• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

More Pipelines or Higher Core Frequency ???

More Pipelines or Higher Core Frequency ???

  • ATi, Higher Core Frequencies...

    Votes: 38 21.5%

  • nVidia, More Pipelines...

    Votes: 139 78.5%
  • Total voters
    177
L

LeviathanV

Limp Gawd
Joined
May 25, 2005
Messages
247
When I first saw ATi new line-up, i was unsettled by the fact that they kept the same number of pipelines than the previous gen. I have not seen this happen from one gen to another since the time when pipelines weren´t an issue...

Now seeing that both cards perform very competitive with one another which company do you think made the rigth choice? nVidia for increasing its number of pipelines while keeping speeds the same, or ATi for increasing clock frequencies and keeping the same number of pipelines?

This is not an ATi Vs nVidia post.... though it might seem like it... is more like the title says... which do you prefer, more pipes or higher clocks...

Well, we should probably edit your poll choices then ;) One is labeled ATI, the other Nvidia - DL
 
I'm thinking more pipes was the better route...nvidia got their parts out sooner and with better availability because they didn't have to fuss around with a new process...and the 7800GTX is only marginally slower than the x1800xt...even though the xt uses more power, runs hotter and takes up two slots...

I think the XT makes a more interesting enthusiast card for its high overclockability...but from a business perspective...nvidia took the much more sensible route...
 
^eMpTy^ said:
I'm thinking more pipes was the better route...nvidia got their parts out sooner and with better availability because they didn't have to fuss around with a new process...and the 7800GTX is only marginally slower than the x1800xt...even though the xt uses more power, runs hotter and takes up two slots...

I think the XT makes a more interesting enthusiast card for its high overclockability...but from a business perspective...nvidia took the much more sensible route...
Click to expand...

Yeah, but ATi took the first step towards the future... I mean whats next for nVidia 48 pipe cards... I mean sooner or later they have to get better performance without increasing the number of pipelines...I think ATi is in a better position for future generation...
 
How so? Having more pipelines is a more solid foundation for the future than simply upping the clock frequency of the gpu and memory. If clock speed was all that mattered none of us would own AMD64 chips but instead intel's high ghz cpus.
 
peacetilence said:
How so? Having more pipelines is a more solid foundation for the future than simply upping the clock frequency of the gpu and memory. If clock speed was all that mattered none of us would own AMD64 chips but instead intel's high ghz cpus.
Click to expand...

I was actually typing something like that when I refreshed the thread :) I totally agree.
 
peacetilence said:
How so? Having more pipelines is a more solid foundation for the future than simply upping the clock frequency of the gpu and memory. If clock speed was all that mattered none of us would own AMD64 chips but instead intel's high ghz cpus.
Click to expand...

ATi did more than increase the clocks for the R520. Though it is really an intel-style chip, power consumption-wise.
 
LeviathanV said:
Yeah, but ATi took the first step towards the future... I mean whats next for nVidia 48 pipe cards... I mean sooner or later they have to get better performance without increasing the number of pipelines...I think ATi is in a better position for future generation...
Click to expand...

That doesn't make any sense...you could just as easily say nvidia is in a better position for the future since their chip already has 24 pipelines and they dont' have to resort to overclocking to get more performance...

What matters is what is available now...and right now the nvidia cards are cooler, quieter, and provide the same performance because they have more pipelines...so I would say that's a win for the more pipelines camp...
 
^eMpTy^ said:
That doesn't make any sense...you could just as easily say nvidia is in a better position for the future since their chip already has 24 pipelines and they dont' have to resort to overclocking to get more performance...

What matters is what is available now...and right now the nvidia cards are cooler, quieter, and provide the same performance because they have more pipelines...so I would say that's a win for the more pipelines camp...
Click to expand...

Something to think about. nVidia is working with 110nm now and gained 8 pipes from 130nm. Imagine what 90nm will do for them, heh. ATI has already blown their load with 90nm, it seems, nVidia has some headroom with their NV40/G70 style superscalar architecture.
 
Im not saying clock frequencies is all that matter... I love AMD, to me is a cheaper way to get similar or better performance in games than intel... but ATi has managed to beat nVidia or in the worst case match it in performance and still keeping the same amount of pipelines...

My point is that nVidia can´t just simply keep increasing the number of pipelines...
 
More Pipes > More Clock Speed

Look at processors now? We are moving away from High Speed single core, and going to a lowerspeed but Multi Core. Sure Clockspeed is good, but its all aobut how much work is getting done in the end.
 
I think more pipes are better because you wouldn't have to rely on an inherently unreliable game of getting very high clock speeds. Getting high clock speeds is good but you can't rely simply on clock speed forever.

Also, I think ATI is making good steps up in designing gpu's better, case in point the ram b/w system of the X1x00, the system is highly suffisticated and I think I read in a review that it can already handle up to 3ghz ram speed without modification. That's sweeet.

~Adam
 
LeviathanV said:
Im not saying clock frequencies is all that matter... I love AMD, to me is a cheaper way to get similar or better performance in games than intel... but ATi has managed to beat nVidia or in the worst case match it in performance and still keeping the same amount of pipelines...

My point is that nVidia can´t just simply keep increasing the number of pipelines...
Click to expand...

Like I said before, they're only at 110nm, and going from 16 to 24 (a 50% increase) gave around a 100% increase in performance. Until unified shaders are required, nVidia will be able to scale their chips and compete with ATi just fine
 
^eMpTy^ said:
That doesn't make any sense...you could just as easily say nvidia is in a better position for the future since their chip already has 24 pipelines and they dont' have to resort to overclocking to get more performance...

What matters is what is available now...and right now the nvidia cards are cooler, quieter, and provide the same performance because they have more pipelines...so I would say that's a win for the more pipelines camp...
Click to expand...

Do you have a link of a comparison of the 7800GTX and X1800XT heat disapation?....not saying your wrong I just havent seen this heat problem with X1800XT your talking about.
 
DASHlT said:
Do you have a link of a comparison of the 7800GTX and X1800XT heat disapation?....not saying your wrong I just havent seen this heat problem with X1800XT your talking about.
Click to expand...

The dual slot cooler is a hint., as is the temps people have with their 125mhz slower clocked X1800XL's.
 
If I remember correctly Nvidia when they first started the Geforce card line basically had more horsepower than any other card and it worked for them just making it faster and faster to brute force run games with raw muscle. When you do this I would think you hit a ceiling on what is actually cost effective and doable tech wise at the time. ATI looks as if they tried a brute force way of making things run better but like anything there are pros and cons. ATI has problems with the new cards being that they are expensive using expensive hardware to run faster the heat and the sheer size space. The price is the biggest factor on the market though. I am sure someone can build the most expensive card using the most expensive stuff on earth at the time but will anyone buy it? Not many if at all. Tradeoffs. Nvidia is about the same on the 6 and 7 series clockwise my 6600gt runs faster than a reference GTX when I over clock it. ATI seems to be always a bit more pricey compared to Nvidia the past two gens and considering the performance remains the best bang for the buck. The ATI AIW cards though are the best product they have IMO and are worth the cost.
 
robberbaron said:
The dual slot cooler is a hint., as is the temps people have with their 125mhz slower clocked X1800XL's.
Click to expand...

Again linkage? proof?....just want proof
 
I would personally go with more pipeline, higher the speed isn't always better as we can see from AMD a lower clock speed beating a Intel CPU which is at a higher clock personally, pipeline and the speed isn't everything it is also the achitecture of the GPU. Sometime Higher clock speed doesn't mean it will perform better sometime all it does it generate more heat, i think ATI is acting like intel to be honest seems like they r just pumping up the clock speed.
 
.::MAGE::. said:
If I remember correctly Nvidia when they first started the Geforce card line basically had more horsepower than any other card and it worked for them just making it faster and faster to brute force run games with raw muscle. When you do this I would think you hit a ceiling on what is actually cost effective and doable tech wise at the time. ATI looks as if they tried a brute force way of making things run better but like anything there are pros and cons. ATI has problems with the new cards being that they are expensive using expensive hardware to run faster the heat and the sheer size space. The price is the biggest factor on the market though. I am sure someone can build the most expensive card using the most expensive stuff on earth at the time but will anyone buy it? Not many if at all. Tradeoffs. Nvidia is about the same on the 6 and 7 series clockwise my 6600gt runs faster than a reference GTX when I over clock it. ATI seems to be always a bit more pricey compared to Nvidia the past two gens and considering the performance remains the best bang for the buck. The ATI AIW cards though are the best product they have IMO and are worth the cost.
Click to expand...

Actually last two gens, they have both been equally ridiculously overpriced ... both the 6800 Ultra and the 800XT-PE came out at $500, and nVidia came out months ago at $600 bucks and ATi just came out the same...

honestly, a 100 dollar increase in top price in just over a year from gen-to-gen ....
Before you know it, next year we´ll be selling a kidney in order to buy the latest 700 dollar vid card for christmass...

its not so hard to believe that in a couple of years, video cards could cost over $1k considering thats about what high end procesors cost now....
 
robberbaron said:
http://www.tbreak.com/reviews/article.php?id=407

Clock comparsons here.
Ask R1ckca1n about the temps of his XL
Click to expand...

From tbreak

One of the things that we’ve always liked about ATI cards is the size and power requirements which are generally lower than nVidia cards. While ATI keeps the power usage to about 100W on the X1800XT, the size is about as big as nVidia’s 7800GTX card- which is a bit too much. Also, ATI utilizes a dual slot cooling solution that is sure to wake your neighbors up at full speed but then again, this is just a reference card and hopefully retail cards will have a quieter solution.

So 100W of power which is maybe as much as the X850 series?.....also this is a reference cooler, 3rd party vendors can (and probably will) drop the dual slot solution....

No where in the review did I see a heat disapasion comparison.....

I just want proof, Im building a new system soon and heat is a factor when I build this new rig.
 
There are pro's and con's to each method. And I don't think there is one method that is best in all cases. Also you have a set transister budget that your are aiming for so everthing that you want to add into your design has to be weighed against your budget and this has a direct impact on your transister count. So you have to consider which is more important to your design? More pipelines or extra features? Again there is no one right answer for every case. Adding more pipes is not going to help if you can not keep them busy so you have to increase everything to ensure all the pipes are "happy" and that eats away at your transitor budget.

Finally on the power comparisons I have seen a lot of bad assuptions. Remember you have a lot of things effecting those results. For example the x1800 die has 21 million more transistors than the 7800 die. So even if everything else was the same (clock speed, ect). The ATI part would draw more power as it has more transistors. Then remember that in these power measurements they are using the WHOLE board. Thus when comparing the x1800xt to the 7800 GTX factor in that the ATI part has x2 the memory and running faster which is going to pull up its power ratings. Thus comparing the x1800xl to the 7800GT is kind of interesting as the both have the same memory speed/size and the have simular power draws eventhough the x1800xl is running 100 mhz faster with more transistors....
 
The way I see it, it's better to design the GPU with more pipelines than higher clock speeds. The reason behind this is that when you shrink the die or move to a new process or even just tweak the current design you can always increase the clock speed, but it is impossible to add more pipes without a major redesign of the GPU.

Now if we apply this to the current generation of cards we have an interesting picture.

ATi - how much higher can clocks go in say 6 months? Memory might be able to go higher, but not the core. And there is no way to add more pipelines to the design.

nVidia - since the GPUs are running ~400Mhz and overclocking to near 500MHz, with some revisions to the silicone or with a die shrink I am thinking they can reach 600MHz default before any overclocking. Will 600MHz increase performance at the same level? Memory is the same as ATi since they are at the mercy of the manufacturer.
 
Finally on the power comparisons I have seen a lot of bad assuptions. Remember you have a lot of things effecting those results. For example the x1800 die has 21 million more transistors than the 7800 die. So even if everything else was the same (clock speed, ect). The ATI part would draw more power as it has more transistors. Then remember that in these power measurements they are using the WHOLE board. Thus when comparing the x1800xt to the 7800 GTX factor in that the ATI part has x2 the memory and running faster which is going to pull up its power ratings. Thus comparing the x1800xl to the 7800GT is kind of interesting as the both have the same memory speed/size and the have simular power draws eventhough the x1800xl is running 100 mhz faster with more transistors....
Click to expand...

QFT
 
if this isnt one of those ati vs nvidia things then how come the choices say "ati, higher core frequencies" "nvidia, more pipelines" :confused:
 
pipes/clock speeds

they are both trying to achieve the same goal, just getting there differently

remember, pipes aren't like they use to be, and neither are clock speeds

the G70 for example has different components of the core clocked at different speeds

we will see more of this definition of pipelines and clock speeds changing over time
 
who cares about pipes or clock speed

dual core gpus need to become standard
 
From the reviews I have read, the 7800 series runs cooler, with less noise, less power consumption and with a single slot design. Looks like more pipes are working better for them. Considering the XT and GTX are about the same in benchmarks and the GT is above the XL. NVIDIA could easily overclock the GTX with a big ole dual slot cooler and slap an extra 256mb of memory on it. Of course, right now they have the fastest card out there with out doing that.
 
xdkimx said:
if this isnt one of those ati vs nvidia things then how come the choices say "ati, higher core frequencies" "nvidia, more pipelines" :confused:
Click to expand...

i was wondering about that too
hmmm

as usual the answer lies in personal preference
:D
 
a link to a page with links to other sites new Radeon reviews. Not sure if anyone wants this or not, but here it is.

http://www.3dchipset.com/#531

I think most of these are the ones that the [H] front page had early october.
 
X1800 vs 7800 is turning in to a Pentium 4 vs Athlon 64 like fight

P4 less work per clock, but higher clocks to keep up ("radical new design")
A64 (or Pentium M) slower clocks, more work per clock, more traditional design

It just kinda remains to be seen if the 7800 is more like a Pentium 3, Pentium M or AthlonXP or Athlon64 vs. the X1800 being a Pentium 4. It will take some time to tell as drivers pan out & newer games are released.


Ramping up clock speeds I think has been shown to eventually come to an end. I think its become obvious that they need to use the process shrink to make a bigger (wider) chip that can do more parallel processing, like dual cores, etc.... b/c you can't keep shrinking the die AND going faster, b/c you end up with a tiny piece of the sun (hot enough for fusion!) in your PC.
 
I prefer more pipes, up to a point. More pipes = more transistors = larger die = easier to cool. High clock speed (presumed fewer pipes/transistors, smaller die) = high thermal density, harder to cool. All other things being equal, the smaller dies are cheaper to produce. However, the higher clock speed requirements to get the same performance out of a smaller chip means that yields aren't as good, so it largely balances out. The manufacturing process's peculiarties have a lot to do with what is the best decision on GPU design.

NV did an excellent job on 7800GTX - they were able to add more pipes and clock speed, while simultaneously holding the line on power consumption :cool: However, this may not be true next time around.
 
This time around, I tend to think that nvidia's way of doing it wins this cycle...
I also think that ati has something really good going with their memory controller.
If they can mature (make it easier for themselves) the process of optimizing how the memory is used in the drivers for specific games, then it will help out a lot.


as far as the heat issue goes that was talked about a few posts ago, I don't know that I've seen any numbers on it, but I get the impression that they run very hot. Which, seeminly, would imply that they are hotter than nvidia's cards. Tough to say without numbers.
 
Ugh. Guys it's not an ATI vs. Nvidia poll.



True in the last genereation you have seen that one company did one thing and the other did another thing. In the corporate world, that doen't make a "rule" or even start a trend. What a company does today has absolutely no bearing on what they are going to do tomorrow. Maybe next gen you will see that Nvidia runs out of die space, and they decide to ramp clocks instead because they got some deal on high speed RAM. Or maybe even something as simple as ATI got a new head engineer, and he decides to implement some new ideas. There are so many forces on what actually gets released, you would be foolish to think that a company at large is going down some idealistic course solely in honor of some great idea. Products just happen to be this way today.

BTW I voted pipelines. Adding pipelines holds to a more traditional ideal of chip design, and actually has way more headroom and less possible production issues than ramping clocks. You also need to consider the rest of the chip logic. It's not just pipelines and MHz. Unfortunately most of the chip logic takes a very long time to make changes to, and it's very expensive to engineer. Take a look at Intel. They used the same core logic from the Pentium in no less than 3 other lines before they redesigned. They mostly got away with clock ramping and minor tweaks (i.e. MMX).
 
LeviathanV said:
Actually last two gens, they have both been equally ridiculously overpriced ... both the 6800 Ultra and the 800XT-PE came out at $500, and nVidia came out months ago at $600 bucks and ATi just came out the same...

honestly, a 100 dollar increase in top price in just over a year from gen-to-gen ....
Before you know it, next year we´ll be selling a kidney in order to buy the latest 700 dollar vid card for christmass...

its not so hard to believe that in a couple of years, video cards could cost over $1k considering thats about what high end procesors cost now....
Click to expand...

Yes I have to agree the whole situation from both companies has gotten out of hand. Right now the major costs to a system are your graphics card and CPU which is rather insane but whatever. They make up around 2/3 of my upgrade costs.
 
vmerc said:
Ugh. Guys it's not an ATI vs. Nvidia poll.
Click to expand...

I know that, my post wasn't completely alluding to your poll.

I voted for clock frequencies btw. There really are a lot of things to look at overall beside those two things.
 
Ultra Wide said:
The way I see it, it's better to design the GPU with more pipelines than higher clock speeds. The reason behind this is that when you shrink the die or move to a new process or even just tweak the current design you can always increase the clock speed, but it is impossible to add more pipes without a major redesign of the GPU.

Now if we apply this to the current generation of cards we have an interesting picture.

ATi - how much higher can clocks go in say 6 months? Memory might be able to go higher, but not the core. And there is no way to add more pipelines to the design.

nVidia - since the GPUs are running ~400Mhz and overclocking to near 500MHz, with some revisions to the silicone or with a die shrink I am thinking they can reach 600MHz default before any overclocking. Will 600MHz increase performance at the same level? Memory is the same as ATi since they are at the mercy of the manufacturer.
Click to expand...

600mhz defualt gpu on a 7800gtx refresh?

not likely even if they went to 90nm process.
 
However, even a die shrink to 90nm, holding the line on everything else, would still be a huge win for NV, as they would get more dies per wafer (and presuming yields are good) more $$$ out of each 7800 they sold.

However, I don't look for NV to move to 90nm any time soon. The are fairly cautious on new processes, test piloting a new process with a mid range GPU before moving the flagship parts to a new process. ATI used to do this as well (9600XT's R360 130nm low-k before 9800XT) and had the same pattern going when they developed the R430 X800(XL) on 110nm, but decided to go straight to 90nm with R520. As we can see, that transitition was far from smooth.

Also, don't forget that the extra transistors added by havign 512MB of DDR3 will contribute substantially to power consumption X1800XTs.
 
You must log in or register to reply here.
Back
Top