin, and even does pretty well against the dual-core Celeron 2955U (Haswell). The 192-core Kepler GPU will probably be
by some margin, too.Perhaps most excitingly, though, the Denver Tegra K1 also looks like it will kick ass in reality, too. It’s important to note that the 64-bit Tegra K1 is the SoC of choice for Google’s upcoming . Android L, which is due out towards the end of the year, is being developed specifically for the 64-bit Tegra K1 — and all signs point to , which is probably an Android L launch device, having Nvidia’s 64-bit chip inside, too.In the meantime, we’re still waiting to see what Qualcomm has planned for its ARMv8 64-bit CPU core. Apple caught everyone off guard when it was first to market with its custom-designed ARMv8 Cyclone CPU — and now Nvidia is proudly touting that it’ll be the first to market with a 64-bit Android CPU. It will be interesting to watch Qualcomm play catch-up when its 64-bit chip is finally ready to go (probably in 2015).
Post a Comment
This site may earn affiliate commissions from the links on this page. .
ExtremeTech Newsletter
Subscribe Today to get the latest ExtremeTech news delivered right to your inbox.
This newsletter may contain advertising, deals, or affiliate links. Subscribing to a newsletter indicates your consent to our
. You may unsubscribe from the newsletter at any time.
More Articles
We have updated our PRIVACY POLICY and encourage you to read it by clicking .Intel plans 64-bit and Extreme Edition Pentium-Ms
This site may earn affiliate commissions from the links on this page. .
user comments 21 comment(s)
superfast ms? (10:13am est wed mar 09 2005)ok, i have a dothan pentium m, and although it is quite good in business apps, it does fall down in gaming.
business stuff runs fast enough as it is.
more cpu intensive stuff doesn't need to be done on a notebook (for what one with that cpu will cost, you can buy another computer).
take away the m's heat and battery life advantages, and there is no longer any reason to buy one over a desktop cpu based notebook.
it sounds like an interesting idea, but i just don't see how intel will pull this off at a price that makes the added disadvantages worthwhile. – by akula
screw mobility (10:24am est wed mar 09 2005)bring yonah/merom to the desktop/workstation world, intel.
once it gets 64-bit extensions and runs at 3 ghz, it will be a serious threat in that arena. – by grover
omg (10:56am est wed mar 09 2005)this is the most retarded idea yet.
lets put a wattage guzzling chip in a laptop
60 freaking watts on a laptop!!!!!
the current pentium-m dothan runs at like 12 watts.
with a laptop, you want low wattage so your battery lasts a while.
i wouldn't be suprised if a laptop with a pentium-mee won't last an hour on battery and is huge and bulky.
once the turion comes out, dell is gonna get screwed on laptop sells.
it may be 25 watts, much higher than the dothan, but it will also be a lot faster. – by rpgman
intel and processors (11:44am est wed mar 09 2005)intel is getting in over their head with the amount of processors they are placing on their road-map.
they are placing more than a dozen new processor designs on their road-map for the next 18 months, not to mention .65nm cpu's.
isn't that a little much?
do they have the fabrication plants to accomidate the demand for all these processors?
don't all those new cpu's require a new chipset?
so, does that mean that all intel owners have to buy a new motherboard in order to upgrade?
time will tell i guess.
i know amd is placing quite a few too, but there are what eight processors?
of those eight, maybe half are going to be really new (d-c opteron, d-c 939 cpu's, and turion) the rest will be upgrades of current cpu's.
yes, they are making all their new and upgraded cpu's on the .90nm fabrication, but they are not trying to drown the market with different cpu's.
if intel can pull this off, they deserve to own the cpu market, but if they have to cut more than a fourth of their road-map, then they deserve to die off.
being ambitious is one thing, but there is a fine line between that and insanity.
intel is walking that line at this point.
just my $0.02. – by chronos
this direction is scaring me off (1:11pm est wed mar 09 2005)i have an aug. 2004 hyperthreaded p4 compaq notebook with a wide screen that weighs over 8 pounds and has 1 hour of battery life.
i hate it even though it's beautiful and really whips through the visual studio compilations.
i wince at the thought of carrying it anywhere
even from my desk to my bed.
if i'd moved down a small notch to a centrino, i'd get 5-6 hours of battery life in a 5-6 pound package.
by the way, notebooks with wireless? priceless (of course, wi-fi will kill you quickly so don't you be getting that unless your time is up. a cable will keep you able). – by kev, a swing dancer
wifi will kill you (1:49pm est wed mar 09 2005)what bs, you nieghbors cordless phone illuminates you more than wifi and your tv a 1000 times as much.
ever get a radiation burn from the sun, most of us have.
how about a radiation burn from wifi, cordless phones, and tv screens?
didn't think so…
if 8 pounds is to much for you to carry from your desk to your bed your health problems go way beyond anything wifi can ever do to you.
add swing dancing to that, omg are you sure you're a man??? – by kev, a girly man
a girl's man (2:18pm est wed mar 09 2005)cordless phone – not so good.tv – x-rays a plenty.sun – melanoma from grand cayman visit as a child, recently removed.
8-1/2 lbs from desk to bed – i'm just saying it's not like hauling my palm m515 or reading a volume from the brittanica.
it's more than a gallon of cider, hot cider, and i don't want that on my lap for long.
i'm not sure i'm all the man i could be.
i do know that some women say their knees get weak when i hold them.
but as for my new dread of hauling an 8 pound notebook around, i suppose i could work out a little.
when i worked at westinghouse for a while, designing a robot arm to clean the interior of a nuclear heat exchanger, the rule was that no man could hold more than 5 pounds comfortably over his head for more than a minute or so.of course, their “men” were all union so that may not be a fair comparison.
but i don't know any culture on this planet that would consider a male who doesn't dance (or “move”), a man.
provider, maybe but a man, i don't think so.
but you know, if you're getting lucky, that's all that really counts. – by kev, a swing dancer
public power for notebook computers (2:26pm est wed mar 09 2005)one thing i've noticed since i've had my 1-hour notebook and the requisite power brick, is that libraries and all sorts of public places really are loaded with power outlets everywhere. – by kev, a swing dancer
yonah (3:17pm est wed mar 09 2005)“yonah will not, however, support amd64”
aw shucks,it won't run 16gb of ram on my laptop?how will i *ever* get any work done.
show of hands for those running out of memory with 4gb on board? – by amd64scam
here here amd64scam (3:29pm est wed mar 09 2005)i don't get the need for 64-bit extensions at this point.
it will mid year before you even have an os that can handle 64-bits and by then all of the 1st and 2nd gen amd64 chips are outdated by the new things amd has in the pipe.
on top of that it will be another year at least before you see any appreciable 64 bit aware software.
the performace boost on 32 bit application with 64 bit extensions is next to meaningless for most work.
add all this up and you see why getting a 64-bit system now isn't future-proofing, but vanity buying.
pretty much the same holds true for the dual cores (unless you need to multitask twenty projects), how many multithreading apps are there? – by pocketpc fan
what i want (3:32pm est wed mar 09 2005)design me a chip that runs around the level of a xscale with a power curve of around 2-3 watts.
then put it in a 65 nm package and bundle between 10-20 of them together as the processor.
now make the os aware of the processor array and allow it to turn off what ever amount of the array it doesn't need at the time.
not only does my app multithread through this array, but it also allows for the use of very little power when needed. – by pocketpc fan
ideal for warming your lap on winter :p (4:57pm est wed mar 09 2005)that's the use i give to my old piii laptop – by yuriha from m閤ico
wow, what silliness (6:44pm est wed mar 09 2005)and ignorance, even.
first of all, only the lv dothans dissipate around 10w tdp, the regular dothans have a tdp of 21w.
secondly, in low frequency mode (at 600mhz), both regular and lv dothans dissipate 7.5w. what does that mean? that means that a p4-ee is going to scale and perform at the low end as well as any non-ee yonah, but it will also be capable of scaling much higher for tasks that require a lot of computational power. i use a p4 laptop (a t30) for everything, and it used to last almost 3 hours on battery when i put the processor into a lower frequency mode, which was fine on the go because all i wanted was to play slime volleyball and listen to music, or similarly trivial tasks that did not require so much cpu power. when working with solidworks i switch it into high speed mode, which still isn't high enough for solidworks to perform particularly well, but it's functional. scalability is good.
and i think this is designed to be a step towards the transition from the netburst-style pentium 4s to a pentium-m style pentium 5 architecture on desktops. i think intel is wary of making claims at the moment because they are unsure of the capabilities of the pentium m architecture (the piii architecture, on which it is based, crapped out at 1ghz originally, and has taken some 8 years or so to reach 2ghz from 1ghz), but i think intel is considering positioning this new high-speed pentium m in the desktop market.
chronos- i agree entirely that intel is planning far too many new processor models. i get annoyed at the complexity and general silliness of their roadmaps. but i don't agree that intel should be rewarded for pulling any of this off. intel is a fine company, but their products are not top-of-the-line technology anymore, and most of their desktop plans are simply continuing the trend of, well, not excellence. i do think that their p-m plans are more reasonable, but if they can't migrate the superior p-m platform to the desktop arena than they deserve little praise.
as for rf being dangerous, well that's a rather silly notion, but not necessarily wrong. it is very hard to judge the effects of any one stimulus over the course of a human's life, because live's are so complicated. it is therefore impossible to determine whether or not prolonged exposure to consumer products that give off significant rf radiation has any significant impact. studies have been done, most of them came back saying that there was little-to-no measurable impact over the coarse of even years, but that doesn't mean there isn't some tiny cumulative effect that may prevent people of our generation from reaching the ripe old age of 100.
as for men dancing, well there's not only nothing wrong with that, but it's a good thing. do you think you're going to pick up women at a club by walking up to them and saying, “duuuuh, beer, sex, now”? even if you are not looking to pick up women, dancing is an energetic activity that men and women can appreciate together, with and/or without sexual connotation. i, for one, hate that our stupid puritan society looks down on dancing.
as for you amd64 scam and pocketpc fan, you do realise that winxp 64 has been in beta for, what, over a year now? microsoft releases all of its beta oses freely, so if you had bought an athlon 64 a year ago, you could have run a completely free and still legal windows os for over a year, while taking some advantage of the amd64 extensions. or if you want a more stable os, go for one of the dozens of alternative oses (including linux distributions) that have supported amd64 since before its introduction. what do you mean no os support? damn near every x86 and cross-platform os supports it. application support for windows is lacking, but you can still run 32bit apps, and most cross-platform apps can be found for amd64.
as for the xscale idea, well it sounds good, except that the arm architecture is very simple, which is great for microcontrollers, but not so much for big cpus. the arm has little to no mathematical computation capabilities, which could be resolved by adding dsps, but at that point you've just made a cell processor with an arm base instead of a powerpc base, and powerpc is far superior for higher-level computing.
and many, if not most, professional applications use multiple threads. on top of that, it doesn't matter whether they do or not, you will still gain from the use of multiple processor cores because it is not possible to have less than 2, probably not possible to have less than 20, threads running in a modern operating system. the additional core provides good, usable processing power that any application can take advantage of.
whew, ok rant over.
but seriously, this pm-ee stuff is pretty cool, i just hope it is reasonably priced.
– by drsengir
hmmmm (7:13pm est wed mar 09 2005)*puts his hand up* i am.
i think the turion will still whip the dothan, for a 2.5ghz chip to run on just 1 volt is amazing, especially when it dissapates just 29 watts.
and, i disagree with the idea that 64bit addressing and computation is a scam, not only is it a performance boost, but it allows alot more data to be crunched at the same time, allowing better looking games, more diverse renderings and so on.
– by headley
no 64-bit os's??? (8:28pm est wed mar 09 2005)i don't think pocketpc fan realizes that 64-bit os's have been around at least as long as the amd64, if not longer.
there's even a 64-bit version of windows xp available for the wonderfully low price of (are you ready for this?) free!
and xp-64 is fully 32-bit backward compatible.
however, if windows isn't your style, there's several 64-bit linux distros readily available.
i will concede that there aren't many 64-bit apps available yet, but that's certainly no reason to not enjoy an athlon-64 with xp-64 now – especially since the amd chip is already available, and will no doubt be much cheaper than any hack job intel manages to push out – ummm, next year? – by athlon girl
re: drsengir (12:04am est thu mar 10 2005)let me rephrase:
if intel can pull off every single processor on their road-map with descent success, they deserve praise.
but, if they fail on more than a forth of their road-map, they deserve get more bad press than microsoft.
back on topic:
i don't get where pocketpc got his info…i have been running windows xp-64 on my amd64 2800+ desktop for about a year now.
i have come across some software incompatibility issues, but i am trying to work through them as they arise.
i can't wait to pick up a copy of the full version when it comes out in april (as long as the price isn't astronomical).
i will be switching to a dual boot system windows xp-64 and fedora core 3 x86-64 :).
i will be running windows xp pro and fedora core 3 x86 on my other desktop, but only until i can upgrade the motherboard and cpu.
i think it will become my amd64 dual core rig :). – by chronos
alright (10:28am est thu mar 10 2005)yes, there have been 64 bit oses out there for a while… but i am talking about the windows environment which like it or not (i don't care either way) is where 99% of these processors are used.
until joe blow can go down to the local box store and get any number of programs for windows in 64-bit flavor the system is still a niche market.
windows 64 bit beta doesn't count, as it is a beta…
– by pocketpc fan
re: drsengir (11:09am est thu mar 10 2005)i think you misunderstood my processor design, i was not refering to an arm processor array, but rather a novel new architecture.
i agree with your assessment of the arm core and it is unsuited for the work.
what i want is a system with each element having the relative computing power of an xscale (reengineered to optimum use in an array) and yet lower power requirements per element than the cell system has been quoted at.
this is not an incremental change, it would require extensive reworking and would be more akin to the budget “supercomputers” labs have made with distributed computing than any particular desktop we have now. – by pocketpc fan
re: pocketpc fan (1:05am est fri mar 11 2005)on the xp64 beta, it is not bug-free, but neither is regular old xp. i have heard minimal compatibility issues from friends who use xp64 beta. i agree it is not viable for businesses, but most business computers don't need to be anywhere near the top-of-the-line anyways, and could stand to run on a celeron or sempron class cpu, which don't support amd64 extensions anyways. as for workstations and servers, well depending on the application, many of those run various unix flavours, many/most of which support the amd64 architecture. so i don't think it's entirely useless, but i do agree that it isn't a spectacular feature that we can't live without, and i agree that most consumers will not benefit from the 64bit extensions.
i see what you are saying now, sorry about the misunderstanding. but the cell still seems to be what you are talking about, with a number of small number crunching units on one die tied together through a powerpc. unfortunately getting the power dissipation or die size figures down is no trivial task. the whole chip has some 230 million transistors in it, with a high percentage of those (something like half) used for logic. that's only 25mtrans per processing unit, and only some 10mtrans per unit for logic, which is not very much nowadays. actually the ppc is something like twice the size of the vector units, so that's more like 46mtrans for the ppc and 23mtrans per every other processing unit. i think the cell is too ambitious in its allocation of silicon real estate and other resources, because that will severely limit the flexibility and usefullness of the processor, but the idea is sound. i like the powerpc, but perhaps they should have used smaller, simpler vector units to save on die space and power reaquirements.
and yeah, i like that distributed design, like modern beowulf cluster supercomputers, but with smaller, simpler components. unfortunately i think it would only serve to confuse the hell out of consumers, and windows would probably not support it.
– by drsengir
re: drsengir (11:09am est fri mar 11 2005)i just haven't figured out why we keep having to make processors bigger and badder.
the move to multi-core is a good one, but i think we need to go much farther.
the os needs to be able to reside in a single core (at a much reduced capacity obviously) but dynamically increase its capabilities as more cores are added.
if each core is a fraction of the size and complexity of the current beasts you could probably easily outpreform a modern x86 architecture or powerpc chip… and i am sure windows won't be doing this anytime soon, so it won't help any “far cry” games. – by pocketpc fan
64-bit intel (6:41am est tue mar 15 2005)i want to know when the pentium m 64-bit comes out, it should really come out no later then the release of longhorn… – by mrpatelio
It turns out Captain Picard is the best DJ
Intel preparing set top box unveil for CES
Holiday easter eggs show up in Google Search
Surface Pro 64GB to start at $899, coming in January
LIVE: Astronauts conduct repairs to the outside of the ISS
This site may earn affiliate commissions from the links on this page. .
subscribe to our newsletter:
This newsletter may contain advertising, deals, or affiliate links. Subscribing to a newsletter indicates your consent to our
and . You may unsubscribe at any time.
We have updated our PRIVACY POLICY and encourage you to read it by clicking .