ATi sux due to AMD?

[wyz135]

I was thinking of something, since AMD bought over ATi a years ago, is it because of AMD, ATi's performance and rating went down? If AMD didn't bought over ATi, will there be a difference?

[Radiator]

The R600 core is to blame for the crappy performance . The fact that it cannot do regular AA with ROPs is to blame , not merging with AMD .
However , newer DX10 games are going to force cards to use their shaders for AA , so buying an ATi card now would make it more future-proof .
But then again , I could be wrong .

[wyz135]

ATi + AMD = The new AMD

That is their fusion project that they're thinking next year, what is that, I heard that is a Chipset powered by both GPU and CPU. Is it ATi graphic card powered by AMD processor? Or AMD processor powered by some ATi technology?

[Radiator]

It'll be a CPU+GPU , but I doubt it'll beat any dedicated CPU/GPU chipsets . Neat idea though .

[wyz135]

last time they came out with a idea what Quad-Core, powered by 4 processors in a single chipset, now another new idea coming out, CPU and GPU powered in a single chipset..... Anyway I hate AMD, I prefer Intel...

[Radiator]

Why do you hate AMD? Lol .
AMD ruled with AMD 64 , intel rules with Conroe , and I'm pretty sure AMD will rule again with the upcoming Barcelona core . But then again , I could be mistaken .

[wyz135]

AMD's CPU really produce damn hell lots of heat, really. I have a AMD Athlon XP and I need 7 cooling fans to keep it cool, and AMD can't really over-clock, AMD's CPU is really unstable. Intel's CPU is stable and you can over-clock it how much you want, but don't over-clock too much

[Radiator]

AMD XP was worse than P4 in my opinion... and neither of them were good - both were heaters and couldn't OC good . AMD x2 is using an older and less efficient architecture than Conroe CPUs .
But AMD's new chip called Barcelona should be ... powerful .

[wyz135]

for Intel, it's common to see people Over-Clocked up by 600MHz or higher, for AMD, even trying to squeeze out 200MHz is difficult

[Hatman]

Athlon XP kicked the penitums butt! My dads is never over 30degrees idle with stock heatsink, if you need 9 cooling fans I think you should have a closer look at your chip!

Clock for clock Athlons VS Penitum was even more in AMD's favor then core2duo to Athlon X2's even now.


They arent that bad though, if you check out THWG review on the 65nm Black Edition 5000x2 you'll see it hit 3.3ghz or so while running quite cool, they still have their kicks left. Im not expecting AMD Phenom to do very well, its clock speeds may be ok for retail but now when overclocking is so easy, Intel is still an obvious choice. Especially with the new 45nm cores.

TBH as far as ATI goes I think they just shoved as much as they could onto their cards so they just had a product while they got to 65/55nm, now they have 55nm and can potentially kick NVIDIAS ass using high end dual cards, since their power consumption is super low.

[Radiator]

for Intel, it's common to see people Over-Clocked up by 600MHz or higher, for AMD, even trying to squeeze out 200MHz is difficult

Well , AMD's performance isn't measured in Ghz or anything . The 5000+ number shows how it's performance is , roughly compared to the chip's competition .
As for Athlon XP vs. Pentium 4 's , it really depends on which generation of Pentium 4's and Athlon's you're talking about .

[wyz135]

anyway I still prefer Intel, and I really like the Core2 Extreme, seems powerful and I hope I had one

[Radiator]

Yeah , Conroe certainly does beat the shit out of AMD's chips .
But in terms of gaming performance , the difference is around 1-2 FPS max .
Conroe is more energy efficient though .

[kagebushin93]

i agree wif wyz135 as intel has a more stable processor as for a amd...it produce quite lots of heat & is not capable of overclocking so much..it can be overclock though but its really not a good idea.

[kagebushin93]

I was thinking of something, since AMD bought over ATi a years ago, is it because of AMD, ATi's performance and rating went down? If AMD didn't bought over ATi, will there be a difference?
And btw,do u tink AMD is the only fault?ATi also makes the responsibility to improve on his OWN company .Just put the matter aside on AMD buyin over ATi causing it to degrade in performance.

[Alucard]

I'm just waiting for ATI's new DX10.1 cards and Intel's new 45nm Processors. Then I can build my new system. Now what would be really interesting is if Intel got ATI. But Intel is working on making its own GFX cards.

[Radiator]

DirectX 10.1 doesn't add almost anything though .
I'd just wait for the next generation of high-end cards ( R700 , G100 ) .

[kagebushin93]

hey wads the hurry when DX10.1 is just about to be released, its still expensive trying to buy DX10 cards,y dun u just w8 for even better cards..first generation version of a card is often not so good...

[Alucard]

I never get the first gen of anything, I always get the next or 3rg gen of cards. Im not in a hurry to buy anything. And when I do it last me a good time before I have to upgrade again. Im still gonna get me a x1950 till more things become DX10 required and not just supported. So yea till then DX9 its still looking good to me.

[-RK]

Well, DX10.1 seems like it's gonna flop on games... it doesn't add much and forces AA.

On the discussion of AMD Proccys vs Intel...

The pendulum swings back and forth. The Athlon XP couldn't keep up with the P4 Prescott for long, but the Athlon 64 outdid just about anything that Intel could field for a while. Right now the 65nm Core 2s are kicking around the Athlon X2s. Who can say who will come out on top next?

It seems like architecture has been the big kicker since the Athlon 64's prime (although the integrated memory controller didn't hurt).

[Hatman]

ATI sux due to its shader clock being locked to the core clock whereas nvidias is like 2x as high.

If anythings to be learned from the 8800GT, its that shaders = performance.

[Radiator]

ATI sux due to its shader clock being locked to the core clock whereas nvidias is like 2x as high.

If anythings to be learned from the 8800GT, its that shaders = performance.

That's why ATi put an outrageous number of shader processors on their video cards , to make up for it . ATi's card's shader horsepower is still much higher than that of nVidia's card's .

[-RK]

ATi Cards seem to have some serious architecture design issues compared to NV cards.

They FAIL at AA and that locked shader clock is something to compensate for...

They seem to be learning pretty quick though. 38x0 series made a pretty serious improvement. Of course, so did the 8800GT...

[Radiator]

ATi Cards seem to have some serious architecture design issues compared to NV cards.

They FAIL at AA and that locked shader clock is something to compensate for...

They seem to be learning pretty quick though. 38x0 series made a pretty serious improvement. Of course, so did the 8800GT...

That's because the R6** cards do their AA with their shaders , while the GeForce 8 cards do their AA with their ROPs .
And also , even though the ATi cards seem to have a large number of shader processors , they don't really have an efficient architecture .

[Coguar]

I think that Ati will be a Better with AMD, look 3870 vs 2900 XT same perfomance but one 2900 XT spend power like 3870 crossfire..

They just need litle time to full functionality..

[Radiator]

The HD 3870 is a fair bit faster than the HD 2900 XT , by the way .

55nm process cut down the power usage and heat output , and 55nm process and the 256bit memory bus really cut down the cost of producing them , hence they are pretty cheap .

[MrWizard6600]

Well, DX10.1 seems like it's gonna flop on games... it doesn't add much and forces AA.

On the discussion of AMD Procs vs Intel Procs...

The pendulum swings back and forth. The Athlon XP couldn't keep up with the P4 Prescott for long, but the Athlon 64 outdid just about anything that Intel could field for a while. Right now the 65nm Core 2s are kicking around the Athlon X2s. Who can say who will come out on top next?

It seems like architecture has been the big kicker since the Athlon 64's prime (although the integrated memory controller didn't hurt).

never were wiser words spoken.

Its identical in the graphics card industry. The Radeon 9800/9700/9600s outperformed and undercut everything from the Geforce FX 5 series. Infact, the situation at the time was very similar to the situation were in now, only with reversed roles. Nvidia promised a radical new design for the FX5 series, but it massively under-delivered. The Radeon 9 series was the clear choice across the board. I was a bit of an Nvidiot at the time, so I was somewhat oblivious to the negative comments made against the FX 5 series, but it was terrible. The FX5200, a highly touted product, barely outperformed a similarly priced last-gen product, the Geforce Ti4200 (sound familiar? HD2900XT from X1950XTX?).

the Radeon 9600 remains, to this day, the best selling card ATI has ever released.

The Radeon X series against the Geforce 6 series was a rather interesting battle, with Nvidia touting SM3.0 as best they could, since only NV chips supported it at the time, all the while ATI's Radeon's were mopping the floor with Nvidia's solutions. Nvidia was able to save some face with the re-introduction of SLI: people who wanted the best, could buy two 6800s and run em in tandem, outperforming the X800/850XT PE, which were superb cards, and performed excellently if you could ever get your hands on a crossfire master. But you couldn't.

The X1k series vs the 7 series was neck-and-neck. ATI adopted a similar 16 pipe technology as their X800 cards, only this time they supported SM3.0 calls, and shrunk the die (130nm to 90nm), allowing for much higher clocks. Nvidia made only a partial die shrink (130nm to 110nm) but introduced an extra 8 pipes for a total of 24. The X1800 and 7800 were completely neck and neck everywhere you looked. When ATI tweaked the pipe architecture for 16pipes with 3 shaders on each pipe (for 48 shaders), we ended up with a monster card that Nvidia just couldnt answer until they completed the die shrink (110nm to 90nm) on the 7800 to give us the 7900. The X1900 and 7900, again, performed neck and neck across the board, but this time for separate reasons. The X1900, with its mass of shaders, choked on any Texture heavy games but exceeded at any shader intensive games. Vice-versa for the more traditional 24piped G71 (7900GTX).

This year, Nvidia made quite a showing: the 8800 was a superb product. Ati shot high with its response. They promised us a beast: a graphics card with a 512bit bus and a massive amount of super powerful stream processors. Unfortunately, something went arai in engineering, and the ROPs failed and we ended up with a card that had ROP work being done on the general purpose shaders (stream processors). Both these GPUs are so sophisticated at this point people have modded em to get running versions of Unix, the operating system, to run on the GPUs alone.

Nvidiots take note: we only now see the fallout of what happens when ATI cant respond: a stagnent market. Nvidia has not released a refresh to their product line because they have no need to. In my books at least 2007 will go down as the most stagnant year in recent GPU history, with only a handful of new products, and nothing ground breaking, not even a refresh. It looks like there will be no 8900GTX. The G92 might have been, and indeed seems to showcase some technologies that would fit a refresh perfectly (tweaked architecture (change to TMUs), and a die shrink), but, alas, there is no enthuisast G92. As it stands, the G92 is a freak. A proper G9*, with a 512bit or even 384bit bus, would have fit the 8900GTX bill perfectly, but alas, its not to be.

Next up, we have D9E (aka G100) vs the R700. It looks as though ATI is going to try to take us in a modular direction, with "stackable" chips. The R700 is in itself not a very powerful chip, but early rumors would suggest that it is so modular that it is possible to cram FOUR of them onto a single wafer, allowing them to run in tandem (could you imagine 4 such cards running on say 790FX? 4 GPUs per card, and 4 cards... 16 R700s!! AMD has been hinting at this in road maps. They call it the "Leo" platform (like the spider platform)). The G100 looks to be just another beast from Nvidia. We have heard rumors of a card (single SKU) which can produce over a teraflop of floating point calculations in a single second, from Nvidia, as the 9 series. There are few rumors about the D9E aside from the obvious: speculation about 256 stream processors (2x the amount on the 8800 GTX), built on a 65/55/45nm process, and a bigger/faster memory bus, and faster (GDDR5) memory.

The pendulum swings in the graphics card market just like it does in the CPU market. The Swinging goes back much further then I've shown you, I just don't know who was superior. I'm sure any of the admins could fill us in on the Geforce 4 vs Radeon 8, Geforce 3 vs Radeon 7, Geforce 2 vs Rage, Geforce vs ... well I have no idea, your going really far back now (like, pre-formation of Nvidia and ATI far back. Like.. 3DFX was instyle and Voodoo still had money far back).

[Radiator]

The HD 2900 XT IS significantly faster than the X1950 XTX , lol .
Agreed with the rest of it though .
I can't help but wonder how the hell are they going to stack 4 GPUs on one PCB while still having a small enough PCB to fit in most computer cases . And the heat.. lol .

[wyz135]

ATI sux due to its shader clock being locked to the core clock whereas nvidias is like 2x as high.

If anythings to be learned from the 8800GT, its that shaders = performance.

Oh.... you're so wrong, you can't determine the overall shader operations by looking at their Shader Clock alone, ATi has always crazy amount of shader processors, and look at the Radeon HD2900XT's shader operations, it beats the GeForce 8800Ultra, which means it can produce better image quality than the GeForce 8800Ultra, but what it lacks is speed. Getting it may be more future-proof because later games will be using damn lots of shader power

[Radiator]

Actually , nVidia's cards have 2x the shader processor count , than the R6** cores . The 320 stream processors is a lie . It actually has 64 shader processors , that can do 5 things at a time , but well.. the architecture isn't that good itself .

The 8800 GT allocates it's shaders differently from the G8* cards , hence the good performance with a lower amount of shader operations and 256bit memory bus .

[-RK]

But the 8800gt and G92 are just well designed period.

[Radiator]

But the 8800gt and G92 are just well designed period.

No arguements there...

Designed better than the G80... if it wasn't it would have gotten so close to the 8800 GTX , considering the amount of ROPs and the narrow memory bus ( 256bit only ) .

[wyz135]

never were wiser words spoken.

Its identical in the graphics card industry. The Radeon 9800/9700/9600s outperformed and undercut everything from the Geforce FX 5 series. Infact, the situation at the time was very similar to the situation were in now, only with reversed roles. Nvidia promised a radical new design for the FX5 series, but it massively under-delivered. The Radeon 9 series was the clear choice across the board. I was a bit of an Nvidiot at the time, so I was somewhat oblivious to the negative comments made against the FX 5 series, but it was terrible. The FX5200, a highly touted product, barely outperformed a similarly priced last-gen product, the Geforce Ti4200 (sound familiar? HD2900XT from X1950XTX?).

the Radeon 9600 remains, to this day, the best selling card ATI has ever released.

The Radeon X series against the Geforce 6 series was a rather interesting battle, with Nvidia touting SM3.0 as best they could, since only NV chips supported it at the time, all the while ATI's Radeon's were mopping the floor with Nvidia's solutions. Nvidia was able to save some face with the re-introduction of SLI: people who wanted the best, could buy two 6800s and run em in tandem, outperforming the X800/850XT PE, which were superb cards, and performed excellently if you could ever get your hands on a crossfire master. But you couldn't.

The X1k series vs the 7 series was neck-and-neck. ATI adopted a similar 16 pipe technology as their X800 cards, only this time they supported SM3.0 calls, and shrunk the die (130nm to 90nm), allowing for much higher clocks. Nvidia made only a partial die shrink (130nm to 110nm) but introduced an extra 8 pipes for a total of 24. The X1800 and 7800 were completely neck and neck everywhere you looked. When ATI tweaked the pipe architecture for 16pipes with 3 shaders on each pipe (for 48 shaders), we ended up with a monster card that Nvidia just couldnt answer until they completed the die shrink (110nm to 90nm) on the 7800 to give us the 7900. The X1900 and 7900, again, performed neck and neck across the board, but this time for separate reasons. The X1900, with its mass of shaders, choked on any Texture heavy games but exceeded at any shader intensive games. Vice-versa for the more traditional 24piped G71 (7900GTX).

This year, Nvidia made quite a showing: the 8800 was a superb product. Ati shot high with its response. They promised us a beast: a graphics card with a 512bit bus and a massive amount of super powerful stream processors. Unfortunately, something went arai in engineering, and the ROPs failed and we ended up with a card that had ROP work being done on the general purpose shaders (stream processors). Both these GPUs are so sophisticated at this point people have modded em to get running versions of Unix, the operating system, to run on the GPUs alone.

Nvidiots take note: we only now see the fallout of what happens when ATI cant respond: a stagnent market. Nvidia has not released a refresh to their product line because they have no need to. In my books at least 2007 will go down as the most stagnant year in recent GPU history, with only a handful of new products, and nothing ground breaking, not even a refresh. It looks like there will be no 8900GTX. The G92 might have been, and indeed seems to showcase some technologies that would fit a refresh perfectly (tweaked architecture (change to TMUs), and a die shrink), but, alas, there is no enthuisast G92. As it stands, the G92 is a freak. A proper G9*, with a 512bit or even 384bit bus, would have fit the 8900GTX bill perfectly, but alas, its not to be.

Next up, we have D9E (aka G100) vs the R700. It looks as though ATI is going to try to take us in a modular direction, with "stackable" chips. The R700 is in itself not a very powerful chip, but early rumors would suggest that it is so modular that it is possible to cram FOUR of them onto a single wafer, allowing them to run in tandem (could you imagine 4 such cards running on say 790FX? 4 GPUs per card, and 4 cards... 16 R700s!! AMD has been hinting at this in road maps. They call it the "Leo" platform (like the spider platform)). The G100 looks to be just another beast from Nvidia. We have heard rumors of a card (single SKU) which can produce over a teraflop of floating point calculations in a single second, from Nvidia, as the 9 series. There are few rumors about the D9E aside from the obvious: speculation about 256 stream processors (2x the amount on the 8800 GTX), built on a 65/55/45nm process, and a bigger/faster memory bus, and faster (GDDR5) memory.

The pendulum swings in the graphics card market just like it does in the CPU market. The Swinging goes back much further then I've shown you, I just don't know who was superior. I'm sure any of the admins could fill us in on the Geforce 4 vs Radeon 8, Geforce 3 vs Radeon 7, Geforce 2 vs Rage, Geforce vs ... well I have no idea, your going really far back now (like, pre-formation of Nvidia and ATI far back. Like.. 3DFX was instyle and Voodoo still had money far back).

lol, MrWizard6600, I love your statement of the history of battle the between ATi and Nvidia:)

[Radiator]

I wrote something similar several months ago on a blog in gamespot... lol .

[wyz135]

looks like ATi just kick nvidia hard in the ass with their upcoming release of Radeon HD3870 X2...... I wonder what will nvidia do to answer for that....

[wyz135]

Well , AMD's performance isn't measured in Ghz or anything . The 5000+ number shows how it's performance is , roughly compared to the chip's competition .
As for Athlon XP vs. Pentium 4 's , it really depends on which generation of Pentium 4's and Athlon's you're talking about .

The AMD's way of measuring speed is the same old traditional way, by GHz, the 5000+ you're talking about is the model number. AMD use model number because their core speed is always slower than intel's. Intel also used the same method for the Core2 series, using numbers, but in different systems like E4300, E6600, E6750 etc. Saying Pentium4 vs the Athlon XP, actually is NorthWood vs Athlon XP, Prescott and Cedar Mill vs Athlon 64

[Deathmarx]

Different system though. AMD use Hyper Transport and Intel still play around with FSB. It's like, you've got a wall. AMD use a phone and ring from one side of the wall to the other. When AMD's technology gets better that 'phone' becomes quicker. Intel on the other hand just make a giant hole in the wall, and as their tech progresses they just make the hole bigger. I'm also going to say I like Intel, however their tactics of spamming the CPU market is absolutely shocking. E8400 and E8500 anyone? (new 45nm Penryns) they're like 100MHz difference with an extra $100 slapped ontop.

What ATI need is a decent card. with 512 bit Memory Bus (future-proofing don't flame me :p) and 1GB of the Hynix GDDR 5 clocked to outrageous speeds. Something in the esque of what the X1950XTX was, this huge, monster of a card. That's what they need.

[wyz135]

Different system though. AMD use Hyper Transport and Intel still play around with FSB. It's like, you've got a wall. AMD use a phone and ring from one side of the wall to the other. When AMD's technology gets better that 'phone' becomes quicker. Intel on the other hand just make a giant hole in the wall, and as their tech progresses they just make the hole bigger. I'm also going to say I like Intel, however their tactics of spamming the CPU market is absolutely shocking. E8400 and E8500 anyone? (new 45nm Penryns) they're like 100MHz difference with an extra $100 slapped ontop.

What ATI need is a decent card. with 512 bit Memory Bus (future-proofing don't flame me :p) and 1GB of the Hynix GDDR 5 clocked to outrageous speeds. Something in the esque of what the X1950XTX was, this huge, monster of a card. That's what they need.

AMD always critisize FSB to be a slow technolgy and bottlenecking CPU overall performance.

Saying that ATi sux due to AMD is not true, after the merge, the boss of ATi remains the president of the GPU department, only thing is that all money earned by ATi will go into AMD's pocket, that's all

[Radiator]

Different system though. AMD use Hyper Transport and Intel still play around with FSB. It's like, you've got a wall. AMD use a phone and ring from one side of the wall to the other. When AMD's technology gets better that 'phone' becomes quicker. Intel on the other hand just make a giant hole in the wall, and as their tech progresses they just make the hole bigger. I'm also going to say I like Intel, however their tactics of spamming the CPU market is absolutely shocking. E8400 and E8500 anyone? (new 45nm Penryns) they're like 100MHz difference with an extra $100 slapped ontop.

What ATI need is a decent card. with 512 bit Memory Bus (future-proofing don't flame me :p) and 1GB of the Hynix GDDR 5 clocked to outrageous speeds. Something in the esque of what the X1950XTX was, this huge, monster of a card. That's what they need.

The HD 2900 XT uses a 512bit memory bus ... and the 1 GB GDDR4 one was as fast as the 512MB GDDR3 one in almost all applications .
A card needs a good solid core before getting riddiculously high memory bandwidth . The bandwidth on the HD 2900 XT was mostly left unused ... the 256bit HD 3870 with a slightly higher core clock beat the HD 2900 XT in the majority of applications . There wont be a need for a 512bit memory bus anytime soon , mate .

[-RK]

As for AMD's naming system... it is a bit outdated, but there is more to it than that.

Long, long ago, in the time of the athlon XP the number was supposed to correspond with the clock speed of Intel CPU it was competing with. An Athlon XP 1500+ was supposed to be at least (thus the "+") as effective as a p4 at 1.5 Ghz. Of course, things didn't always work out that way...

AMD really should have changed the naming system with the original Athlon64. It just blew away the P4.

[Deathmarx]

A card needs a good solid core before getting riddiculously high memory bandwidth . The bandwidth on the HD 2900 XT was mostly left unused ... the 256bit HD 3870 with a slightly higher core clock beat the HD 2900 XT in the majority of applications . There wont be a need for a 512bit memory bus anytime soon , mate .

How long ago were we saying we wouldn't need a 128-bit/256-bit ?
Very valid point about the good solid core however. The 3870 >is< a good step in the right direction, price cutting Nvidia :D

[Radiator]

I said not anytime soon , not that we would never be needing a 512bit memory bus .

[-RK]

maybe if they make a card with 2 high clocked g92 based chips using the same memory bus...

that's a scary thought though, it'd handle Crysis just fine...


edit:
oops... ATi beat me to it. HD3870 x2 is now out (not on the same memory bus but still...)