More “Passive” Pixels.
There is a popular saying that I first chanced to hear during the opening story of a highly regarded British portmanteau horror film entitled “The House That Dripped Blood”. The story concerned a struggling author attempting to compose his latest masterpiece about an escaped homicidal maniac who prowled the countryside for hapless prey that he savagely strangled to death before laughing harder than a hyena at a heard of submissive wildebeests. The saying goes “one picture is better than a thousand words”.
Almost half a century after this flick fulfilled the cravings of copious scare starved fear fetishists and some 90 years after the said phrase bloomed into a full blown cliché, humanity existed in an era where the importance of visual stimuli took such marked precedence over printed text that the latter’s elevated rarity caused many disgruntled linguists to state that the proverb should be revised or rather, reversed to…
“One word is worth 10 thousand pictures of tacky weddings, drunken parties and poorly framed plates of pasta in front of ill-focused landmarks.”
The original quotation is often erroneously credited to Frederick R Barnard, who applied its earliest version “A Look is worth 10 thousand words” in December’s edition of “Printer’s Ink” in 1921 followed by the famous paraphrase in the same publication in March six years later. No definitive author has yet been identified, though an article in the New York times pre-dating both of Barnard’s references suggest its source to reside within Japanese folklore.
So, how exactly does this convoluted preface take root in the immediate context? What possible reason could its conveyor have for such a cumbersome and superfluous introduction. A conveyor who’d spent most of the preceding three months lurking on the fringes of cutting edge computing and taking multiple tumbles of an ever expanding collection of bicycles. Sloppy thinking, concussion, a general disinterest in the core subject matter of this lacklustre example of early 21st century technology journalism?
Perhaps it was a cocktail of all three. For when any nerd worth half the silicon on their brain elects to address the release of a major component, they enter a state of profound reflective prior to committing characters to the ether. A pattern of thought that forces them to question the aggressively circular nature of their sporadically financed hobby and consequently, of life itself.
In an industry more fickle a cat in a crowd of kindhearted pensioners and as repetitive as time delayed cable channels showing repeats of quiz shows, the question of how to combine insight with engaging originality shall remain an eternal enigma.
So, imagine my myriad emotions when I learned that our two slumbering giants of graphical stardom had stirred, roused their ruthless ranks of polygons and were pitching toxic teraflops across a battlefield scarred by decades of bitter and bloody rendering. Think of this jaded hack’s crestfallen consciences when the garrulous guff which can be taken with a generous fistful of sodium gave way to cold hard statistics and the necessity to sensationalise a slab of PCB with a slightly speedier square silicon on it than the slab it is replacing became impossible to prevaricate.
Interpreting various quantities of micro-components is scarcely a task that stimulates radical creativity. Beyond those raw and ravishing figures, two fundamental choices exist. One, determine and reveal how the latest tsunami of transistors translates into real-world performance and two, explain in fantastically scientific detail precisely why this performance has once again increased by just right percentage to ensure that all credible tech sites presented with review samples are at least 90 percent satisfied.
Considering that 99 point stupid percent of enraptured enthusiasts could’t care less about the latter and the former could be succinctly conveyed in a single sentence, one cannot help but be baffled by the hardware hack’s eternal desire to swoon over data, features and specs throughout podcasts longer than a Wagnerian opera and articles that make a dissertation on astrophysics read like the idiot’s guide to a black hole.
So, exactly how does this distended preface relate to its core content? Simple. There isn’t any. Core content that is. Sorry. Not this time I’m afraid. Not next time and likely not thereafter, or at least until its creator is blessed with prophetic powers, for that it the only way he will be able to divulge anything but desperately aged news.
You see, in 2016, the year these ponderously chosen words drifted into cyberspace, the digital rumour mill ground out gossip, leaks and previews that, far from incendiary exaggerations, proved comprehensive and accurate enough to supplant the vast majority of official publicity, leaving formerly indispensable reviews suspended in pathetic obsolescence, other than for a minuscule proportion of elite technologists seeking PhDs in advanced silicon engineering. That is why in the immediate context, one picture will speak for one thousand podcasts:
And here it is.
So, do you now comprehend my reluctance to dig this subject deeper than a coal mine in a ravine, or more fittingly, then its own family tree? Can you spot it? That miraculous feat of micro-engineering? Pause for a moment to assimilate its unique elan. Marvel at how its bold and radical design commands immediate attention, despite an illustrious heritage in close company.
Observe how it stands apart. The chassis moulded masterfully to elegant and refined proportions. The finesse, an unparalleled embodiment of enchanting and extraordinary craftsmanship. The embossed Geofrce logo pulsating with verdant vibrancy, offset sharply against a stark onyx backdrop. The top, an exquisite encapsulation of aluminium allure, with the heat sink’s nickle coated fins shimmering through a flawless expanse of robust acrylic. The squirrel cage fan, a revolutionary evolution in the concept of rotational heat regulation, engineered to exhaust degrees as efficiently as dispel pretentious sarcasm…
Whoosh! There it goes, and with no further time to squander, here come some hot soft facts.
Oh here we go! It’s one of those tortuously dense diagrams that looks like the contents of a Lego slime monsters handkerchief. And just when you thought your crippled corneas were to be spared a shred of mercy, here’s another one.
At this point in the faint and far off past, any articles or podcasts featuring these slides would segue into an exhaustive explanation regarding clusters of cores and wadges of cache harboured in a hoard of streaming multiples.
Thankfully a delicate infusion of aggravation and common sense, evoked the inspired notion that a regurgitated thesis of nano jargon might be better expressed by labelling the diagrams and linking to previous explanations. GTX 680 GTX Titan GTX 980 GTX Titan X.
If you cannot muster the patience to delve into the past, then I pray you already possess the knowledge to comprehend the following, for if not, your patience shall wear ever thinner.
As legions of graphical gurus observed with baited breath, their beloved leader bade a belated farewell to an era that had played host to two major architectural transitions and spanned over four years. In June 2016, after aeons of customary leaks, rumours and frantic forum fiction, his brooding diligence culminated in the creation of Pascal, his swiftest and sveltest GPU to date.
Its cause, unsurprisingly, was to augment performance in recreational simulations that typically involved destroying legions of living organisms and massive chunks of planets, whilst assuaging the wrath of eco-warriors by consuming as little energy in the process. It achieved this ironic feat via assiduous loyalty to its mechanical heritage with the appreciable exceptions of an attenuated die, two additional Streaming processors and energy footprint more frugal than two regular butcher’s bulbs (for reference, that’s 200 watts).
The sole gripe that even the most rose-tinted critic could rationalise was Old Green Eye’s reversion to conventional DDR memory instead of the radical “High Bandwidth” breed premiered by his Ruby Rival’s “Fiji” chip a year earlier, though this shortfall stemmed from canny commercialism and not incompetence. DDR5X, an enhanced refresh of its ancestral counterpart, boasted twice the bandwidth per pin and could be implemented using established designs with no major physical alternations, thereby minimising production and purchase costs.
Moreover, enthusiasts who braved bankruptcy to boost their benchmarks were endowed with an 8 gigabyte frame buffer, doubling up on the Fiji, which had suffered notable criticism for stuttering when tasked to render the richest eye-candy at optimum resolutions.
Nevertheless, in relation to it’s predecessors, the Pascal’s genetics had evolved as much a fossil in a stasis chamber, thus in a bid to preserve precious linguistics, we shall proceed with another paraphrase of this article’s founding motto.
Behold, for one table is worth a thousand white papers.
Notes on noteworthy upgrades.
The hardware hack’s existence is a tormented one.
Whether or not they are professionals acting for prominent news sources, or extrovert v-bloggers promoted and bankrolled by fellow hobbyists, each is doomed to ardours hours of dragging sliders, tapping run buttons, realising they’ve forgotten to check the box marked “Hyper Quasi Rotoscoptic anti-aliasing”, a feature which last years GPUs are perfectly capable of but for now remains exclusive to those whose phobia of jagged edges takes precedence over their overdrafts.
Discovering that enabling this function adversely skews results, re-running the benchmarks, hitting reset after attempting to activate a triple screen setup and being “blanked” on all three counts, enduring cries of bias from furious fanboys for snubbing “optimised” drivers that were released five minutes after your review was published.
Conducting the tests yet again only to reveal little if any performance discrepancy, tweaking power, fan speed and temperature via a convoluted thermal control system which the manufacturer claims is revolutionary but really just makes makes things louder and hotter. Testing the cards in a case, on a bench, or in the freezer. In twos, triplets and quads.
Extracting data from video files big enough to cripple a corporate server farm and despairing when the specs of refreshed parts have are swiftly leaked into the ether and that in a matter of weeks, this tortuous nightmare will begin once again.
Weather a clinically cultivated office or catastrophically cluttered man cave, abundant fallout from aeons of a geek’s misadventures is inevitable. Leaning towers of boxes serving as dog-eared timelines of a transient industry and yet, arousing memories as sacred to them as those conjured by their grand mother’s glass menagerie.
Once The Pascal’s NDA had been lifted and billowing banks of blue smoke regressed to reveal scuffed worktops strewn with scarred chassis, stray thumbscrews and useless user manuals, the following granules of veracity became immediately apparent.
When it came to racing around the cape of Cuda Carnage, Old Green eyes newest and thriftiest flagship carved out a 30% lead over S.S. Super Maxwell, whilst plain old Maxwell House chugged through a choppy wake a further 40% behind it’s unbridled successor.
Two way SLI continued to showcase Nvidia’s mastery over the sweet science of scaling, provided the Pascal’s ravenous rasterizers were blitzed with a bazillion pixels to prettify.
NVidia were swift to discourage investing in triple and quad card configurations on the basis that all gaming support for such environmentally reckless profligacy was to cease.
The strident majority of swivel chair cynics welcomed the emerald empire’s continued efforts to preserve mother nature’s paradise but for contrasting and selfish reasons.
“3 and 4 way SLI” they claimed, “Was only ever for compulsive narcissists whose irrational pride was nourished by their rigs attaining benchmark results that 99.9% of their 99 thousand infatuated followers could only get within 99 country miles of if they sold their houses and a selection of internal organs.
In truth, a large proportion of these sneering anti-socialites had long lusted after equal luxuries and were secretly rejoicing that with their cessation, also came relief from restless nights terrorised by the proverbial green-eyed monster.
To further vary the digital vista, such nascent VR headsets as HTC’s Vive or the perpetually hyped Occulus Rift were ill-equipped to exploit multiple GPUs, leading to rife speculation that Nvidia may soon be to compelled to prioritise solo performance.
For rampant renderers, spoilt science students and AI aficionados with a penchant for profound learning, the company initially planned to issue a unique “enthusiasts key” available to users upon request that could be downloaded and applied to the firmware of individual cards, thereby unleashing their powers of parallelism when in groups of three and upwards. However, it was subsequently decided that the cost and complications of implementing such an arcane procedure were not worth the additional influence it would afford the corporation over such relatively scarce communities.
Besides, a large chunk of supplementary resources had already been allocated to stylising a fashionable “High Bandwidth” SLI bridge, deemed essential to ensure data integrity and optimise frame rates when combining cosmic resolutions with reams of ocular opulence.
The bridge was roundly condemned by wiseacres an elaborate gimmick to scam the wealthy and gullible, though several forensic studies appeared to demonstrate that its presence did yield significant benefits under the stated conditions.
Another feature that none besides certified insiders were competent to clarify in several hundred PowerPoint slides was “Fast Sync”. A sophisticated hardware level frame pacing system intended to complement Nvidia’s proprietary G-Sync technology.
Under circumstances where the GPU rendered frames at speeds exceeding the monitor’s refresh rate, those that would previously have been transmitted to the display in between scan cycles and appeared torn to the user were stored in a virtual buffer and selectively designated for output to the primary buffer with remainder being discounted. The result, in theory, was an image free from the imperfections induced by a lack of traditional v-sync as well as the input latency commonly associated with its presence.
For those still sceptical of the Pascal’s carbon cutting credentials, one final party piece stood poised in its pipelines. Prior to the of existence of Symmetric Multi projection (SMP), any enthusiast looking to broaden their vision via the virtues of Nvidia surround required a video subsystem potent enough to process multiple copies of the same image and transfer specific portions of each to to the central and peripheral displays. SMP surmounted this cumbersome drawback with remarkable simplicity, doing for triple-panellists what a holographic mirror would for a dodgy chorus dancer.
Instead of several copies of the same image being processed separately, that image’s geometric parameters could be cloned, cropped and allocated in accordance with the entire scene’s composition and the number of monitors it occupied. Since this procedure was occurring before the data was delivered to the GPUs rasterizing and shading engines, only one effective image required their attention, thus, both workload and power consumption were significantly reduced.
Before you slump into stupor, please rouse your ravaged retinas for long enough to assimilate that approximately 2500 words and six pictures have elapsed since this article’s contrived introduction An average reader’s journey to this closing paragraph will have occupied roughly 20 minutes of their superfluous time, though a genuinely “average” reader would have very likely scoffed at the first sentence before indulging their considerably trendier tastes with snarky swivel chair punditry regarding diets, celebrities on diets, celebrity pets on diets and dietary advice from celebrities with allergies to everything except exotic Burmese cat hair and their own eating disorder. For the rest, who prefer to digest specs and stats from podcasts comprised of over 8000 words and no pictures whatsoever, the tech word’s tenacious task forces were all too delighted to oblige.
Note that if you are clicking these links several centuries after this article’s publication you may be confronted with such harsh truths as.
“The page you are looking for tragically perished beneath a tsunami of inevitable complications born of man and womankind including but not limited to unsustainable resources, a lack of perceived popularity, cyber rogues with idle hands and the existence of corrosive elements.”