![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
Radeon HD 4800 card pictured
- Thread starter Silus
- Start date
alg7_munif
Supreme [H]ardness
- Joined
- Oct 9, 2006
- Messages
- 5,862
Noo.... I want to see Kyle in a pink tutu
Jakalwarrior
2[H]4U
- Joined
- Aug 6, 2005
- Messages
- 3,189
time to invest in amd stocks? like that one guy posted he bought it for 8 bucks maybe i can may some money on it. but i dont have a clue on how to trade however
With the market going to crap and ATI not being near enough to pull AMD out if debt, I wouldnt
im still pissed I bought some stocks (NOT Amd) last monday when the market was sort of climbing, only to have it take another dive.TheBluePill
2[H]4U
- Joined
- Aug 17, 2005
- Messages
- 3,773
If the 4xxx series is released in early and AMD can pull of 3+ghz off of their next round of Phenom's, they have a chance to get back in the market recoup some of their lost ground.
Missing a product cycle for both their Processor and Graphics divisions were very painful, but not un-recoverable.
Missing a product cycle for both their Processor and Graphics divisions were very painful, but not un-recoverable.
- Joined
- Jun 25, 2007
- Messages
- 7,268
That is true. Believe it or not, both AMD and ATi when they were seperate companies have faced harder times and have survived. Rumors are winding around saying that the 45nm spin of Phenom puts AMD back in the game and even gives Penryn a run for the money. More rumors say that RV770 is up to twice as fast in texture and AA dependant situations. I don't know if there's any truth to that, but I am expecting the worst, but certainly hoping for the best.
undertheradar
[H]ard|Gawd
- Joined
- Oct 26, 2004
- Messages
- 1,792
Oh yeah it is... I was looking at the stats from the most recent Steam survey, and the ATI 3xxx's are in the basement. Of course, considering how much longer the 8xxx nVidias have been out, its not that much of a shock, but still, the 9xxx nVidias already had some percentage points!!!
TheBluePill
2[H]4U
- Joined
- Aug 17, 2005
- Messages
- 3,773
Yuppers, they didnt compete with the 8 series in any real way until recently.
Zorachus
[H]F Junkie
- Joined
- Dec 17, 2006
- Messages
- 12,547
Here is my educated guess as to the speed of this 4870x2, due in June, mark my words it will be no slower than this.
I say for sure the speed = to 2-3870x2's in Crossfire, + add 15% for the clock increase this beast will have at over 1ghz
So check a 3870X2 benchmark, times it by 2 or check a 3870X2 Crossfire bench and add 15%, and I bet that is where it falls approx
I say for sure the speed = to 2-3870x2's in Crossfire, + add 15% for the clock increase this beast will have at over 1ghz
So check a 3870X2 benchmark, times it by 2 or check a 3870X2 Crossfire bench and add 15%, and I bet that is where it falls approx
NKDietrich
Supreme [H]ardness
- Joined
- Jul 25, 2004
- Messages
- 5,442
The most interesting part of this is:
Guess those leaked spec's a while back were fake.
Guess those leaked spec's a while back were fake.
TheBluePill
2[H]4U
- Joined
- Aug 17, 2005
- Messages
- 3,773
If they fix AA and squeeze a boost out of it while nvidia just does a G92 refresh, ATI could very well take a marginal lead...
Care to elaborate on that? I'm not sure what you mean. The 2900XT didn't suck because of crappy drivers, it sucked because it was hot, slow, and used a ton of power - drivers can't work miracles...
The architecture isn't really the problem. The chips can blast through multisampling if they're excessively powerful in contrast to NVIDIA's chips, but, so far, AMD hasn't been hammering the 'excessively powerful' part hard enough.
AMD's still banking on MSAA being an exclusively shader-based process on all hardware, probably, but I doubt it's going to pay off for 'em until 2009.
AMD's still banking on MSAA being an exclusively shader-based process on all hardware, probably, but I doubt it's going to pay off for 'em until 2009.
well the last steam survey was quite a while ago i think they said november? or something.
Yeah, that Steam survey was back in November 07, so it does not include the HD3x series of cards. I'm of course assuming they don't run the survey very long, because if started in mid November, and was 2 weeks long, some 3xxx series cards should have shown up.
you're absolutely right about that, but drivers do fix anomalously unusual performance like the 2900xt in stalker. the asus branded card came with the game packed in, and even had its artwork silk screened on the cooler, but it took a few driver refreshes before a single xt could run the game at 1680x1050. a lone 320 gts at the time was capable of running the game at 1600 rez with max settings, but the xt required a drop to the next widescreen spec at 1440x900 to achieve the same frame rates. stalker was arguable the toughest game to run at the time of the xt's launch, much like crysis is to the x2's launch. the x2 still sucks on crysis, but some of the beta driver leaks show good improvements (or so people say).
Your spelling and grammar is a joke.
Zorachus
[H]F Junkie
- Joined
- Dec 17, 2006
- Messages
- 12,547
lol, but what am I laughing for. I have made some silly posts before as well that didn't make sense.
Its been out for almost 2 years since the X1950XTX, The reason it seems not even off the shelf is that the fact that nvidia ignored it completely, perhaps because GDDR3 quickly hit the same speeds, and likely at tighter latencies too no less. GDDR5 will be what GDDR4 was supposed to be.
undertheradar
[H]ard|Gawd
- Joined
- Oct 26, 2004
- Messages
- 1,792
Actually, that is not correct, since that is the last time a survey was 'completed', but there is still one going on right now:
"This survey began November 13th, 2007. Last updated: 3:53am PST (11:53 GMT), March 03 2008...
#1 NVIDIA GeForce 8800 126,898 15.02 %
#3 NVIDIA GeForce 8600 67,960 8.05 %
#6 NVIDIA GeForce 7900 37,281 4.41 %
#7 NVIDIA GeForce 6800 34,601 4.10 %
#35 ATI Radeon HD 3870 3,530 0.42 %
(the only 3xxx series to even show on the charts, although I take back what I said before, the 9xxx hasnt shown up on it either).
See for yourself:
http://www.steampowered.com/status/survey.html
That mental image disturbs me greatly.
sorry english isn't my first language
seems you aren't that good either
NVIDIA didn't ignore it. It's just that GDDR4 was still too expensive at the time and fitting their graphics cards with it, would mean either more expensive cards or smaller profit margins. GDDR3 was cheap and wasn't too far off GDDR4 in terms of performance. As we saw later, GDDR4 really doesn't help much, except in power consumption.
It's what R700 is supposed to be anyway. R600 enhanced, with AA fixed, an increased number of ROPs and, most likely, an increase in Stream Processors.
Silent.Sin
Gawd
- Joined
- Jun 23, 2003
- Messages
- 969
I wouldn't believe any of the specs floating around right now, but I suppose they are feasible. All of you people wanting AMD to ditch the current R6xx architecture...why? The current 38xx family is pretty f'in sweet, and exactly where AMD needs to be.
I think the reason that NVidia and AMD have been milking these "old" architectures for so long is to try and recoup their R+D costs for them. Whatever kind of nonsense you guys are thinking is happening with the lack of competition is ridiculous. If NVidia had a better architecture ready that could MAKE them money you can bet your ass they would have released it. Kind of the first rule of business, no? You don't sit back sipping margaritas and wait for your competition to catch up and have the chance to overtake you...just doesn't make sense. Some companies (read: AMD) might be forced to release their hardware a tad earlyish (read: 2900XT) just in order to stay in the game, but believing NV DIDN'T release hardware because they pwnz0red the game is pretty laughable.
Problem is that developing and releasing a new architecture is damn expensive. Especially considering the amount of software engineering that has to go on with these things--hell the drivers for 1.5 year old hardware STILL aren't perfect. Fact is that these things are beasts with more logic circuitry than current CPUs have. Not only that, they've actually BECOME what CPUs were generally considered the masters of in the past, general purpose processors.
I think from this point on we're pretty much done seeing entirely new architectures for each generation of GPU. What we'll see is incremental improvements on the same architecture in order to maximize what's possible with it. Only when it becomes too cumbersome to add on new features will we see a brand new core design. Wait that sounds familiar...oh yeah that's right, that's what happens with CPUs.
I firmly believe that the R6xx architecture was created somewhat as an experiment in modularity, a foundation to build on for the forthcoming core-level parallelization of GPUs. Especially the R670 refresh which really only decreased size and power requirements from the R600. Clearly the 3870x2 was on the roadmap from the get-go; it wasn't an answer to the year old 8800Ultra, it was a proof of concept of single-card crossfire. The scaling with those chips is pretty damn good in situations where multi-gpu is taken advantage of. The drivers aren't quite there yet, but they're getting up to speed pretty fast.
The "problem" with the original R600 core was that it was too big and power hungry for such a multi-gpu solution. The technology just wasn't there yet to make a ~700 million transistor core small enough, cool enough, etc to make a 2900x2 type card. AMD solved that problem and now we see the fruits of that labor. Next up is the performance of those cores, which, according to the allmighty [H], has been taken care of at least in the area of AA (which was really the biggest stumbling point of the R6xx series).
I anticipate that this summer is going to be something to look forward to, the battle is going to be very heated and we will see some really nice performance improvements. Now if only EA hadn't pushed Crysis out so early and actually waited until it was done it would have better matched up with this release schedule...fancy that, EA screwing up..but that's for another thread.
I think the reason that NVidia and AMD have been milking these "old" architectures for so long is to try and recoup their R+D costs for them. Whatever kind of nonsense you guys are thinking is happening with the lack of competition is ridiculous. If NVidia had a better architecture ready that could MAKE them money you can bet your ass they would have released it. Kind of the first rule of business, no? You don't sit back sipping margaritas and wait for your competition to catch up and have the chance to overtake you...just doesn't make sense. Some companies (read: AMD) might be forced to release their hardware a tad earlyish (read: 2900XT) just in order to stay in the game, but believing NV DIDN'T release hardware because they pwnz0red the game is pretty laughable.
Problem is that developing and releasing a new architecture is damn expensive. Especially considering the amount of software engineering that has to go on with these things--hell the drivers for 1.5 year old hardware STILL aren't perfect. Fact is that these things are beasts with more logic circuitry than current CPUs have. Not only that, they've actually BECOME what CPUs were generally considered the masters of in the past, general purpose processors.
I think from this point on we're pretty much done seeing entirely new architectures for each generation of GPU. What we'll see is incremental improvements on the same architecture in order to maximize what's possible with it. Only when it becomes too cumbersome to add on new features will we see a brand new core design. Wait that sounds familiar...oh yeah that's right, that's what happens with CPUs.
I firmly believe that the R6xx architecture was created somewhat as an experiment in modularity, a foundation to build on for the forthcoming core-level parallelization of GPUs. Especially the R670 refresh which really only decreased size and power requirements from the R600. Clearly the 3870x2 was on the roadmap from the get-go; it wasn't an answer to the year old 8800Ultra, it was a proof of concept of single-card crossfire. The scaling with those chips is pretty damn good in situations where multi-gpu is taken advantage of. The drivers aren't quite there yet, but they're getting up to speed pretty fast.
The "problem" with the original R600 core was that it was too big and power hungry for such a multi-gpu solution. The technology just wasn't there yet to make a ~700 million transistor core small enough, cool enough, etc to make a 2900x2 type card. AMD solved that problem and now we see the fruits of that labor. Next up is the performance of those cores, which, according to the allmighty [H], has been taken care of at least in the area of AA (which was really the biggest stumbling point of the R6xx series).
I anticipate that this summer is going to be something to look forward to, the battle is going to be very heated and we will see some really nice performance improvements. Now if only EA hadn't pushed Crysis out so early and actually waited until it was done it would have better matched up with this release schedule...fancy that, EA screwing up..but that's for another thread.
Zorachus
[H]F Junkie
- Joined
- Dec 17, 2006
- Messages
- 12,547
Well said and spot on.
I agree makes sense to me
We will see VideoCards go the way of Intel and AMD, a new generation only every few years, but once a year a minor refresh.My bad, I stopped at the first video card chart on that page! Too close to release maybe? Not sure, I'm not a fan of making excuses for manufacturers.
AMD needs to be behind nvidia's "next-gen" mid-range? Yeah, perfect place for them
Falling behind because your hardware sucks as you stagnated and the competition didn't is even more expensive (case in point: 3dfx).
Somewhat yes, somewhat no. Since video cards have gone unified, I don't think we will see sweeping changes. I *do*, however, think there will still be some major architecture revisions, just not necessarily in one GPU cycle (eg, the memory system may stay the same while the individual SPs get a re-work, or vice versa).
Please don't try to compare CPUs to GPUs - it just doesn't work. At all.
Multi-GPU designs are hackish and suck. If they are the future, I want no part of it until they fix scaling (>1.9x in *EVERY* game with no game-specific tweaks needed).
That, and the R600 is slow. A die shrink only made it viable because then it became cheap - its performance certainly didn't improve. Kind of a side-step in my opinion.
Silent.Sin
Gawd
- Joined
- Jun 23, 2003
- Messages
- 969
kllrnohj,
Don't want to quote directly cuz that would be a gigantic post. But anyway...
1) AMD is not in a bad place. It's not the BEST place, but it certainly isn't bad. They compete at nearly the same level price wise. The fact that it's labeled as a "mid-range" vs "high end" product has nothing to do with price/performance ratio. Lowest price of a 9600GT I just did a quick search for is $170, 3870 is $190. Not too far off.
Also I meant most of my comments to be directed towards their development process and roadmap, not necessarily as they stand against the competition. They are in a good place for keeping inline with where they want to go in the future.
2) If AMD's hardware has stagnated then so has NVidia's. The only major change they've had to the "high end" is the G92 core, which in essence was the exact same thing that happened with the R670 core, smaller and less power hungry. The 9800 series is supposed to simply be a refresh of the G92 core...so what is the argument again?
I don't really get your more expensive comment, they aren't anywhere close to being in a 3dfx type situation. If you want to look at sales figures check out OEM numbers, these halo products represent a very very small fraction of the company's overall market strength. 3dfx was caught in the wrong place at the wrong time where the OEM market was absolutely and completely owned by Intel for the most part with the scraps going to some other vendors, including ATi. If 3dfx didn't have the products to compete with NVidia in the high end performance market (which they didn't), they had nothing else to fall back on.
3) I was comparing the development cycles of the two coming very close to each other now, not directly comparing their architecture per se. Obviously the GPU is still a specialized processor that is very different from x86, but the use of them has become something with a much broader range. Something that software outside of games can make use of. You can't tell me you don't see similarities between this and FPU development in the past. It's exactly the same direction GPUs are heading. Which leads me to...
4) If you don't want any part of multi-gpu solutions I suggest you start or fund a company to do otherwise and do it quick. They might seem "hackish" but there is no possible way that the industry can sustain the previous method of making bigger and better by brute force alone. You'd be looking at gigantic development times to create the next gen 1bil+ transistor monolithic core. To make a simple and elegant design that works and be able to connect them together is what AMD wants, and what it's been trying to form with the R6xx. Just what exactly did you expect to see in the Fusion processor? Or bulldozer platform if that ever becomes a reality.
Yes, there are diminishing returns on simply adding more GPUs, but the benefits are still there and the only other option is hugely expensive and time consuming. I say they will go with the cheaper option that still offers increased performance and the ability to create entire product lines based on the number of cores available, rather than how much performance is available in wholly different core types. Also a striking similarity to the CPU market.
Sucks for the driver developers, looks pretty awesome for the consumer. Like anything else that's a paradigm shift it takes time for developers to wrap their heads around it, but eventually it will get there. How many applications were out there for the consumer that could make use of multiple CPU cores 3 years ago? How about now?
5) That was exactly my point. I don't agree with AMD's marketing calling it an entirely new series of products, but you can't really change that. It was their engineers doing exactly what they needed to do in order to get this core ready for its future destinations and that's what you need to realize. It might not have wound up as the performance champion, but they've invested too much into it to start from scratch on something that works better. We'll see what they can do with performance enhancements come R700 now that they've got the size and power thing taken care of.
Don't want to quote directly cuz that would be a gigantic post. But anyway...
1) AMD is not in a bad place. It's not the BEST place, but it certainly isn't bad. They compete at nearly the same level price wise. The fact that it's labeled as a "mid-range" vs "high end" product has nothing to do with price/performance ratio. Lowest price of a 9600GT I just did a quick search for is $170, 3870 is $190. Not too far off.
Also I meant most of my comments to be directed towards their development process and roadmap, not necessarily as they stand against the competition. They are in a good place for keeping inline with where they want to go in the future.
2) If AMD's hardware has stagnated then so has NVidia's. The only major change they've had to the "high end" is the G92 core, which in essence was the exact same thing that happened with the R670 core, smaller and less power hungry. The 9800 series is supposed to simply be a refresh of the G92 core...so what is the argument again?
I don't really get your more expensive comment, they aren't anywhere close to being in a 3dfx type situation. If you want to look at sales figures check out OEM numbers, these halo products represent a very very small fraction of the company's overall market strength. 3dfx was caught in the wrong place at the wrong time where the OEM market was absolutely and completely owned by Intel for the most part with the scraps going to some other vendors, including ATi. If 3dfx didn't have the products to compete with NVidia in the high end performance market (which they didn't), they had nothing else to fall back on.
3) I was comparing the development cycles of the two coming very close to each other now, not directly comparing their architecture per se. Obviously the GPU is still a specialized processor that is very different from x86, but the use of them has become something with a much broader range. Something that software outside of games can make use of. You can't tell me you don't see similarities between this and FPU development in the past. It's exactly the same direction GPUs are heading. Which leads me to...
4) If you don't want any part of multi-gpu solutions I suggest you start or fund a company to do otherwise and do it quick. They might seem "hackish" but there is no possible way that the industry can sustain the previous method of making bigger and better by brute force alone. You'd be looking at gigantic development times to create the next gen 1bil+ transistor monolithic core. To make a simple and elegant design that works and be able to connect them together is what AMD wants, and what it's been trying to form with the R6xx. Just what exactly did you expect to see in the Fusion processor? Or bulldozer platform if that ever becomes a reality.
Yes, there are diminishing returns on simply adding more GPUs, but the benefits are still there and the only other option is hugely expensive and time consuming. I say they will go with the cheaper option that still offers increased performance and the ability to create entire product lines based on the number of cores available, rather than how much performance is available in wholly different core types. Also a striking similarity to the CPU market.
Sucks for the driver developers, looks pretty awesome for the consumer. Like anything else that's a paradigm shift it takes time for developers to wrap their heads around it, but eventually it will get there. How many applications were out there for the consumer that could make use of multiple CPU cores 3 years ago? How about now?
5) That was exactly my point. I don't agree with AMD's marketing calling it an entirely new series of products, but you can't really change that. It was their engineers doing exactly what they needed to do in order to get this core ready for its future destinations and that's what you need to realize. It might not have wound up as the performance champion, but they've invested too much into it to start from scratch on something that works better. We'll see what they can do with performance enhancements come R700 now that they've got the size and power thing taken care of.
TheBluePill
2[H]4U
- Joined
- Aug 17, 2005
- Messages
- 3,773
This is actually the Most telling Metric of where ATI Stands and is probably an acurate sample of the entire enthousiest market;