The Gigahertz Race is Back On

An anonymous reader writes "When CPU manufacturers ran up against the power wall in their designs, they announced that 'the Gigahertz race is over; future products will run at slower clock speeds and gain performance through the use of multiple cores and other techniques that won't improve single-threaded application performance.' Well, it seems that the gigahertz race is back on — a CNET story talks about how AMD has boosted the speed of their new Opterons to 3GHz. Of course, the new chips also consume better than 20% more power than their last batch. 'The 2222 SE, for dual-processor systems, costs $873 in quantities of 1,000, according to the Web site, and the 8222 SE, for systems with four or eight processors costs $2,149 for quantities of 1,000. For comparison, the 2.8GHz 2220 SE and 8220 SE cost $698 and $1,514 in that quantity. AMD spokesman Phil Hughes confirmed that the company has begun shipping the new chips. The company will officially launch the products Monday, he said.'"

Comments

More Power for What?

[pipingguy]

This reminds me of the sign at the local breakfast shop (paraphrased): "Use coffee: do stupid things faster".

Yeah, this is cool, no doubt. How many users actually *use* how much power they already have? I use a lot, but it's mostly dependent on the graphics card.

Re:More Power for What?

[zaibazu]

Displaying myspace profiles. The CPU load they produce is astonishing.

Re:More Power for What?

[Name Anonymous]

How many users actually *use* how much power they already have? I use a lot, but it's mostly dependent on the graphics card.

You're correct that people don't need this much power for their desktops but there are still plenty of uses for more speed in servers and for certain other applications. Actually I think the correct phrase is "most people don't need.... and at that, it may be inaccurate. Someone who does heavy video work can certainly chew up a lot of processing power. Heavy image work can use a lot of prcessing power in bursts.

Then there is the big fact that progammers these days are sloppy and waste resources. A machine that is faster than one needs today will only be adequate in 2 or 3 years given upgrades to all the programs. (Am I being cynical? Maybe, but then again, maybe not.)

Re:More Power for What?

[billcopc]

You know what's funny about Beryl/Vista ? They're doing the same stuff the Mac was doing years ago on puny hardware. I mean really, how frickin' hard is it to draw a window as a texture on a pair of triangles ? Seriously!

Programmers are sloppy, because sloppy is all the industry wants to pay for. Way back in the day when CPU cycles were super expensive, programmers were paid better money and given the time to tweak the crap out of everything, because if they didn't, the app would run dog slow and people wouldn't buy it. The problem is that somehow, people now tolerate underperforming software. They see it as a reason to upgrade... good god, they actually fall for it! Gee I certainly remember surfing the web on a 486 with 8mb of Ram back in the day. Now my OS needs a good 50-60mb to itself, and that's after I ripped out all the cruft. Normally it would be 100mb just for sitting idle with a background image and a neon-colored task bar. Gee uh, where'd all my system resources go ? Does it really require 7.3 million bytes to house a TCP/IP stack when some embedded devices pull it off with oh, 6kb or so ?

The truth however, is that if we were to write code as tightly and meticulously as we did in the 80's and 90's, software would perform, on average, at least 5 to 10 times faster than today, excluding hard bottlenecks like disk access and network bandwidth. It would also take 50 times longer to write the software, and I'd say less than 1% of people who call themselves "programmers" are even able to write such finely tuned code. Everyone doing VB ? Out. Everyone doing RAD ? Out. All you Ruby on Rails weenies ? follow me to this dark alley *BLAM*

I remember spending hours on little loops, with a CPU reference manual and a calculator. Sometimes I did little time sketches to figure out the best way to stagger memory accesses so as to not starve the execution pipes. Often times that meant weaving two disparate functions together, one being memory-hungry, the other CPU hungry. Together they filled each other's latency pockets, and my routine ran nearly thrice faster as a result. No C compiler I've ever seen could do such kinky things. Heck one time I even wrote a little assembler demo whose code executed twice: forward, then backward. The opcodes and data were carefully selected to represent valid instructions when reversed. It was more than a nerdy trick, it allowed my routine to fit entirely in the CPU's on-die cache, which gave it a huge speed boost but more importantly, it enabled a lowly 486 to mix 48 sound channels in real-time. Today's Cubase can't even handle a couple dozen channels without stuttering and/or crashing, on computers over 100 times faster than a 486.

Gigahertz a bad thing?

[jibjibjib]

I was kind of hoping the gigahertz race would end so Microsoft would have to stop making each version of Windows slower than the last.

Re:NO NO NO NO NO

[Soul-Burn666]

So take a top end 3GHz model and underclo it and reduce its voltage. You still get good performance, with lower power consumption.

You'd be surprised

[Moraelin]

You'd be surprised how much more _can_ be made with a CPU.

E.g., sure, we like to use the stereotypical old mom as an example of someone who only sends emails to the kids and old friends. Unfortunately it's false. It was true in the 90's, but now digital cameras are everywhere and image manipulation software is very affordable. And so are the computers which can do it. You'd be suprised the kind of heavy-duty image processing mom does on hundreds of pictures of squirrels and geese and whatever was in the park on that day.

And _video_ processing isn't too far out of reach either. It's a logical next step too: if you're taking pictures, why not short movies? Picture doing the same image processing on some thousands of frames in a movie instead of one still pictures.

E.g., software development. Try building a large project on an old 800 MHz slot-A Athlon, with all optimizations on, and then tell me I don't need a faster CPU. Plus, nowadays IDEs aren't just dumb editors with a "compile" option in the menus any more. They compile and cross-reference classes all the time as you type.

E.g., games, since you mention the graphics card. Yeah, ok, at the moment most games are just a glorified graphics engine, and mostly just use the CPU to pump the triangles to the graphics card. Well that's a pretty poor model, and the novelty of graphics alone is wearing off fast.

How about physics? They're just coming into fashion, and fast. Yeah, we make do at the moment with piss-poor approximations, like Oblivion's bump-into-a-table-and-watch-plates-fly-off-superso nic engine. There's no reason we couldn't do better.

How about AI? Already in X2 and X3 (the space sim games) it doesn't only simulate the enemies around you, but also what happens in the sectors where your automated trade or patrol ships are. I want to see that in more games.

Or how about giving good AI to city/empire building games? Tropico already simulated up to 1000 little people in your city, going around their daily lives, making friends, satisfying their needs, etc. Not just doing a dumb loop, like in Pharaoh or Caesar 3, but genuinely trying to solve the problem of satisfying their biggest need at the moment: e.g., if they're hungry, they go buy food (trekking across the whole island if needed), if they're sick, they go to a doctor, etc. I'd like to see more of that, and more complex at that.

Or let's have that in RPGs, for that matter. Oblivion for example made a big fuss about how smart and realistic their AI is... and it wasn't. But the hype it generated does show that people care about that kind of thing. So how about having games with _big_ cities, not just 4-5 houses, but cities with 1000-2000 inhabitants, which are actually smart. Let's have not just a "fame" and "infamy" rating, let's have people who actually have a graph of aquaintances and friends, and actually gradually spread the rumours. (I.e., you're not just the guy with 2 points infamy, but it's a question of which of your bad deeds did this particular NPC hear about.) Let's not have omniscient guards that teleport, but actually have witnesses calculate a path and run to inform the guards, and lead them to the crime. Etc.

Or how about procedurally generated content? The idea of creating whole cities, quests and whatnot procedurally isn't a new one, but unfortunately it tends to create boring repetition at the moment. (See Daggerfall or Morrowind.) How about an AI complex enough to generate reasonably interesting stuff. E.g., not just recombine blocks, but come up with a genuinely original fortress from the ground up, based on some constraints. E.g., how about generating whole story arcs? It's not impossible, it's just very hard.

And if you need to ask "why?", let's just say: non-linear stories. Currently if you want, for example, to play a light side and a dark side, someone has to code two different arcs, although most players will only see one or the other. If you add more points and ways you can branch the story (e.g.

Re:You'd be surprised

[cerberusss]

nowadays IDEs aren't just dumb editors with a "compile" option in the menus any more. They compile and cross-reference classes all the time as you type.

I program in C, you insensitive clod!

Re:NO NO NO NO NO

[level_headed_midwest]

Low thermal dissipation is a much more prevalent theme for AMD than it is for Intel, especially outside of the notebook sector. Yes, Intel has some 1.06 and 1.20 GHz Core 2 Duo ULVs for laptops and those have a 10-watt or so thermal dissipation while AMD's lowest-TDP mobile chips rate in at 25 watts (Turion 64 MT/Sempron.) Intel also has the single-core Core Solo series at 1.06-1.33 GHz that dissipates 5.5 watts. However, those chips are very rarely seen in any notebooks larger than a 12" screen size. You'd be much more likely to see a 31-watt Core Duo or 34-watt Core 2 Duo sitting in an average laptop than a 10-watt C2D ULV. AMD's Turion X2s have similar TDPs, ranging from 31 to 35 watts. All of the processors have similar frequency and voltage scaling mechanisms and battery life is roughly similar.

For desktops, most of Intel's newer Core 2 Duo processors have an average thermal dissipation of 65 watts. The fastest Core 2 Duo, the 2.93 GHz X6800, has a 75-watt average TDP. The quad-core chips range from 105 watts for the 2.40 GHz Q6600 to the 130-watt QX6700. These chips have a very reduced version of the SpeedStep that Intel puts in its laptop chips. The lowest core speed of the 800 MHz FSB chips is 1.20 GHz and 1.6 GHz for the 1066 MHz FSB chips. AMD's current new desktop processors start from a maximum thermal dissipation of 35 watts for the single-core Sempron EE and go up to 45 watts for the Athlon 64 single-cores (Lima), 65 watts for the Athlon 64 X2 models from the 3600+ to the 5200+, 89 watts for the 5400+ and 5600+, and 125 watts for the X2 6000+ and FX-70 series. The AMD chips all clock down to 1 GHz at idle. AMD also rates the chips on their absolute maximum thermal dissipation rather than an average thermal dissipation like Intel does, so a 65 watt AMD chip will usually end up drawing less power than an Intel 65-watt chip. The AMD chips also draw significantly less power at idle due to their lower clock speed.

The scenario is much the same for servers. AMD has their High Efficieny line of dual-core chips that draw 68 watts, the normal line that draws 95 watts, and the SE line that draws 125 watts. Intel has a few low-voltage Xeons, but those are very uncommon and pretty much limited to blade server vendors. AMD sells its Opteron HEs through a wide range of vendors.