The Case of the Befuddling Base Clock
By 2014, many initial doubters, some fonts of knowledge on popular podcasts probing cutting edge technology, were stubbornly shunning all Haswell wholesomeness in favour of their six year old i7-920, out of which they had cranked a shade south of 5ghz from an officially authorized 2.66. Meanwhile, there were others that had parted with pounds to secure “extremes” (the i7-965) , all of whom were engaged in an quintessential quintet of over clocking splendour, with the RAM , CPU, QPI, Base clock and Uncore multipliers agreeably divorced and awarded absolute autonomy.
The forging of Sandybridge in 2011 brought forth another Intel bait and switch special. The privilege of adjustable multipliers to main stream processors at the expense of a base clock tightly tethered to every bus in the chipset. The PCIe, USB, and SATA controllers had converged into a military parade behind the memory and CPU. Nudge the “base” drum one hurt too high and reduce the regiment into cacophony of brassy dissonance. Liberated processors were distinguished by a “K” at the end of their call-sign and artfully targeted at those with even the tiniest tendency to tamper.
Extravagant enthusiasts crossing the same bridge by “Extreme” express a year later, once again faired slightly better.
The introduction of ratios, also called “straps”, effectively created a”phantom” base clock linked solely to the CPU and enabled two elevated, “virtual” tuning thresholds that were transparent to the preposterously petulant peripheral buses, all of which remained supervised by the actual system clock .
Ah! Hold hard a moment. Barely parsecs ago there was just the base clock. Now we’ve got a phantom base clock and a “system clock”. Is this crude technocracy, or extraordinary attention to narrative variation with a yearning to evoke captivating suspense?
I know. Frustrating isn’t it? Technology’s most glorified gurus garner grand gratification from attributing names to things that either don’t exist or already have more names than an identity thief with dissociative identity disorder. Nevertheless, I’m afraid what was previously the “base clock” has evolved into asymmetrical twins and the consequences of expounding on the mechanics of two “base clocks” are sure to induce repugnant confusion. Therefore, to sustain clarity in our journey through the oncoming paragraphs, some revised definitions are imperative.
Base Clock 1: – This will be referred to as the “base clock” and randomly preceded by the word “internal” so as to emphasise its exclusivity to the CPU and isolation from external encumbrances. Picture it in the same way as you would closed circuit television, in the sense that its signal is not broadcast to the outside world or in the current context, the CPU’s hypersensitive surroundings!
Base Clock 2: This shall be called the “system clock” or “host clock ” and should be considered the “true” base clock. It is what remained visible to the rest of the motherboard along with its fleet of finicky buses.
Fear not, following a assortment of statistically fortified tables, your mind will be ineffably enriched. But for now, its…
Back to the Straps
Why, you may ask, was such a function devised? It would certainly have been useful for those impeded by imprisoned multipliers, but in staunch keeping with tradition, Intel had ensured that every shackled processor was also strapless and since unbound CPUs already offered comprehensive governance over their multipliers, where was the benefit in employing the base clock for more its bare essentials?
To paraphrase a cliché borne from journalisms tireless obsession with the negative , any granularity is good granularity.
Historically, processors of every age and pedigree have manifested richly varied patterns of behaviour when persuaded to violate their indigenous directives. Quite simply, they were complex creatures and for all who aspired to garner gigahertz, copious catalysts to customize generally delivered more positive returns than a BIOS whose menu’s were a wash of greyed out simplicity. One of the rules assiduously applied by specialists of many a generation, whether they be cranking the core or raising the RAM, was to keep every modification besides those necessary to nominate the desired frequencies as minuscule as possible.
In the case of Messrs Sandy and Ivy Bridge of non E descent, their cores were committed to scale in steps of 100mhz in accordance with their multipliers. A value of “x12” yielded 1200mhz (12 x 100) and one of x24 would give 2400mhz (24 x 100).
For fully audited members of the extremists’s lodge, these supplementary straps of 1.25 and 1.66 – some called them gears – allowed for previously untenable “effective” base clock speeds of 125 and 166 mhz to serve as raised starting points in addition to the 100mhz default. In other words, the CPU’s multiplier theoretically possessed markedly greater manoeuvrability and users could meticulously manipulate the cores’ frequencies over three separate ranges of approximately 15%.
It wasn’t until mid 2013 and the début of the intensely hyped Haswell the Intel saw fit to furnish his flourishing main-stream fraternity with the same delights.
The Haswell-E platform was intrinsically no different, and by this time, venerable motherboard moulders were vastly more enterprising, ambitious and perceptive, with many proficiently re-packaging the facility as an attractive USP.
The imperious, octo-dexterous 5960x – Haswell-E’s holiest – was stated by numerous impartial reviews to have obtained speeds approaching an impressive 4.5ghz . Promptly following its release and a brief spell under the fingers of feverish fiddlers, these claims were unanimously regarded as genuine. Such a pronounced injection of unconditional pace might have hovered beyond reach were it not for the base clock’s higher reference levels being successful in retaining a degree of unity.
The sole and risky option would have been to set the multiplier to x45, leave the host clock chugging along at 100mhz (45×100=4500mhz) and prey that no unpleasant bolts from the blue conspired to plague the poor ham strung user for imposing settings that strayed so far from stock.
However, in ascending to the middle strap of 1.25, hence triggering an internal base clock rate of 125mhz, this same tenacious technophile could decrease the multiplier to x36 (36×125=4500mhz), thereby ensuring closer alignment with its factory obligation but at negligible expense to speed.
The second and more pressing problem partially alleviated by our metronomic mediator concerned the accelerating evolution of memory. As new standards of RAM were conceived and commercialized, they were accompanied by a selection of independently approved and universally compatible defaults, ostensibly to avoid misunderstandings between manufacturers, whole-sellers, consumers and any other indispensable elements in the retail process.
The organization enlisted to formalize these vital guidelines was J.E.D.E.C. established in 1958 and actively engaged in similar roles throughout the electronics industry ever since.
J.E.D.E.C. ‘s draconian deliberations eventually divulged official recommendations for voltage, bandwidth and latency and were a conservative precursor to the aforementioned XMP profiles, formulated by Intel and pioneering memory vendors. Once formally ratified, they were indelibly written to the SPD Roms on the relevant parts during OEM production. Thereafter, it was up to illustrious brands such as Corsair, G.Skill, Patriot and others, to liaise with Intel and embrace direct responsibility for any optimistic deviations.
Did this deter their confidence? Not for a nano. Those Devilish DIMM die-hards abandoned convention more readily than a graffiti artist in the Pope’s bathroom and hastily unleashed their mightiest modules. According to Jedec, the maximum enforceable frequency for a juicy stick of DDR4 to yield a respectable combination of durability and fleetness was 2133mhz, exactly the same as they had endorsed for the final incarnation of DDR3.
In the initial period of DRAM’S fourth tenure, Corsair’s stunning stall consisted of several packages. Their revered “Dominator” lineage had sprouted a premium “Platinum” off shoot, which the company continued to cultivate with considerable conviction.
Advertised speeds ranged from at 2400mhz, already well in excess of JEDEC’s declarations, right up to an unprecedented 3400mhz for an limited set of four by fours arrayed in scintillating heat-spreaders of cinnamon-orange and reportedly verified to operate at their rated optimum by Corsair themselves. A rusty piece of DIMM and distant history for a peachy “Grand Mariner”, see what did there? Obscure allusions to the colour “orange”? Well, what better to stir up memories than memory itself?