How will ATI respond to Fermi? (poll included)

[happy medium]

The x1950xtx was the better performer (especially with newer sharder heavy games) but the point was that the 8800 launch had completely different circumstances. Nvidia didn't have to say anything to hype up their upcoming 8800 or squish any negative rumors regarding it. They had no competition out before the launch, nor in the upcoming months after the launch, much less an entire line up of parts out to compete against before they had even a single 8 series card out. In short Nvidia could afford to say nothing about the 8800, doing so didn't lose customers or hurt them.

With Fermi, I just don't think it makes business sense to sit tight lipped while your competitor steals DX11 enthisast customers one after the next, unless Fermi doesn't do enough to justify hyping. If the rumors are true, what can Nvidia say?

Yes, I see your point also but, Fermi does beat a 5870,and I'm sure Nvidia knew this for 3/4 months so why not tell people. Wierd ha? Why hide it? Hey our card is faster, heres the early benchies, wait for us. But nothing..why hide a winner?

 

[cbn]

Yes, I see your point also but, Fermi does beat a 5870,and I'm sure Nvidia knew this for 3/4 months so why not tell people. Wierd ha? Why hide it? Hey our card is faster, heres the early benchies, wait for us. But nothing..why hide a winner?

Maybe Nvidia was deciding on where to set the voltage/power? (Or something along those lines)

If rumors are true a 188 watt ATI video card (HD5870) will end up faster than a 225 watt Nvidia card (470 GTX) as far as average FPS goes. (However, Minimum FPS maybe be better for the Nvidia card)

P.S. Has anyone ever considered what the overclocking headroom differences will be between Fermi and Cypress?

 

[happy medium]

Guess we lived on different planets? http://www.dailytech.com/article.aspx?newsid=4441

This one is subdued compared to most.

It was released just after that though. 3 or 4 weeks and they didn't know jack shit.

Read all the rumers and guessing in the comments on that page at the bottom.

 

[sandorski]

Yes, I see your point also but, Fermi does beat a 5870,and I'm sure Nvidia knew this for 3/4 months so why not tell people. Wierd ha? Why hide it? Hey our card is faster, heres the early benchies, wait for us. But nothing..why hide a winner?

One reason could be because the amount of better performance was not high enough to keep people waiting for its release.

 

[happy medium]

Maybe Nvidia was deciding on where to set the voltage/power? (Or something along those lines)

If rumors are true a 188 watt ATI video card (HD5870) will end faster than a 225 watt Nvidia card (470 GTX).

P.S. Has anyone ever considered what the overclocking headroom differences will be between Fermi and Cypress?

To be honest I really don't care how many watts it pulls or how hot it is or ATI for that matter.

I put high end video cards in my high end case with good airflow , with my high end psu.

If rumers are true the gtx won't have to overclock much ha?
I do like the feeling of getting a real good overclock .(looks at his 5750 in signiture)

 

[happy medium]

One reason could be because the amount of better performance was not high enough to keep people waiting for its release.

Lets hope so man or we are all screwed. I really want a new card ,and I really don't want to crossfire. Or pay an arm and a leg for 50% more perfromance.

 

[cbn]

If rumers are true the gtx won't have to overclock much ha?

What rumors are you talking about?

Are you talking about Fermi's minimum frame rate performance with tessellation turned on? (Fermi is super strong in that category)

 

[ronnn]

It was released just after that though. 3 or 4 weeks and they didn't know jack shit.

Read all the rumers and guessing in the comments on that page at the bottom.

So the forum posters knew jack shit? Major shock there. I must just be smarter. Not saying I knew the details (just not a detail person), but I knew it was on time, on budget and worked very well. At the same time ati was a mess.

 

[happy medium]

So the forum posters knew jack shit? Major shock there. I must just be smarter. Not saying I knew the details (just not a detail person), but I knew it was on time, on budget and worked very well. At the same time ati was a mess.

True that!

 

[happy medium]

What rumors are you talking about?

Are you talking about Fermi's minimum frame rate performance with tessellation turned on?

Good question.

The one where the gtx 470 is about even with the 5870 and the gtx 480 is 10% faster already.

or the many others Ive seen.

What I meant was if one gpu (gtx 480) is beating another gpu (5870) by 10% from the start with crappy probrobly very beta drivers, it won't need much of a overclock to beat a overclocked 5870 either, when all is said and done. Amd is on it's 5th set of drivers for the 5870 and many hot fixes as well,remember that.!

Sound logical?

Edit: ALso Nvidia has the game makers lined up to optimize thier games on thier hardware and historicly has great drivers.

 

[cbn]

I am surprised nobody voted option number 2.

 

[happy medium]

I am surprised nobody voted option number 2.

Mabe on the 6870. I don't see it happening this round. Games are using more core and not as much bandwidth latley. Mabe thats why they castrated the 57xx series so badly. Figured this was the direction the game developers are going to go?

I like option, up the core speed and lower prices. Both 1 & 4

 

[cbn]

Games are using more core and not as much bandwidth latley. Mabe thats why they castrated the 57xx series so badly. Figured this was the direction the game developers are going to go?

Well Fermi has 50% more bandwidth than Cypress and does better with Tessellation as far as minimum frame rates go.

This begs the question how much of that minimum frame rate performance is due to memory bandwidth/VRAM amount and how much of it is due to the core?

I just don't see why ATI can't do the same for its own products?

 

[mfenn]

Well Fermi has 50% more bandwidth than Cypress and does better with Tessellation as far as minimum frame rates go.

This begs the question how much of that minimum frame rate performance is due to memory bandwidth/VRAM amount and how much of it is due to the core?

I just don't see why ATI can't do the same for its own products?

Radeon 5000 series owners could do an experiment to figure out what the bottleneck is.

1. Underclock the core and memory by the same proportion (say 75%). Measure the performance.
2. Return the core to stock. Measure the performance.
3. Underclock the core again and return the memory to stock. Measure the performance

This would give you a good idea as to whether a particular gave is memory bandwidth or shader limited.

I'm willing to bet its the shaders, but I could be wrong.

 

[cbn]

Radeon 5000 series owners could do an experiment to figure out what the bottleneck is.

1. Underclock the core and memory by the same proportion (say 75&#37. Measure the performance.
2. Return the core to stock. Measure the performance.
3. Underclock the core again and return the memory to stock. Measure the performance

This would give you a good idea as to whether a particular gave is memory bandwidth or shader limited.

I'm willing to bet its the shaders, but I could be wrong.

BFG10K did an experiment with HD5770 where he underclocked core and memory by 20% and then measured the performance drops using average FPS.

After doing this he found HD5770 a fairly balanced card with a slight bias to the core (In other words he did find memory bandwidth to limit performance).

 

[happy medium]

BFG10K did an experiment with HD5770 where he underclocked core and memory by 20% and then measured the performance drops using average FPS.

After doing this he found HD5770 a fairly balanced card with a slight bias to the core (In other words he did find memory bandwidth to limit performance).

However, he did not measure minimum FPS. Fermi's main advantage over Cypress according to the Unigine Benchmarkwith tessellation turned on is in the category of minimum FPS (not average FPS).

I am wondering how much ATI could gain with the faster memory chips? Would this help close the major gap Fermi has over Cypress?

I think it would close 2 gaps though, performance and price.
All card being equal, more people will buy Nvidia.
If Amd keeps the price lower they sell cards. Just like last round.

For Amd to compete well against the Nvidia name and PR machine they have to sell similar performance and cheaper.

 

[cbn]

I think it would close 2 gaps though, performance and price.
All card being equal, more people will buy Nvidia.
If Amd keeps the price lower they sell cards. Just like last round.

For Amd to compete well against the Nvidia name and PR machine they have to sell similar performance and cheaper.

How much would it cost ATI to do a small run of Cypress with 7 Gbps GDDR5? Does anyone know the price difference?

 

[happy medium]

Depends on how much new memory cost. If it was cost effective, I think they would have done it 5 months ago.

 

[cbn]

Depends on how much new memory cost. If it was cost effective, I think they would have done it 5 months ago.

ATI didn't have a competitor 5 months ago.

 

[happy medium]

ATI didn't have a competitor 5 months ago.

New faster memory cost more money but why not use the new memory 5 months ago. I'm sure the prices where the same or close to the same as they are now.

They would have launched at over 400$,mabe thats too much? Shit we didn't even need the cards that fast then. For what games? MW2? I play that at high settings with my 5750. Mabe Crysis?,but that game is 2 years old ,who cares?

 

[SlowSpyder]

What rumors are you talking about?

Are you talking about Fermi's minimum frame rate performance with tessellation turned on? (Fermi is super strong in that category)

A couple of things that I think are worth pointing out.

In the Heaven bench, when Nvidia ran it they did not have AA turned on. It makes you wonder, why would they not have AA turned on? This is a high end part, anyone who buys it is going to want performance with AA. I'm willing to bet Nvidia ran the bench in a manner that was to show their part in the absolute best light. I'm guessing that when you use this card with settings that you would likely use on a high end part that the scores are closer then what Nvidia showed. Also, as others have said in different threads, tesselation isn't used in a way it would be in games in the Heaven benchmark. I really think people are looking too muc into that Nvidia video with the Heaven benchmark. It was created by Nvidia using settings that no one with this level card will use on a synthetic benchmark. I would definitely wait for DX11 games to be tested at settings you'll actually use by a third party in this case.

Good question.

The one where the gtx 470 is about even with the 5870 and the gtx 480 is 10% faster already.

or the many others Ive seen.

What I meant was if one gpu (gtx 480) is beating another gpu (5870) by 10% from the start with crappy probrobly very beta drivers, it won't need much of a overclock to beat a overclocked 5870 either, when all is said and done. Amd is on it's 5th set of drivers for the 5870 and many hot fixes as well,remember that.!

Sound logical?

Edit: ALso Nvidia has the game makers lined up to optimize thier games on thier hardware and historicly has great drivers.

My guess is that Nvidia will launch with pretty mature drivers. Remember, this part was supposed to be out in, what, October? Nvidia has had working silicon for months. Nvidia will have had the advantage of being able to tweak and mature their drivers through many months of delays of an actual product launch.

 

[mfenn]

BFG10K did an experiment with HD5770 where he underclocked core and memory by 20% and then measured the performance drops using average FPS.

After doing this he found HD5770 a fairly balanced card with a slight bias to the core (In other words he did find memory bandwidth to limit performance).

The memory of the 5700's was cut down much more than the core, and the result is balanced. Which would mean that the 5800's have an abundance of memory bandwidth.

 

[happy medium]

A couple of things that I think are worth pointing out.

In the Heaven bench, when Nvidia ran it they did not have AA turned on. It makes you wonder, why would they not have AA turned on? This is a high end part, anyone who buys it is going to want performance with AA. I'm willing to bet Nvidia ran the bench in a manner that was to show their part in the absolute best light. I'm guessing that when you use this card with settings that you would likely use on a high end part that the scores are closer then what Nvidia showed. Also, as others have said in different threads, tesselation isn't used in a way it would be in games in the Heaven benchmark. I really think people are looking too muc into that Nvidia video with the Heaven benchmark. It was created by Nvidia using settings that no one with this level card will use on a synthetic benchmark. I would definitely wait for DX11 games to be tested at settings you'll actually use by a third party in this case.




My guess is that Nvidia will launch with pretty mature drivers. Remember, this part was supposed to be out in, what, October? Nvidia has had working silicon for months. Nvidia will have had the advantage of being able to tweak and mature their drivers through many months of delays of an actual product launch.

Brings me back to the "why didn't they show the benchmarks" earlier argument.
If they had good drivers, (as you say) and a faster card, release the benchies and stop AMD from selling so many direct x 11 cards back in November.

So your saying Amd drivers were not mature when they released? They managed to increase speed quite a few percent from Sept. ha?

I guess Nvidia will see no performance in the upcomming 5 driver releases?
AMd has had 5 months and at least 5 driver releases. They should be at there cards optimal perfromance wouldn't you think? Mabe they been holding back performance for the Fermi? I'll add that to the rumer pile.
Or are you saying that nvidia has mature drivers for a card with no final clocks up untill ..what 2 weeks ago?

 

[happy medium]

The memory of the 5700's was cut down much more than the core, and the result is balanced. Which would mean that the 5800's have an abundance of memory bandwidth.

I'd have to disagree. The 5770 core and memory is exactly half of a 5870. 128bit vs 256 bit. 1/2.

Same for the 5750 and 5850, half exactly core and memory.

 

[mfenn]

I'd have to disagree. The 5770 core and memory is exactly half of a 5870. 128bit vs 256 bit. 1/2.

Same for the 5750 and 5850, half exactly core and memory.

You are right. For some reason I was thinking that they turned the memory clocks way down on the 5700's.

I retract my previous statement.