PC Gaming Going the way of the Dodo?

[W1nTry]

No i can't and never can consider the limitation to the same hardware a benefit.

On the contrary, limitation is a necessary evil in MANY cases, but on the topic to which we refer, if PC hardware was not advancing at the rate it is we'd actually get a MS OS that could run better on existing hardware with  improvements to its functionality as well. Somewhat like the last revision of Mac OSX for the POWER tech which actually ran faster, better and had LOWER hardware requirements than the previous version. I am figuring apple foresaw their limitations and worked to suite, then of course there is the intel thing but anyways.

The xbox had 2 slow shaders. We're up to 96 fast shader units in the top pcs. 733MHz pIII vs dual core 2.6GHz fx60 with a huge list of improvements; 64MB DDR200 vs GB's of DDR400->DDR2667 PLUS the video memory. Then there's sound, and drive improvements etc. Between these 2 levels isn't a bunch of empty space either. It's years of advances each year farther beyond what the xbox was capable of.

Actually there isn't that great an improvement in sound. And DDR2 is not such a great leap save in power (which is also reaching ridiculous heights with dual core and SLI offerings, which is prolly the biggest reason to jump, have you seen a LARGE leap in performance from DDR2 so far?) but there are LARGE leaps in power consumption in particular in the intel arena, I wait for XDram. Compare an Audigy to a Xifi, we don't see anywhere near the improvement in sound that we do in most other aspects of PC hardware, sure we have 6.1, 7.1, optical i/o EAX2, etc, but you need decent speakers to truly hear the difference and thats more money. And imho halo still looks better on a Xbox than on a PC (but then I believe thats bungie's fault).

As far as claiming that pc code gets unoptimized (at least compared to consoles...), drivers are optimized on a per card basis, and engines such as Doom3, Source, crytek all boast many revisions that tweak them for the cpu. The xbox versions of unreal2003, doom3, and morrowind are the same engines, very cut down. It is the power instruction set, not exactly the power architecture. The power architecture for example has a traditional pipeline + cache setup, plus strong core that does the "thinking" (shhhh!! it's thinking!!!). The Cell based chip has working spu's that can do a lot of simplistic operations quick - but they dont have their own cache, and have to be directed to a small shared cache. It's bad for latency. period. Still - as far as the engine and cpu performance, remove it. At the paltry resolution it supports, it can make a difference, but pc gamers wont play at that res. I barely play at 1024x768 anymore

As far as playability on older hardware, a Geforce 4 can still handle a lot of games - it cannot play Fear well of course, but at least it can run fear quake 4, or half life.
One of the longest lasting video cards was the 9500. The early 9500 could have been bios flashed to a 9700, the equivalent of 6600's in today's gaming. you dont HAVE to buy a card each generation. I went from GF2, to ti4200 to 6800GT.

I both agree and disagree here. Quake 4 imho runs better than Doom3, but still doesn't run as well as I would like. This is a subjective argument in particular though. On my 9800Pro with 2600+ droom3 ran 'decently'. But imho, the thing that made Doom3 an enjoyable game was atmostphere. The enemies are not that smart, the levels aren't that different but the atmosphere of the game was what made it one I was willing to buy. And that was running @1024x768 with medium settings. And I can assure you I dropped below 30fps every once in a while which is QUITE the annoyance. If you had to tone down the settings to play, the game loses the effect and isn't worth playing. Whats the sense in playing farcry without the water effects? I don't dispute you can play games on your 9500 with was a GOSU video card, but gaming is alot more than just staring at a screen with things moving on it. Do you play COD2 on ur 9500 higher than 1024x768 with the bells and whistles?

Doom 3 was very playable on a 9600 - which was one of the common cards.
The FX series CAN be incredibly powerful if coded for. It featured a core that technically WAS more powerful than the 9x series of similar spec, but should really have been paired with an itanic because it featured a VLIW core. Unless instructions were ordered in a particular way, performance was horrible. Carmack did a great job getting it to run doom3 well on it - making full use of the 2 texture per pipeline functionality it offered, which helped attribute to it's speed. Now this isn't a good argument to put to consoles because of the nature of very long instruction word processors. Actually nvidia made several mistakes, including not following dx9 spec and forcing either 16 or 32bit precision, when dx9 called for 24 - so obviously on quality modes, it was slower, and on faster modes, well, 16 bit is noticably lower quality.

I HAVE a vanilla 9600 on another rig home with another 2600+ and to play Doom3 on it is near torture. Even my younger brother doesn't bother to play it cause its just not up to mark.

Every card before it, and every card since has been standard. With the VLIW mistake - even transmeta went belly up. Another factor was ATi's lack of good OpenGL drivers. If those software companies dont want to use DirectX in favour of open GL, do you think they'd switch to a platform with less of a choice? ATi has improved in OpenGL to the point where ati's solutions compete VERY well to nvidia's ($:$) in doom3 and quake4 - when was the last time you saw THAT in a console? Sofware does make either system - but the hardware limitations hit faster than you think, even on the pc where the abstraction makes it easy, yet driver tweaks make it fast.

ATI's solution has improved in OpenGL courtesy of its 512 ring-bus memory controller. As I recall when the X18xx first came out they were WAY behind Nvidia and still are to a lesser extent. Thats all to do with what each manu chose to develop for. Its still fact that Nvidia's implementation of OpenGL is better without the advance in memory architecture. I don't quite follow ur comment about THAT on a console? Consoles are static, hardware upgrade and advances aren't something you buy a console for, you do get improvements, but thats strictly optimization.

I would say safely I get from 15-80% increase in performance and they're both rigs that are mid level. So you mean to tell me that in 1 yr with not 100% increase in performance I am paying more than 1x the former? In the least it should be bout the same wouldn't u say?

I'd say
1. you dont shop like i do
2. you didn't shop for gaming performance, you shopped for overall performance - as evident by the expensive memory, and a full new system, even though all you needed there was the videocard

MOST ppl don't shop like you. I used prices in accordance with a standard ONLINE etailer and one of the LARGEST ala newegg.

Dont bother trying to compare a top end rig with any console. It's a different league altogether.

Agreed

Your illustration is very flawed.
Are you joking by looking at price increases in top end rigs? $750 for a top rig back then???!? First, you're comparing the a $200 intell 2.8 chip, which by the time it was $200, was far from top end mind you to an fx55. By the time intel had the c series, they had the 3.06 with HT, and a full line up to the 3.2GHz - and the top wasn't $200. Then you specify overclockers ram for the fx55, but not for the 2.8C? the 2.8C can overclock a much higher percentage regularly compared to the fx55, which because of it's unlocked multiplier does not even need much higher than 433Mhz, if you rely on multiplier first, then bus to bring to a maximum. When the Ti series was in it's prime, the Ti4200 was $250 - i know, i bought the cheapest one (an MSI) for $240. The Ti 4600 was $400+, the 5900 also debut over $400 (at least, the highly oc'd versions edged past $500). Finally, now that there's competition, you can increase the chip to FX57 in your illustration as it's $810 on pricewatch. change the mobo to a crossfire mobo and drop in 2 x1900xt's for as fast as 7800gtx512 speeds, but $500 less.
The figures are just off on the older rig, and could use some tweaking on the upper one. The times are off. Back in jan 2002 I purchased the 1700+ for $121 - it's all on my newegg invoices. this was the time of the geforce 4 - and they were pricey. I waited several moneths for inflated priced to drop before buying my ti4200.

Well I was trying to rem what was the top Intel cpu then so 3.06 not 2.8 my bad but what was the diff in price? 50-80USD? nothing too substantial I am sure. We're NOT talking overclocking, if we did, then you could take a 2.66 and OC with the proper cooling to over 3.06 and get the performance for a decent price difference. We're going strictly on top to top, not OCed to top to Top. You're comparison there would be flawed. I agree with the X1900XTX instead of 2x7800GTX, but fact is a top of the line rig is SLI or crossfire, so what about X1900XT master and slave? if there was a Ti4600x2 back then I would have factored that into the equation, I just used what was available.

Back in 2000 - Max pc made their Rig of the year - it cost them $12,000. With the pc you can spend as much as you want. That does not make it optimized for gaming (if that's the main concern)

Agreed, but you can build rigs now at ridulous prices much higher than 12G it b/c we have more options now.

let's see - 1 year to see 100% increase in performance from my old rig

I have a 6800GT/3400+ i got a year ago(Jan) - because the bang for buck ratio of the 3400+ was a lot better than the 939 3500+. My rig cost me $1k

for $1000, purely gaming wise i'd
1. get an epox sli and reuse my chip.
2. 2 above stock 7800GTXs
-------------
$998

When I stated 15-80% it was the overall figure for the system, not strictly graphics. Keep that in mind. Remember I went from 2600+ to 3500+ and the platform changes.

and probably a lot more than doubles my graphics power. Yes I'd use the same original base, because it's easily enough to handle today's and much of tomorrow's games. In fact, i can probably get near double performance with a single x1900xt - and that would make my motherboard choice cheaper as well. for a much lower price - say $600 total. Look at Baego's upgrade. His money went to get gaming performance and he got it. finally, reread the priing post on the last page. Your post leads one to believe pricing is still high if you want to play games today, it isn't. you dont need SLI, you dont need the most expensive card, you dont need a very expensive CPU.

See point above about overall system performance increase.

Finally, I'm never going to say a console is a bad thing - quite the contrary. I just dont agree with a lot of the arguments against the pc. The main arguments for the console is it's ease of use AND price - but with pc prices getting lower and lower - just think! $600 today for a 7800GT wielding pc! - and console prices getting higher, the pc is looking better with time!

Agreed and I hope it gets better (though I honestly want a Nintendo Rev, which will be still affordable and have what games really lack lately, INNOVATION).

my last few posts were agains these arguments
1. price
imo, price is not a card to play. Anyone here with a decent enough pc - and that's most of you - can run a cheap $100US x800 video card to play any game you desire today. If you dont have a pc, then a decent enough one that would handle any game need only cost $500 or so. A console IS cheaper, but not nearly the order of magnitude most people make it out to be. It is NOT cheaper when it is a video card generation or two into it's life, and the consumer already has a pc with enough cpu power and memory.
2. power
the power of graphics is in the graphics card. CPU&Software to a point, but the increased resolution a pc can display, the rapid development of graphics etc, leaves behind the console shortly after it debuts, even when they tweak the bejeasus out of it. Tweaks on the personal pc can make a seemingly weak system into a much more powerful contender. 9500 to 9700, 8/4 6800le to 16/6 6800 (pipe/vertex unlocking), 2800+ from 1.6 to 2.4Ghz, flashing the x800gto˛ to an x850xt pe etc etc.

Here's my responses:
1. You could buy a 6200TC card and play any game these days, but why would you WANT TO? gaming is again not just about staring at a screen with moving objects. Gaming is an experience which comes to vary GREATLY with the quality of audio, video and environment. Why don't ppl play in arcades as much as they used to back say 10yrs ago? atmosphere and price. I stopped going to arcades and in fact had a disccusion yesterday with some colleagues about how arcades got too many sprangers and it just wasnt' worth it to go anymore. You wouldn't go to a place where you could get robbed, why play something like doom say without the lighting effects? its a strange and extreme comparison but its just to stress overall experience has alot to do with gaming.

2. I agree that the PC will ALWAYS without FAIL reduce a console to a smoldering pile of ash in terms of performance, but as time progresses, to play new games u NEED new hardware to keep up, or suffer the loss in details, performance, etc. On a console you KNOW the new games are gonna play on ur box. How many PCs can say that? And PCs also suffer thigns like OS upgrades which usually need better hardware, driver issues, hardware compatibility issues. Even if you build a PC JUSt to game, as you keep updating software you are BOUND to run into problems, its a GIVEN. Rarely is there a PC that NEVER suffers from incompatibilites if its constantly upgraded.


On the actual thread topic, PC gaming is doing great with Online games like WOW and other MMORPGs, even online play is booming. I would say PC gaming is undergoing a transformation, but consoles are doing DAMN good too and will prolly do better as time goes by.

[Beomagi]

Keeping your hardware up is like purchasing a second console. You're right, upgrades would occur every few years, but during then, you're talking about mind numbing graphics compared to a console. How many games that actually NEED 6800 power can you even imagine running on an xbox? With the xbox, you upgrade to the xbox 360 - I upgrade my videocard from the 4200 to the 6800 and i'll wait for another year at least before i have to purchase my next graphics card.

If you're bringing up top end rigs or parts, you're not comparing the pc to the console. That's a perspective people already into pc's would look at - a console person just wants to play games as he doesn't get the usefulness of having a pc with all the bells and whistles. That's why i still say $800 can net you an xbox killer, with the sole purpose of playing games. If you say you need a full upgrade for better system performance, you're making pcgaming (strict pupose of this thread) out to be a lot more expensive. As enthusiasts, you'll have a lot better overall performance, but people who dont think of their pc as their game machine, and barely use it wont. You said you spent $1k and your rig performed 15% to 80% faster - as with the nature of this thread,
1. if it were sole gaming performance, there were better ways to spend money to get it
2. if it were a full system performance upgrade, it's not accurate to use it when looking at cost for gaming performance.

http://gatt.carigamers.com/index.php?topic=6027.msg120816#msg120816
all of my prices and link ARE to newegg...

Back then, intel's top chips would cost you $500 to $700. this is WHY AMD had such a fanbase - cheap compeditive chips. And i would bring up overclocking because of the ram you chose Low latency ddr4200?? unless you're talking overclocking, anything above DDR3200 would be extra money not used.

[W1nTry]

My question is, would 800USD have gotten you an Xbox killer when the Xbox debuted? will a current 800USD gaming rig best an Xbox360? you see the time cycles of a console are 5-7 years generally and you mentioned yearly upgrades on a PC gaming rig. You see its FAR more feasible for the console. One other thing i'd like to point out is that with the level of graphics we're reaching and the rate at which ATI and Nvidia are revamping their product lines, how long will it be before you hit the visual ceiling on all powerful consoles and PCs? then what? From perspective of console to gaming pc, if you've hit the ceiling, then consoles become the obvious choice if its strictly gaming. And don't tell me we have decades to go before we hit the ceiling, we're rapidly approaching photorealistic real time games, just look at project gotham racing. Rem when CPU manus said that a dedicated video processor wouldn't do what they would be capable of? look at how more complex the GPU is than a CPU, look at how not too long ago, render farms were the only way to render 3D video and now we're doing Shrek level quality vids on desktops using your video cards. We'll reach sooner than you think.

[Beomagi]

$800 for a current rig - yes, i posted how earlier, using the sempron64 and x1900xt, AND i went over technical aspects of how the video solution is more powerful - double memory bandwidth, dedicated memory, faster core, etc.

[W1nTry]

You're saying THAT will beat an Xbox 360?.... hmm aright... i'll take your word for it.

[Beomagi]

It will - look at the core of an xbox and the core of an x1900xt. The core of an xbox is slower - AIW specs, and the memory bandwidth is HALVED. The Memory is shared for textures, system, AND game itself, and if you think the sempron gives weak single core performance, I've posted several benchmarks and overclocking results earlier in this very thread, on the last page.

[W1nTry]

Isn't the Xbox 360 core video solution using high speed local memory (as opposed to GDDR? or at least in addition to/OR XDram?) and isn't it also running on Rambus technology (I KNOW the PS3 will, I can't rem for the 360 however). And single core performance of a sempron higher than that of a 3.2GHz Power CPU??? not taking into account its a triple core for later multithreaded games? Beo you've just lost some plausibility on that one, in the LEAST the POWER arch is comparable (stress least AND i'd prolly compare to the Northwood as opposed to Preschott) to the P4, you're telling me that a sempron would best a 3.2GHz P4 in computational power? I will also remind you that POWER tech though using the antiquated FSB runs generally @ 1/2 the CPU Freq? aka 3.2GHz==1.6GHz FSB. Look at the G5s, the multiplier for all intents and purposes is 2. 1.6GHz==800MHz, 2GHz==1GHz FSB, etc. I still say its not an apples to apples comparison but I find your argument somewhat difficult to substantiate.

Update:
I've done a lil googling and found some specs, so a 800USD gaming rig gonna beat:
Xbox 360 technical specifications
Look at the Memory bandwidth... yeah... i'm sure the sempron will beat it.

[Beomagi]

It is NOT a Power architecture - it's a CELL architecture, using a CORE unit to divy workload to two SPUs. And i've Gone over the cpu differences Before HERE
http://gatt.carigamers.com/index.php?topic=6027.msg120828#msg120828
More info - when an spu requests memory, it checks a tiny local store - if there's a miss, it checks l2 cache - a puny 512k that's shared among both spus - so the spu accessing memory more rapidly has the capability to starve the other on cache. Then L1 (L2 then L1 sound backwards eh?) is checked. THEN it gets to the PPC core, which for cell cpus, dont process the into, it's more of an instruction router, THEN it hits memory. Carmack commented on the cell, himself and said he'd rather have used standard x86 processord because they were up to twice as powerful for common engines. Even so, i STILL showed how and why the cpu performance removes itself.

Take a look at performance benches in DoomIII I previoulsly posted HERE
http://gatt.carigamers.com/index.php?topic=6027.msg120819#msg120819
I went over the resolution/graphic tweaks making the cpu less of an issue in my previous posts, using actual benmarks of real world games.

No matter what kid of Ram it's using It's STILL rated at having only 22.4GBps with is only equal to 700MHz memory at 256bit width interface in the original x800 or 6800 - And BECAUSE of the architecture - the northbridge is the video chip - it's SHARING that link. Compare to almost 50GBps on the x1900xt.

IF you want to continue calling it a power architecture (which it isnt), then I'll do you a favour and repost the same links from the last page
http://www23.tomshardware.com/cpu.html?modelx=33&model1=238&model2=212&chart=70
http://www.xbitlabs.com/articles/cpu/display/cpu-games2_3.html

do you think the p4 840EE or the 955XE really has a huge advantage in gaming? It's immensly capable at parallel threads able to handle 4 discrete threads easily. You yourself already said the FX57 is still the gaming leader - a single core processor, and the sempron can overclock to give you sufficient single core power for any game, compared to other single core processors. The Sempron does not have to best the higher end p4 3.2 GHz cpu because the GPU would be the limiting factor in games, once again, check the links i've double posted here

I already said in what way the Cell cpus can be used HERE
http://gatt.carigamers.com/index.php?topic=6027.msg120834#msg120834
To truly push itself ahead of single core procs making use of it's multimedia abilities. Thats the only place where such memory bandwidth can be used via the cpu - we saw that back when the P4 increased it's bus from 533 to 800MHz. Benchmarks showing the performance of faster and faster memory, show performance leveling off FAST because the cpu doesn't need it. How much faster did doubling the bandwidth make your games? Do you think 939's dual channel interface was THAT much faster than 754's single channel?

[W1nTry]

Dude, the 360 is built on POWER tech NOT CELL tech, did you even look at the link I put?. And why did you post a bunch of links to Benchmarks of AMD and Intel based processors? we're making a comparison of the Xbox360 processor and the consoles gaming capabilites vs the 800USD rig you grafted??? I don't follow your logic. Think about this carefully before you reply, MS is using SONY/IBM/Toshiba's CELL tech? note the key partner there being SONY??? The Xbox 360 CPU is built around the PPC core:
Xbox 360 CPU

[Beomagi]

I keep posting links to show that the CPU is not going to make much if any playable difference in gaming. And it doesnt matter how massivly parallel the core get, which is WHY i'm posting links comparing the sempron to the fx57 to the 840EE and 955Xe.

It does look like i'm wrong about the cell architecture in the xbox. I apologize. I do remember reading about the design capable of handling 6 threads. Looking for the article. Even so, I'll still say, the CPU would not be the end all for gaming performance, At the mid range resolution of 1280x1024 in the links posted, the ability of multi thread capable processors or super fast single core processors is almost indistinguisable.

[Beomagi]

reports such as this were abundant before - harder to find now.

Each core will also be able to act on two threads at once. Think of threads as customers at a grocery store waiting in a checkout line. Each core would be like a checkout clerk who can work with two customers are once, thus shortening the wait time. Each core's ability to handle two jobs simultaneously means the chip can act like it is in fact six chips. Each core will operate at 3.2 gigahertz, which is comparable to the processing speed of Intel's fastest Pentium processor.

Analyst Kevin Krewell of Instat/MDR, which is hosting the Fall Processor Forum, says the new IBM chip shares much of its lineage with the Cell Processor and other chips in the PowerPC family that have come before. "The basic core at the heart of this chip is very similar to that in the Cell Processor," Krewell says. "There's definitely some design re-use going on here."

http://yahoo.businessweek.com/technology/content/oct2005/tc20051025_898864.htm
another cell/xbox tidbit
http://everythingapple.blogspot.com/2005/03/xbox-2-xenons-cpu-core-cells-ppe-core.html

[W1nTry]

The way I look at it, is that multiple cores in gaming comes into play with the amt of simultaneous work being done (no duh I know) but to be a little more specific, these consoles (save the rev) does not have a physics engine, thus the CPU would be handling that. A multi-core processor can then feed the gpu as per norm, perform additional calculations for the environmental intereaction and some else say, more intelligent in game AI. Which would create an overall better gaming experience without a performance HIT. I hope this sheds some light to how gaming IS affected by the CPU in consoles. Without an Ageia physics engine like the PC, teh consoles only hope is to have multiprocessing capabilities. At the same time a more powerful single processor that is CISC (as POWER) will result in a difference in the video performance than a single RISC (intel/amd) based processor. True the increase with CPU power becomes somewhat linear at say FX-57 and up stage (with the exception of at res over 2000+x1200+ where it still benefits to a measurable degree) we don't have a true indication of how the video abilities of a system will scale with POWER tech. Its been my experience that games on an apple were always smoother, ALWAYS and this is from personal experience.

I am unsure where those Cell/xbox articles get their data but I would say the PPC lends much more to Cell than in the reverse. Take into consideration that the Cell is being jointly developed by IBM(makers of PPC for over a decade now), SONY and Toshiba?? IBM stands to gain even more from this processor see IBM to use CELL in blade servers and is the more dominant partner in the technological department (they ain't called BIG BLUE for NOTHING) also see IBM intros POWER 6. You seem to forget IBM gave AMD the ability to transition to SOI and 90nm. Heck the PS3 will have a PPC based processor inside it. The memory tech inside cell is RAMBUS see: RAMBUS inside cell. I don't see CELL being the child that spawned the parent (POWER)

[Beomagi]

Indeed you are right, we need better people to code multithreaded game ai/physics and such so that multi core chips get used better, and physics MAY be a big hit if they can do it properly - so far, i've not seen any demos or evidence to support it - yet. Multicore can handle having threads that do a lot of extra background stuff without any one thread becoming a bottleneck, but nogame today is coded with that in mind, so of course, in the future parallelism can render the xbox's cpu fast enough to a point where single cored cpus are bottlenecks rather than the graphics card, but it's not so yet.

As for the power architecture - you can skip it - look at the more basic design. it's a standard cpu to northbridge to everything else connection. At 1280x1024 (not a high res by today's PC standards) when the links show cpu's levelling off, a change in architecture would NOT help. Levelling off means the cpu is waiting for the videocard.

It is because of that, and because we dont do as much as can be done with multicore CPU's that the sempron WILL suffice.

Wikipedia says DDR, It doesn't matter. What does matter the the throughput and the 360 is less than half that of the x1900xt.

[W1nTry]

Well I think u meant the video card is waiting on the CPU but the bottom line is a bottleneck between the two. The interesting thing is OS X moved the desktop onto the video cards. Thats why aqua is so gorgeous and windows looks like 0.5MP camera shot @ 1600x1200 res. If I find the link showing how they did it i'll put it here, but in essence they basically feed the screen to a OpenGL api and the Video card renders the desktop in real time (I could be wrong on the actual phrasing of this wrt api). Bottom line is they are utilizing the tremendous bandwidth between the GPU and Video ram instead of CPU to Video which is much reduced compared. Here is where speculation comes in. With the dawn of HT3 (that's HyperTransport 3 to n00bs like crixx who think its Hyper Threading), we will be seeing much greater bandwidth between CPUs. There is even speculation that for instance on an Opteron which sports 3 HT lanes, in a multi-cpu setup, 1HT is used to system memory, 1HT link to other Opterons for coherent communication and 1 left for whatever they want really (possible co-processor). Nvidia uses HT to connect the nforce 4 intel northbridge to the AMD nforce 4 north/south bridge to get SLI. With things like HT3 and coherent HT we could get the bandwidth we need to not bottleneck CPU-GPU links....however that will NOT solve the CPU processing being the deprecated factor. I hear what you're saying about the memory bandwidth of the 360 being less than that of a X1900XTX (however this is GPU to Vram), I do however wonder how the increased throughput between CPU-GPU will affect its overall performance anyways. Look at this, the bandwidth that the 7800GTX512 has is greater than that of the X1900XTX, but the ATI solution is better by virtue of implementation. Same way here, the Xbox360 overall architectual implementation with superior system bandwidth might win out over sheer bandwidth within a single component. The following is taken from a link i posted previous:

Memory Bandwidth • 22.4 GB/s memory interface bus bandwidth
• 256 GB/s memory bandwidth to EDRAM
• 21.6 GB/s front-side bus

Currently the A64 platform and Intel (save maybe opterons in 4U and up) don't exibit that kinda bandwidth in the FSB or HT bus and 256GB/s memory bandwidth to EDRAM if I remember correctly this reffers to local memory available to the videocard so I am unsure whether ur bandwidth superiority of the X1900XTX is actually true for the case of the 360. I am leaning towards the 1900xtx not being faster as this same local memory is being used for the PS3 (RAMBUS tech) which is currently superior to GDDR3. its associated with XDram.

[Beomagi]

ATI's x1900XT beats the 7800gtx512 because of the increased shader power and fillrate which they deemed to limit the card.

That goes under another thread - where ndivia claims texture to shader ratio is 1:1 or 2:3, and ati is claiming it's 1:3, and so far, their solution looks like they're right(er). (so far)

I hear what you're saying but so far every game that i've seen is not affected much by
1. lowering my hyper transport multiplier, decreasing the system bandwidth (this does nothing of course to gpu to vmem throughput.
2. single channel/dual channel - decreasing the cpu to ram bandwidth (hits is the biggest impact of these 3, and it's still not great)
3. agp 4x/8x, pcie 8x/16x - I get No improvements in speed when i go from 4x to 8x - except in one circumstance. higher bandwidth to the gpu really helps with one main thing Textures - and at this point, i'm at a crawl anyway, no matter the card.

Most of the Gpu memory holds textures. The fatter the AGP or PCI-e pipes, the faster you can load them. Thing is only when you're running a game with textures much much higher than your card can hold would you see this kick in. I noticed this when playing battle for middle earth. On my 128 MB card - i'll play to a point, then hit an amazingly slow area. on my 256 MB card (both at 6800 class, on is unlocked, the other is a GT, so other than memory speed which is close, memory size on the vanilla is half - 128MB) I encounter no problem - cutting textures to the second highest lets me pass that point keeping my fps up - the order of magnitude from the drop is astonishing. This occurred in several games and i was able to test em because of the 2 setups. Textures are loaded at the start of the level usually. With low texture memory, those that arn't are loaded when needed so running into a part of a level can cause slowdown if there's a lot of refreshing to do - or it can jsut cause the game to slow to a crawl when the sum total of all textures visible on the screen exceed the alloted memory on the card. At that point every read would miss the vmem, and check main memory.

That would be a strength for the xbox360 because it's memory is a LOT faster than DDR 3200, But it would NEED that advantage because the memory is shared, thus giving less room for textures than a true 512MB card. In many games, today, the game itself takes up half a GB easily (bf2 runs best with over 1GB of memory).

Vertex instructions and shader instructions take up soooo little bandwidth it's surprising. For that i dont see a bottleneck between the cpu and gpu. The bottleneck would mean that going from 8x to 4x (or 16x to 8x in pci-e) would severly hinder graphics performance, whicch it doesn't. The bottleneck lies in the card itself - ati's x1900xt showed tha shaders were indeed slowing down the x1800xt - nothing else changed, cept the number of shaders per pipe. Video cards have several "bottleneck" factors - The memory size only hinders performance if textures and lightmap data are too great. If it isn't then it's nothing. Memory speed AND bit width (multiply for throughput) is a big one - since the frame buffer resides in gpu memory. processing texture, processing shader instructions, processing verticies are 3 possible bottlenecks on one unit - the core.

[Redlum08]

also, remember, Microsoft is losing $300 on every Xbox 360 they sell. Cost of Making an Xbox 360 = $700

[shivadee]

all first party companies MUST loose money when selling a console. Always happens..same with SONY, Nintendo and MS. Over time however the same hardware that was expensive to manufacture at the time is much cheaper to make OVER time. THey break even after about a year or so....sometimes more like in the regular XBOX case

when XBOX launched LIVE in November i believe....the next year by the 3rd quarter they broke even and have been making profits.

It will be the same with the PS3 and rev......Simple profit and loss marketing plans....you must expect a loss at first....hence each company when launching a system must have huge capital.....

by the way...whatever happened to Infinium Labs and that POS they were supposed to bring out?

[TrinireturnofGamez]

 infinium labs  was a cover up to steal money from investors ...  Hardocp opened up one of their ''samples''  and found PC parts.. athlon XP ... run of the mill motherboard , video card etc.   opening up another one   found a pentium 4 ...  pretty obvious that they were just slapped together to impress stockholders and investors every few months ..   
   Even if the Phantom ( aptly named) did release it would have a few problems : Bieng a PC  and running PC games it would be quickly outdated , plus the hard drive was too small for a hardcore pc gamer ... only 80 gigs i believe..  not nearly enough , The system was supposed to use a Geforce fx 5600 , trying to play any new game on that card right now will result in a nightmare of lag even on medium settings ...