Two Sides of the Same Screen.

«»

1 2 3 4ALL

Tomorrow

Happy 5 kers: Here are the screenshots

Disgruntled 5Kers: Interesting.  So yours is throttling too.

Happy 5Kers:  Sorry?

Disgruntled 5Kers: Observe  the lower  meter.

Happy 5Kers:  The core trace?

Disgruntled 5Kers:  Yes, note where that line levels off, and then begins to oscillate?  It should a steady reading, virtually flat, like the memory.  These chips were binned at 850mhz though it would seem they perform heavy tasks hovering below 800.  Now look at where the fluctuations occur relative to your temperature, exactly the same as with me, high 90s to low 100’s.

Happy 5Kers:  I do, but that’s not throttling, at least not in the critical sense, its completely intentional.

Disgruntled 5Kers:  Am I speaking to Tim Cook’s iPhone or Tim Cook himself?

Happy 5Kers:  Nice one.  Really though, this is precisely what I meant by accurate and efficient methods to constrain energy consumption and excessive heat.  All the latest GPUs have it built-in.  The Kepler, The Maxwell, The Tonga and the rest.  Frequencies and voltages are dynamically regulated to ensure the cards remain within their respective TDPs and thermal boundaries, depending on which is violated.

Disgruntled 5 Kers:  I don’t need a lesson.  I’m an ex-PCer myself, defected to the lodge of the enlightened two years ago, I’m vividly aware of what’s happening and why.  The question is, should it be happening at all and moreover, under vanilla settings.    Should I have to intercept what is obviously a carelessly installed and slackly calibrated cooling system.

Happy 5Kers: 99% of the time you don’t have to, just let the limiter do its thing.  I know It sounds like a cop out but the granularity is so fine and the algorithms so agile that there’s no discernible performance hit.

Disgruntled 5Kers:  Subjective territory there I think.  In any case, thank you, you’ve now proven conclusively that my niggle is not isolated, we’ve all been taken for quinces.

Happy 5kers:  The niggle is your niggling obsessiveness.

Disgruntled 5Kers:  So you believe its normal and acceptable for a GPU to reach throttle governed equilibrium when tasked with anything more taxing than an email.

Happy 5kers:  Amusing, but also grossly exaggerative.  You had me launch two of the most notorious resource hogs in software history, composed with the sole purpose of lighting up the dashboard and the consequence is a minuscule reduction in GPU speed.   I’m more convinced now then ever that all is working exactly the way it should be.

Disgruntled 5Kers: Sorry, but a graphics chip that spends most of its mid to high load time spewing Fahrenheit faster than Brutal Doom spawns enemies will likely not yield long term stability.  How many times have we been assured all is peachy only to have peripherals buckle and burn within a year.  How much scrutiny do these units receive before they’re discharged? Remember the 2011 iMac? Distorted blue lines and artefacts or, if you were really lucky, a display as blank as a Hermit’s Twitter feed?  Goodness me, what a coincidence, another AMD GPU, and a mass recall. These machines not engineered to last, they’re mass produced to keep customers perched on the precipice of satisfaction until next year’s model is available.

Apple doesn’t care about sustained reliability.  Their designs, configuration and choices of parts revolve entirely around profit.  Most devices that live beyond the dawn of their successors only do so when Apple’s financial incentives happen to favour a manufacturer that supplies solidly machined parts. It was pure good fortune when Nvidia secured the contract for previous models.  I’d wager Apple’s testing parameters incorporate nothing other than what’s likely to expire within the initial year of ownership.  After that its a calculated risk as to who will actually bother to pursue a legitimate complaint and how many will be too busy, lazy and profligate to resist trading up for the present “latest and greatest”.  They gamble on our naivety and exploit our patience.

Happy 5kers.  Come on.  How often is a typical iMac user going to subject their rig to the sort of regime you’re contemplating? Prime 95 and Furmark morning, noon and night.  What did you order an iMac for? To stick in a server farm? Bit-coining?  And why single out AMD?  Nvidia has implemented  thermal and wattage perimeters on every one of their cards since the GTX 580. Even the Maxwell has a threshold of 80 degrees and when running Furmark, it too will throttle, just read Anandtech’s review.

Disgruntled 5kers:  Wrong that’s the Maxi’s default top temp.  It’s absolute limit is 95 deg, says so on nvidias site. You touched on GPU boost and Powertune earlier. Look carefully at Afterburner, notice those greyed out controls for Voltage and Power limit and how the core and memory clock sliders are dragged all the way to the right? On discrete Nvidia cards, all four of those fields can be modified.

There’s the option to manually raise the and lower the TDP and thermal thresholds over a specific range. For top tier 780s and 980s, power limits can be lifted to within 106% and 110% of their standard values and the temperatures elevated from 80 to 95 degrees. On top of that there’s Boost Clock, which pumps an extra shot of nitrous through the core under permissible conditions. On my previous iMac, I couldn’t change the voltage or thermal targets, but the GPU and DRAM fields were adjustable. Plus, I had Boost Clock, in other words like all Nvidia cards, the 780m was conservatively clocked, giving me headroom to work with and potential to engender a healthy FPS bonus.

Powertune works under similar principals, as you must know. The difference is that that the R9’s advertised figures for frequencies and temperature are their definitive parameters, meaning far less scope for supplementary pace other than at the expense of golden silence.

And that’s exactly where we stand. What use is 850mhz if a combination of mismanaged design, sloppy integration, lacklustre quality control – whether intentional or otherwise – and uncompromisingly stringent implementation, eliminates the slimmest chance of sustaining optimal performance for anything other than mundane desktop duties.

Happy 5kers: Prime 95?  Furmark?  What about Linpack or AIDA64 nobody with a speck of sanity would classify those apps as “anything other than mundane desktop duties”.  They demand  priority that’s off the scale and hardly reflect typical usage.  Who in the Milky Way spends whole weekends stressing out their comps with what are essentially diagnostic tools and only recommended for efficiently verifying legitimate defects.  Now, if I were getting visual corruption, sporadic reboots or everything was grinding to a halt within minutes of running them, then I’d be perturbed enough to complain, but as things stand, I feel more than fulfilled.

Remember when OCCT was the “gold standard”? That had a tortuous graphics test and front end for Linpack , which Intel themselves use to audition processors prior to release.  It was hilarious how many people were perplexed when their massively over clocked chips bailed out or suffered heat stroke quicker than a , when they hadn’t encountered a sniff of instability throughout months of work and play.  These programs were not intended to simulate a regular user’s requirements.  That fact that incessant tweakers got a hold of them was one of the reasons why manufacturers began taking practical steps to protect their products from misuse.

Try to see things from their point of view.  If I bought a car then immediately drove to my local garage and hooked it up to one of those rolling roads, you know, for monitoring horsepower.  Then, instead of taking measurements in the usual way,  I left it spinning on the rollers at maximum revs and a tyre blew. Should the manufacturer be culpable, or should I?  Is that part of a normal motorists daily schedule?  The rolling road to a car is the equivalent of OCCT or Furmark to a computer.

Fortunately for the automotive industry,  very few clients have the cash or space for one in their back garden, so they don’t have the means to thrash their hardware as frequently or relentlessly as as we do.  If they did, you can bet your life we’d see an assortment of pre installed governors, trips and limiters to prevent an avalanche of returns from compulsive budding mechanics on the basis of their turbocharged marvels being unfit for purpose.  We, by contrast, do have ways  to subject components to undue strain.  Prime, OCCT, Furmark and other similar utilities  have become de rigeur for the mainstream user.  Folks I know, perhaps like your good self, run them for hours and hours in succession and days on end.  They should use them appropriately, to diagnose faults, but honestly, it would appear they’d rather seek or, dare I suggest, actively encourage a foundation for dissatisfaction.

«»

1 2 3 4ALL

Pages: 1 2 3 4