Improving the PC as a gaming platform: the hardware


— 12:03 PM on December 28, 2012

So, we meet again. Last time, we were, ahem, discussing religion: the 10 commandments of PC games. That article focused on the oft-encountered annoyances in PC games themselves. Now, we're going to look at something a bit broader. We're going to explore how the PC can be made a better gaming platform, starting with the hardware.

Even die-hard PC gamers will recognize the, shall we say, essential issue with this platform: it's often far too difficult to (a) buy a gaming PC, (b) pop in a disc and play a game. For all their faults, consoles do many things right in this department. As a the quotation goes, "Good artists borrow; great artists steal." So let's see if we can't steal a few good ideas, shall we?

A hardware baseline
One of the main reasons people don't game on their PCs is because their graphics card is often more anemic than Kate Moss on a diet—and all hail Cthulhu if the processor doesn't play second fiddle to that. It's pretty hard to get excited about games when they run like molasses.

Old enough to remember computing in the 90s? That time when most computers only produced strings of beeps as sound, when 640x480 was a good new year's resolution (that pun is as old as SVGA, so I'm entitled to it), and when having an optical drive was a lustful pecadillo. Things we take for granted today, like spoken word and videos, were rare and exotic attractions. When technology advanced, the industry came up with a certification specification to ensure punters didn't miss out—and consequently spent more on better PCs. That spec was called MPC, short for Multimedia Personal Computer. The first version of the MPC spec said, in simple terms:

  • Thy computer shalt be blessed with a sound card and speakers.
  • Thou shalt be provided a CD-ROM drive in which to receive silver discs.
  • Thy processor shalt not be completely crap.

At the time, this spec meant a lot—and, to be honest, I think it worked marvelously. We need something like that again. People wanted MPC, everyone sold the better hardware, and everyone was happy. Let the powers that be come up with a new baseline specification. Call it MPC-HD or whatever acronym the marketing Nazgûl want to give it. I'm fine with whatever, as long as it gets the job done.

The new spec would clearly involve some compromises, since you can't simply step up and demand that every new computer feature a Radeon 7970, 32GB of RAM, and a six-core CPU with Hyperthingamabobs. However, let's take a page from our own System Guide's Econobox. MPC-HD could set the bar at, say, a Radeon 7770 graphics card ($120 or so) and a Core i3-3220 processor (around $130). Those components provide solid gaming performance at 1080p in the vast majority of titles, even with anti-aliasing enabled. They would be a perfectly reasonable baseline to aim for—one that provides many times the horsepower of current-generations consoles.

Setting a baseline would make life easier for developers, as well. Let's imagine MPC-HD has multiple levels, and when publishing your game, you can simply state that the minimum requirement is MPC-HD Level 1. That's easy for developers to code for, easy for buyers to follow, and easy for manufacturers to advertise and profit from. One can only wish.

Proper performance comparison mechanisms
Since I might as well ask for a unicorn, here goes another crazy idea: we need something akin to SPEC benchmarks for consumer hardware. There, I said it. Any moment now, I expect someone to kick in the door in and take me to the loony bin, tires screaming and sirens blaring.

In Wikipedia's words, the Standard Performance Evaluation Corporation is a "non-profit organization that aims to 'produce, establish, maintain and endorse a standardized set' of performance benchmarks for computers." SPEC benchmarks are used chiefly in server and workstation computing, and they come with very strict sets of rules and procedures. Now, Futuremark and friends already offer synthetic gaming benchmarks, but there's room for improvement. We need more true-to-life workloads and an emphasis on inside-the-second frame latencies.

I'm not advocating that SPEC itself take up the reins in the consumer space. The markets are fundamentally different, and the consumer space evolves and morphs at a much faster pace. However, if the gaming industry could come up with its own standards body, buyers would have a much easier time comparing different CPUs and graphics cards. Speaking of which...

Stop the numbering circus act with the graphics cards
Discrete or integrated, nobody knows what graphics solution to choose anymore. No, I'm not talking about people like you and me who actually read hardware reviews. I'm talking about the vast majority of the market, composed of regular, non-technical people who simply want to buy or upgrade a computer. That apparently simple task has become an ordeal. I mean, you know there's a problem when even my most computer-literate friends come to me for clarification—and I often can't help without looking at Scott's graphs.

Imagine yourself as a typical shopper. You don't read reviews, you don't know what's what, and you have no recourse except for asking the sales clerk—a move as wise as looking Medusa in the eyes and complimenting her hairdo. You read the labels around you: 7970, 560 Ti, 5850, etc. You go for a 5450, thinking you're getting a good deal. Surely it's better than the 4870 in your friend's old computer; it even has more of those gigabyte things. Except you unwittingly bought a piece of crap (for gaming, anyway). When the disappointment kicks in, you're going to throw your hands up in the air and say, "Why didn't I just buy a console?"

Sometimes, people don't even look at the model numbers. They think, "Ooh, an Nvidia graphics card with 2GB of memory," and they leave it at that. Those people inevitably end up with a weaksauce-flavored model packed full of RAM chips to trick the unwary. But it's not the buyer's fault, guys.

To the graphics cards manufacturers: get a grip. Here are three things that should, nay, must happen. (And yes, I know I'm asking for a lightsaber to go with the aforementioned unicorn.)

  • Model numbers ought to make sense in terms of performance, especially when going from one generation to the next.
  • Lower-end models ought to be clearly marked as such, so as to avoid confusion.
  • Card makers ought to stop packing huge amounts of memory onto otherwise useless cards.

The status quo might help you push low-end cards, but I'm very doubtful it's healthy for the market in the long run.

Stop the circus with CPU model numbers, too
When my friends bring up Mulholland Drive and sing its praises, I always challenge them to summarize the plot on a piece of paper for me. In the same vein, please explain Intel's model numbering scheme. Make it a pretty story. Couldn't figure it out, either, huh?

This ends up being a rehash of the graphics card situation. Model numbers never seem to match up with performance, so people have to look at other attributes like clock speed or core count. We then end up with talk like, "But this one here has four cores, surely it must play games better than the one with two"—or the good old, "But this one has more gigahertz!"

Sigh.

GPU makers need to get their act together on the driver front
Yes, I know, this is technically a software topic. However, graphics cards and their drivers are symbiotic.

I won't mince words: graphics drivers need work. Far too often over the years, I've run into bugs and compatibility issues. Insufficient testing often seems to blame. Way back when, before game-specific optimizations became the norm, testing wasn't too difficult. However, GPU makers' obsession with inflating benchmark numbers or FPS scores in the latest triple-A games has led us down this path, where driver issues have become the norm rather than the exception.

For example, I remember spending many, many hours troubleshooting an apparent overheating problem with a Radeon, only to find out that the driver's Catalyst AI (snicker) had a memory leak in Unreal Engine games. No, I hadn't fiddled with any settings there. I also had to play Civilization V in DirectX 9 mode for months because Nvidia's drivers decided to crash every time another civilization stopped by for a chat. To this day, a thread in our forums about AMD's aspect-ratio upscaling still gets hits because AMD hasn't fixed the problem. Similarly, my gamma settings still aren't applied reliably at bootup on my GeForce-powered system.

Many problems, so few solutions
It's 2012. PCs have been entrenched in homes for the better part of two decades, and these ridiculous problems persist. Does it really have to be this way?

These are tough nuts to crack, but they ought to be cracked sooner rather than later. PC gaming is in a position of strength right now, since current-gen consoles are stuck with circa-2005 hardware. However, the next generation is upon us, and it may offer enough of an improvement to make folks reconsider the value of their complicated and prickly gaming PCs.

My ideas are wild and crazy, but rarely do easy solutions appear for difficult problems. What are your thoughts?

   
Register
Tip: You can use the A/Z keys to walk threads.
View options

This discussion is now closed.