Windows gaming, 64-bit battle, OSX

The first two relate to Microsoft Windows XP slows down gamers

Hi Mr. Editor dude

Reading through the story now, about the refresh rate for Windows XP and OpenGL - I thought that I'd just mention that this problem can be fixed by turning off V-sync.

Again though, you'd need a third party application, or you could use the official Detonator drivers from nVidia's site. These, unlike the ones included in Windows update, allow you to change settings in the nVidia control panel under display properties, but the drivers themselves are not WHQL certified.

Many gamers out there though, run the latest possible drivers, even beta ones.

Just thought that it would be worth a mention.

Cheers

Christo van Gemert - Staff Writer - PC Format Magazine

Says Dennis Jakobsen:

This is an issue in ALL versions of windows.. well.. since directX and OpenGL games started..

You can override the refresh value in the DirectX control panel.. it just doesn't work.. it's still 75.. :)

Quake3 and all Quake3 engine based games (RtCW, Jedi2, MOHAA, etc.) do have the ability to set refresh rates for OpenGL. r_displayRefresh "xx" is the command to do it.. replace xx with your favorite refresh rate of choice..

But it has even more implications than you mention in your article. afaik all drivers have vertical sync turned on by default to avoid tearing. this 'feature' in combination with a refresh of 60 also gives a max. fps (frames per second) of 60.

I know that DCC/DCC2 is some sort of monitor Plug'n'Play thing, that WindowsXP uses to set the maximum refresh rate at any given resolution.

Why this is not being used for games in both DirectX and OpenGL i have no idea..

ARRON ROUSE generated a heap of letters with Giant Killing: the 64-bit desktop

Arron,

For the desktop and workstation world, it is my belief that processing power is less an issue than the ability to connect up at high bandwidth the various processing resources that are available.

This is one of those cyclic things, but in this particular case, access to memory and to other outside processes has fallen far behind the growth of processor power.

Many of the problems that we face, such as voice processing, are limited by their inability to handle large amounts of RAM, rapidly. We are trying to make up for this with algorithmic sophistication, but the problem is not going to go away. HT is one of the things that will help this situation, if not solve it. Further parallelization of processes is another, and things like HT aide the efficient and inexpensive accomplishment of that role.

Another area that is strongly affected by this type of problem is robotics.

You actually could extend this to any area that has strong, real-time demands where latency is a major issue. It will be interesting to see what consumer level applications are enabled by this new fabric of processing.

Everett L.(Rett) Williams

"XANREL writes:

"If AMD can't make a real dent in corporate sales straight away, that leaves the consumer market. You can put money on it that a 64bit version of Doom III will run a hell of a lot faster than the 32bit version. The hardcore gamers will be falling over themselves to upgrade."

You've gone the wrong way to come to the right conclusion. A 64bit processor doesn't actually run normal programs faster. A 64 bit program runs the same on a 64bit processor as a 32bit program does on a 32bit processor, assuming the only difference between the processors is the number of bits.

There are two real purposes for 64bits:

The first and primary reason for consumer use is memory addressing. 32 bits = 2^32 = 4294967296 possible combinations of 0's and 1's. This means that since we need an address for every byte of memory, we can only have 4294967296 bytes (4GB) of RAM. This is certainly not a problem today where even the most taxing games might use 1GB during their highest-stress modes and even then very, very rarely, but give it a year, maybe two and the top of the line game will start throwing out of memory errors on any machine with less than 33 bits. Remembering that each bit doubles the number of numbers you get, 64 bits gives you 18446744073709551616 (18EB), so don't expect to run out of address space on a 64-bit machine for a long, long time. This is also why big servers need 64-bit processors: 4GB of RAM just doesn't cut it for anything high volume.

The second reason is simply large numbers. If you routinely deal with numbers in excess of 4.3 billion then 64-bits is faster. That doesn't mean that 32-bit machines can't deal with numbers that big, just that they have to fsck around doing more operations any time you want to do anything with such a large number. Consumers don't really have any use for numbers that big, so it's not really that important, but this is the main reason almost the entire scientific community runs on 64-bit machines.

AND JOE Kraska writes:

I read your article regarding the Hammer, and I think you're spot on. Intel has screwed up in a major way. Two decades of semiconductor evolution has demonstrated a very clear trend: the consumer-oriented products chase the high-end, acquire what was previously conceived of as a performance computing capability, and move upward.

Only very rarely does it really move the other way, except as R&D which gets eaten up by the consumer-oriented technologies.

Intel should have known this; their entire prior path to success was predicated on it, as a matter of fact. They elected to depart from the prior pattern, and boy oh boy are they going to pay.

MICHAEL PETTENGILL says:

Good article, but there is a nail for the coffin that you didn't mention.

Check out "The Innovator's Dilemma: The Revolutionary National Bestseller That Changed The Way We Do Business" by Clayton M. Christensen at http://www.amazon.com/exec/obidos/ASIN/0066620694

IA32-64 is an inferior solution for the workstation, server, and mini-mainframe market, but because its incorporated into an existing competitive product, it can deliver this inferior solution for less, and AMD and the channels players that will sell will settle for a significantly lower margin.

This merely replicates what the minis did to mainframes, and what PCs did to minis.

IA64, on the other hand, is just a workstation/mini/mainframe chip entering a mature market with the hopes of penetrating the PC market.

The last attempt with that strategy was Alpha. Alpha promised to deliver workstation/mini/mainframe super iority at the cost of a high end PC. Intel and channel partners simply upgraded the PC chips and systems to maintain both price superiority and attain functional parity with the Alpha "PC" solution.

The ironic thing about Alpha's history is that as its competitors fell away, eg., MIPS and Precision, Alpha's share of the mini/mainframe market fell, with SPARC and Power being the primary benefactors.

Christensen covers the case of a competitor entering an existing market, and history would suggest that Alpha would have a real struggle. And history did in fact, repeat.

Other than Intel having potentially deeper pockets, its not clear how Intel can hope to deliver a new chip into an existing mature market.

If its targeted at the PC market, it's inferior in terms of cost (higher) and performance (lower for existing apps).

If it's targeted at the workstation/mini/mainframe market, it's inferior in terms of performance (PC solution running in IA32 Windows mode) for a substantial portion of the total market and inferior in terms of pricing to support the channel player's need for profit margin to push a new product into a mature market.

Intel is paying Compaq big bucks to port its software (eg., VMS) to IA64, but Compaq can't justify manufacturing IA64 systems - instead the porting work is being done using HP IA64 systems, bought BEFORE the merger was approved by the respective BODs.

HP _has_ to deliver IA64 systems to seed the next step in what is already a decade long strategy to get HP out of the microprocessor chip business. But given HP's losses in its RISC hardware business, HP doesn't have the resources to drive IA64 to success in this market, and that's while HP has a larger market share than Compaq had with Alpha.

(Curiously, Compaq's Alpha business was generally profitable; the reason for exiting the Alpha business was the projected future costs of delivering higher performance compared to the projected higher performance of IA64. I forecast that Compaq's Alpha business will begin losing money as the cost of seed units and payments to partners for porting from Alpha to IA64 start to kick in.)

Bottom line is that IA32-64 is continues to keep the PC chip as a disruptive influence while IA64 isn't disruptive, but a new entrant into a mature, and consolidating, market.

SPENCER T. KITTELSON writes:

Excellent speculative analysis on the Hammer series. AMD does indeed need a big player to support their chips properly.

I wonder if Broadcom is for sale or if their IP can be had at a reasonable fee? That would really put INTC in a world of hurt.

AND BRIAN MACK says:

Great article. There wasn't a lot of new information, but you did an excellent job of pulling it together.

Let me say upfront that I am an AMD fan. Every one of my desktops since my first K6-233 to my current Athlon 1GHz has been an AMD machine.

Every machine I build for friends or family has been an AMD system. They make chips that have an almost unbeatable price/performance ratio.

They are not without problems, though.

AMD has 2 things going against it in the marketplace.

1) They have the perception of being "an imitator" and "unreliable". Anyone that has used computers for any length of time knows that the unreliable label can be just as easily be placed on Intel. As for the imitator part, techies and pundits have longer memories about these things while the general public could really care less. That's why the AMD PR ratings have come back (for better or worse).

2) Do they have enough production capacity? Intel was roasted a few years ago when their volumes of processors dropped while switching from one process to another. That happens and is to be expected from time to time, but Intel has many fabs. AMD has one. Can AMD reach the market share goals that they set using the Dresden fab? If the answer is yes, that's great.

Now extend that question out 2-3 years. By that time Hammers will be in full production and the venerable Athlon may well have gone the route of the K6. Let's introduce a die shrink into the mix and the outlook is not so cheery. I would like to see them add another fab even if it is a joint venture such as the rumored deal with Fujitsu.

I hope AMD can overcome the obstacles.

SAYS RANDY HUBBARD:

I've been promoting the same theory about AMD changing the CPU landscape forever, since they intro'ed the Athlon. Any way you slice the bread, AMD is the CPU technological leader at the moment. They can respond with a cheaper, superior product to anything Intel releases in the P4 or Titanic line. Intel is betting the farm on as yet undeveloped technology in an effort to leapfrog ahead of AMD but that may fail miserably. Even if Intel does get lucky enough to hit a home run with some new technology, it would appear from all sources that AMD has a few new technologies still left up it's sleeve to counter Intel's efforts and AMD's strategic allies in the PC industry keep growing by leaps and bounds.

The times, they are a changin' for the better now that Intel has some serious competition and in fact is playing second fiddle in many market segments. Consumers will be the winners. Now if we could only purchase a quality O/S to run on the X86 platform and watch Microsucks fade into the sunset, life would be really, really good !

AND, LASTLY, "Dave" writes:

Hi Aaron,

Great article.

Did you read ATI, Nvidia porting drivers for AMD-OSX hybrid? Apple could be a great addition to AMD's list of customers. It could lead to a market for AMD's embedded products among other interesting opportunities. Apple could easily sell a dual/quad hammer box at the high end, workstation or server.

Maybe there's a deal in which AMD also FABs PPCs for Apple; that would be an interesting twist... Apple has been trying to buy PPC IP from Motorola ever since they had that falling out over the clone market (brilliant and bold move on Jobs part -- mac clones were a bad idea for Apple IMO).

In Giant Killing, notes the word 'proprietary':

"It [Intel] was going to produce a cheaper chip than their competitors using economies of scale by persuading seve ral companies to use their chip instead of everyone using proprietary architectures."

Note: the Itanium architecture is proprietary. maybe just 'other architectures'.

"With 64bit processors heading for the consumer market, it is obvious that proprietary architectures like the UltraSPARC will be pushed further upmarket."

Note: The SPARC architecture is open. UltraSPARC is indeed a proprietary implementation, as is almost every CPU on the market. The SPARC architecture is found almost exclusively in expensive implementations, but 'expensive' is not necessarily synonymous with 'proprietary'.