Some hard Fermi performance # info

[pasta4u]

What report shows the GTX380 launching in June?

semiaccurate.com which has been correct on everything about fermi up to this point is reporting it.

Also some others in forums are confirming it and they are more than just forum trolls. Do some poking around. gts 360 in late march perhaps early april (when volume actually starts to hit) and then June for gtx 380. They are hoping on revision B3 to give them much better yields on the part.

I'm not going to respond to the other part of your post because I do agree with it for the most part. But 5870s will not easily hit 1200 on the core. Pushing the voltage to 1.35(which is more than a slight boost), most cards will top out at anywhere from 1050-1075. The only way I cans see a 5870 hitting 1.2Ghz is if you're using sub-ambient cooling. Like the Phenom IIs, Cypress seems to respond better to lower temperatures than voltage.

I've seen 5870s at 1.2ghz using the stock cooler and this http://www.thermalright.com/new_a_page/product_page/vga/vrm/product_vga_cooler_vrm.html

Most spend time cooling the wrong thing.

A refresh will of course not only come with faster clocks but better board components that will incease the speeds at which they can clock

 

[heatlesssun]

semiaccurate.com which has been correct on everything about fermi up to this point is reporting it.

Charlie D. Yeah, he can be right a lot but at the same time he's NEVER going to anything but bad mouth nVidia, so you do have to take a lot he says with a grain of salt. And remember, he's said the SAME thing about every nVidia for years, too hot, too much power, and by in large its not the huge mess he always makes it out to be.

nVidia very well knows the power issues.

 

[pasta4u]

Charlie D. Yeah, he can be right a lot but at the same time he's NEVER going to anything but bad mouth nVidia, so you do have to take a lot he says with a grain of salt. And remember, he's said the SAME thing about every nVidia for years, too hot, too much power, and by in large its not the huge mess he always makes it out to be.

nVidia very well knows the power issues.

ITs very believable that nvidia will launch a higher end gtx 380 a little over a month after the gts 360.

ATI did the same with its high end. 5870 first and then a little over a month later the 5970. Yields of the lower chip (or single chip in ati's case) will make for higher retail amounts than the higher end part.

Its not like he is talking about 3-6 month delay of the gtx 360. ITs just a little over a month according to him.

 

[Kowan]

Okay, so there is no "report" of the GTX380 shipping later.

 

[RamonGTP]

Actually there is, the report just isn't from nVidia.

 

[Kowan]

An editorial is not a report.

 

[pasta4u]

There is no report of the gts 360 shipping in late March either.

In fact there is no report on when any fermi products will ship nor at what clock speeds.

 

[kadozer]

There's a green carrot on a stick with some nice koolaid on the side from the Nvidia hype marketing team. Come get your fill.

 

[Vermillion]

I do find it funny that people are crying wolf, foaming at the mouth or just being an annoying fanboy.

If Fermi wasn't 20-30% faster then HD5870 I'd be worried because at 6 months late or more it needs to be considerably faster. We can then expect the 5000 series to drop in price and a refresh to pull within about 10% or less of Fermi but cost less. Meaning a win-win for everybody, whether you're a fanboy or not.

 

[LeninGHOLA]

You're missing the point.

This is a reply to LeninGHOLA

Nah, I get it. My problem is that we are trusting Nvidia to be honest with their numbers right now as well as trusting these numbers to be accurate on release day. Not to mention we don't even know which card they were benching, 448sp or 512?. There was a single comparison on a single game, which is a TWIMTBP title. Driver tweaks can easily push performance levels up, yet those same tweaks may not make it into release drivers because performance may suffer in other games. Personally, I can't base my opinion on such a small sample. How can we say the GF100 is 30% faster if only a single bench was done on a single game?

 

[kadozer]

Exactly no solid numbers from the same marketing team that claimed GTS250 > 5870 while back. A single bench on a NVIDIA testing rig (pun intended)
People are too gullible these days...blind lemmings that fall for marketing now thats funny.
I'm not exactly joining the fermi circle jerk just yet... if that makes me a AMD fanboy so be it label me or anyone else that questions the lack of real performance not shown.

 

[XMAN245]

Exactly no solid numbers from the same marketing team that claimed GTS250 > 5870 while back. A single bench on a NVIDIA testing rig (pun intended)
People are too gullible these days...blind lemmings that fall for marketing now thats funny.
I'm not exactly joining the fermi circle jerk just yet... if that makes me a AMD fanboy so be it label me or anyone else that questions the lack of real performance not shown.

While it was a Marketing gimic and a bad one at that, the game they used was B:AA and it was PhysX enabled that has the GTS250 betaing a 5870. So while what they said was true, it was a bad marketing thing just the same.

 

[magoo]

I have some screen capture somewhere running the Far Cry 2 medium ranch with pretty much the same settings on a single HD 5870...........and I'm pretty certain my numbers were better than nvidia's.

I'll hunt them up once I get home.
I do recall that I was impressed that my single 5870 was as good or better than my SLI'd GTX 285s on the same machine,same settings.

So I'm not holding my breath for the GF100, yet.
Like I've said before, if the GF 100 smokes my 5870s I'll gladly sell them and buy two.....I'm not partial either way.....except I'd miss EyeFinity.

 

[XMAN245]

I have some screen capture somewhere running the Far Cry 2 medium ranch with pretty much the same settings on a single HD 5870...........and I'm pretty certain my numbers were better than nvidia's.

I'll hunt them up once I get home.
I do recall that I was impressed that my single 5870 was as good or better than my SLI'd GTX 285s on the same machine,same settings.

So I'm not holding my breath for the GF100, yet.
Like I've said before, if the GF 100 smokes my 5870s I'll gladly sell them and buy two.....I'm not partial either way.....except I'd miss EyeFinity.

Well, if you bought 2 fermis, you would still be able to use your 3 monitor setup as Nvidia is brings nFinity(there stupid sounding name for Eyefinity) for both Fermi and GT200 line cards.

 

[Mr. K6]

Well, if you bought 2 fermis, you would still be able to use your 3 monitor setup as Nvidia is brings nFinity(there stupid sounding name for Eyefinity) for both Fermi and GT200 line cards.

Is it as good as Eyefinity though? Again, something we don't know.

I'm glad to see there's some other people still living in reality here and waiting for more info.

 

[XMAN245]

Is it as good as Eyefinity though? Again, something we don't know.

I'm glad to see there's some other people still living in reality here and waiting for more info.

lets see

support for 3+ monitors(SLI needed this gen) check
Does 3D support in mulit monitor check
Does regular 2D support aswell Check(ATI current implimation of Eyefinity)
Going to have bezel management out of box Check(ATI fix this yet)

Aside from needing SLI for 3+ monitors, which is a problem for those only wanting 1 card, seems to me it gonna be atleast on par.

 

[Mr. K6]

lets see

support for 3+ monitors(SLI needed this gen) check
Does 3D support in mulit monitor check
Does regular 2D support aswell Check(ATI current implimation of Eyefinity)
Going to have bezel management out of box Check(ATI fix this yet)

Aside from needing SLI for 3+ monitors, which is a problem for those only wanting 1 card, seems to me it gonna be atleast on par.

Because you have actual performance data and comparisons in hand, as well as implementation and driver support data, and compatibility data not only for games but also motherboards and other hardware. Again, if you dumb things down to the reasoning of a first grader, everything probably makes sense without any information. I mean, Fermi will probably have the color green on it somewhere, and from stoplights green means "go," so we know it'll be totally more awesome than the 5870 which is red and means "stop." Right!?

 

[Tamlin_WSGF]

lets see

support for 3+ monitors(SLI needed this gen) check
Does 3D support in mulit monitor check
Does regular 2D support aswell Check(ATI current implimation of Eyefinity)
Going to have bezel management out of box Check(ATI fix this yet)

Aside from needing SLI for 3+ monitors, which is a problem for those only wanting 1 card, seems to me it gonna be atleast on par.

Bonus to Nvidia for backwards support as well. It can be good or not, but I am glad Nvidia chose to support multimonitor gaming.

One question for you though:

Do you find a dual gpu solution equal to a single card solution in games? Lets say 5970 vs. Fermi?

 

[heatlesssun]

lets see

support for 3+ monitors(SLI needed this gen) check
Does 3D support in mulit monitor check
Does regular 2D support aswell Check(ATI current implimation of Eyefinity)
Going to have bezel management out of box Check(ATI fix this yet)

Aside from needing SLI for 3+ monitors, which is a problem for those only wanting 1 card, seems to me it gonna be atleast on par.

Because you have actual performance data and comparisons in hand, as well as implementation and driver support data, and compatibility data not only for games but also motherboards and other hardware. Again, if you dumb things down to the reasoning of a first grader, everything probably makes sense without any information. I mean, Fermi will probably have the color green on it somewhere, and from stoplights green means "go," so we know it'll be totally more awesome than the 5870 which is red and means "stop." Right!?

XMAN245 took the information available to him and made logical conclusions. Obviously he and we don't have enough data to compare "nFinity" with Eyefinity vigourously but with what we know it looks to be about on par with Eyefinity with the exception of needing two GPUs.

 

[Mr. K6]

XMAN245 took the information available to him and made logical conclusions. Obviously he and we don't have enough data to compare "nFinity" with Eyefinity vigourously but with what we know it looks to be about on par with Eyefinity with the exception of needing two GPUs.

You're assuming that two different implementation methods will function the same. Maybe they will, but that's quite a leap considering we have no data.

 

[GoldenTiger]

You're assuming that two different implementation methods will function the same. Maybe they will, but that's quite a leap considering we have no data.

Why do you think it won't work roughly the same in effectiveness, sir? Video cards themselves are "different implementation methods", and the evidence we have supports a positive outcome moreso than a negative one.

 

[heatlesssun]

You're assuming that two different implementation methods will function the same. Maybe they will, but that's quite a leap considering we have no data.

And it would be a leap as well assuming that nFinity isn't as good as Eyefinity.

 

[XMAN245]

Bonus to Nvidia for backwards support as well. It can be good or not, but I am glad Nvidia chose to support multimonitor gaming.

One question for you though:

Do you find a dual gpu solution equal to a single card solution in games? Lets say 5970 vs. Fermi?

Personally, I believe you can if the performance to cost is there. Example, I spent 300 bucks for 2 eVGA GTX260 216s and my performance is in better than a single GTX285, which the cheapest one you can get new costs more.

So if the performance of a full blown 512SP part fermi is = or greater than the 5970 and cost less or equal to the 5970, then yeah you can. But should the full blown fermi be slower and cost more, the win would be to get the 5970. This is also dependant on ATI not castrating their profit margins to keep Nvidia from selling cards as it would be hard pressed for someone to warrant getting a full blown fermi part for 500 if ATI reduces the 5970 to say 400.

But again, it all comes down to the end user

 

[magoo]

Well, if you bought 2 fermis, you would still be able to use your 3 monitor setup as Nvidia is brings nFinity(there stupid sounding name for Eyefinity) for both Fermi and GT200 line cards.

If I'm not mistaken, the nvidia three monitor response is "stereo surround" which would by it's definition require 120Hz LCDs and stereo vision goggles. Look here: http://www.hardocp.com/image.html?i...VHN0ekRuc2RfMl8yOV9sLmdpZg==&ref=YXJ0aWNsZQ== which makes it an entirely different animal......they do mention "non-stereoscopic" 3D across monitors of same resolution, but not in much depth.

EyeFinity doesn't require that and I have three Dell 2408s so I'd be SOOL on that front.

If I read the latest beta release by ATI, they include a bezel management solution.....but of course this is not the "official" 10.1 release. We'll see next week what's up when ATI releases the 10.1 Catalyst.

 

[GoldenTiger]

If I'm not mistaken, the nvidia three monitor response is "stereo surround" which would by it's definition require 120Hz LCDs and stereo vision goggles. Look here: http://www.hardocp.com/image.html?i...VHN0ekRuc2RfMl8yOV9sLmdpZg==&ref=YXJ0aWNsZQ== which makes it an entirely different animal.

EyeFinity doesn't require that and I have three Dell 2408s so I'd be SOOL on that front.

If I read the latest beta release by ATI, they include a bezel management solution.....but of course this is not the "official" 10.1 release. We'll see next week what's up when ATI releases the 10.1 Catalyst.

Incorrect, nVidia already has confirmed nFinity requiring SLI and not requiring "3d surround" tech to span 3 monitors:

http://www.nvidia.com/object/gf100.html

...expanding your real estate across 3 panels for the ultimate “inside the game” feeling. Not ready to enter the third dimension? NVIDIA® Surround™ will also be supported with GF100 on non-stereo displays.

 

[Mr. K6]

Why do you think it won't work roughly the same in effectiveness, sir? Video cards themselves are "different implementation methods", and the evidence we have supports a positive outcome moreso than a negative one.

We don't have any evidence, that's my point. In case you haven't noticed, I'm not a fan of extolling something I know little about. The only thing concrete I know is that it will require SLI for triple monitor support, which is already a huge negative for me. Beyond that, I'll wait and see.

And it would be a leap as well assuming that nFinity isn't as good as Eyefinity.

Like I said, using that one metric, that it needs SLI, I'd already say it's inferior. But I can't say anymore until I actually see it in use and some associated performance and quality data. Because I don't agree that it's awesome, you're assuming I think it sucks. I don't; I'm neutral until I have more data.

Do you find a dual gpu solution equal to a single card solution in games? Lets say 5970 vs. Fermi?

Nope, and I think that's the biggest thing Fermi has going for it - it's one GPU. After my GTX295, I'll always choose the fastest single card solution over any dual card solution.

 

[corrupted223]

At least we all know that once Kyle and the boys get their hands on Fermi we will know for sure what it can and cannot do, and what you can convince it to do via overclocking. Until then I will wait to pass judgement on untested hardware.

Regardless of what happens I certainly have a feeling that us consumers are going to benefit from the competition in the coming months.

 

[trudude]

I guess all that needs to happen now is for some games to come out that actually stresses out the hardware. I know that the only game worth mentioning that actually needs a 5870 or better to perform right maxed out @ 1920x1200 is Crysis. Sure Crysis 2 is coming soon, but from what I have heard it is programmed with efficiency in mind so I am fairly certain that most 5850+ owners are going to be playing that game fine. So again I will ask the question WHERE ARE THE GAMES?!?!

 

[heatlesssun]

The only thing concrete I know is that it will require SLI for triple monitor support, which is already a huge negative for me. Beyond that, I'll wait and see.Like I said, using that one metric, that it needs SLI, I'd already say it's inferior.

That said, single GPU solutions, even the 5870 are a bit underpowered to drive newer games at maximum settings with larger screen Eyefinity setups. So yes, you can do multi-monitor cheaper with Eyefinity. But many are going to want more than one card anyway.

 

[Tamlin_WSGF]

Personally, I believe you can if the performance to cost is there. Example, I spent 300 bucks for 2 eVGA GTX260 216s and my performance is in better than a single GTX285, which the cheapest one you can get new costs more.

My thoughts too. Sometimes, its possible to even get larger performance with dual card at same price point, so SLI might not be more costly then a single solution. Especially in cases where you need an adapter. In other situations, where your MB doesn't support SLI, you might get higher costs though.

What I fear most about a dual card solution, is the possibility of sync/lag issues by using ports from two different cards. We know little of Nvidia surround yet, so there are still a lot unanswered questions until its released.

Edit: Rys at Beyond3d mentioned somethings about this:

Surely frame presentation has to be synced across all three heads, regardless of whether it's a combined framebuffer approach or not? I'm struggling to see a situation where the rendering on the peripheral displays is less important than the middle one in terms of frame time.

You really want one GPU for all heads, or one GPU per head. No other combination really makes sense to me in terms of performance.

Wavey's right in that it's not just the clock source that a > 2 head board needs to supply. There's a full display engine needed too (there's a lot going on after the write to the framebuffer).

http://forum.beyond3d.com/showpost.php?p=1377247&postcount=109

Good news for ATI and Nvidia multi-monitor users, is that major upcoming titles will have ingame support for this. This is about Eyefinity, but it should work on Nvidia's solution as well as TH2G and softTH:

http://www.widescreengamingforum.com/forum/viewtopic.php?f=64&t=18046

 

[Snowdog]

That said, single GPU solutions, even the 5870 are a bit underpowered to drive newer games at maximum settings with larger screen Eyefinity setups. So yes, you can do multi-monitor cheaper with Eyefinity. But many are going to want more than one card anyway.

For more casual users like me. I just want three monitor support for non gaming use. My 8800GT provides all the one monitor gaming performance I need.

The end result is that for my next card, Nvidia is completely out because they don't have three monitor support. I don't have SLI and I don't want to cram a crappy PCI card in there. So ATI remains the only game in town and NVidia is SOL.

NVidia will retake the lunatic fringe high end card market(as expected), but does that really matter anymore? I would imagine less and less people are interested in $400+ graphics cards as time goes by.

 

[heatlesssun]

For more casual users like me. I just want three monitor support for non gaming use. My 8800GT provides all the one monitor gaming performance I need.

The end result is that for my next card, Nvidia is completely out because they don't have three monitor support. I don't have SLI and I don't want to cram a crappy PCI card in there. So ATI remains the only game in town and NVidia is SOL.

NVidia will retake the lunatic fringe high end card market(as expected), but does that really matter anymore? I would imagine less and less people are interested in $400+ graphics cards as time goes by.

For a single card three monitor solution your point is well taken.

Lunatic fringe high end? There's nothing wrong wanting fantastic performance. Where are people coming off like $500 is a lot of money for a video card? I was paying that much for them in the 90's, so in today's terms its even less money now.

 

[YeuEmMaiMai]

wow who would have known that Nvidia releases some extremely vague information about their new product and everyone just basically says OGM it's da bomb. We don't even know the setup used video card wise or if this thing was cherry picked or the drivers tweaked to give these performance #s. Do you even know what the visual quality settings were? I'm not buying into this crap until cards hit retail and we actually get to see what they are up to.

 

[GoldenTiger]

wow who would have known that Nvidia releases some extremely vague information about their new product and everyone just basically says OGM it's da bomb. We don't even know the setup used video card wise or if this thing was cherry picked or the drivers tweaked to give these performance #s. Do you even know what the visual quality settings were? I'm not buying into this crap until cards hit retail and we actually get to see what they are up to.

Read the article; all of this is already answered.

 

[heatlesssun]

wow who would have known that Nvidia releases some extremely vague information about their new product and everyone just basically says OGM it's da bomb. We don't even know the setup used video card wise or if this thing was cherry picked or the drivers tweaked to give these performance #s. Do you even know what the visual quality settings were? I'm not buying into this crap until cards hit retail and we actually get to see what they are up to.

This was a press release not a product launch so I'm not really sure who's going OMG since there's nothing to go OMG over at this point.

As hardware enthusitats we like to know what's coming down the pike and speculate and that's all we're really doing at this point.

Until there's something to buy obvisouly I'm not buying anything like everyone else for now. I'll see what nVidia has in the GF100. Now if I have to wait until JUNE to get the high end model, hmm, might be an issue for me.

 

[Mr. K6]

That said, single GPU solutions, even the 5870 are a bit underpowered to drive newer games at maximum settings with larger screen Eyefinity setups. So yes, you can do multi-monitor cheaper with Eyefinity. But many are going to want more than one card anyway.

Not really. I play every game maxed at 2560x1600 and 4xAA, and the only game I don't get 60FPS+ consistently on is Crysis. Not saying that rendering 75% more pixels (24" 3-way Eyefinity) or 200% more (30" 3-way Eyefinity) is going to be the same, but it will easily be playable. One thing I love about AMD cards, I guess from their architecture, is that they handle high resolutions very well. When I first got my 30", I was playing games with no problems at 2560x1600 on my 4870 512MB with 60FPS+; as long as I kept AA off so it didn't over work the memory limitations. That said, there's been tons of tech demos, especially right here at [H], that have demonstrated games playing very well on one 5870 using three 30" screens. Then there's the practical uses beyond gaming for why triple display set-ups on one card are great.

 

[heatlesssun]

Not really. I play every game maxed at 2560x1600 and 4xAA, and the only game I don't get 60FPS+ consistently on is Crysis. Not saying that rendering 75% more pixels (24" 3-way Eyefinity) or 200% more (30" 3-way Eyefinity) is going to be the same, but it will easily be playable. One thing I love about AMD cards, I guess from their architecture, is that they handle high resolutions very well. When I first got my 30", I was playing games with no problems at 2560x1600 on my 4870 512MB with 60FPS+; as long as I kept AA off so it didn't over work the memory limitations. That said, there's been tons of tech demos, especially right here at [H], that have demonstrated games playing very well on one 5870 using three 30" screens. Then there's the practical uses beyond gaming for why triple display set-ups on one card are great.

You're making my point as I said MAXIMUM settings. One 5870 simply isn't enough for most new games at 5760x1200 and maximum settings.

 

[Snowdog]

For a single card three monitor solution your point is well taken.
Lunatic fringe high end? There's nothing wrong wanting fantastic performance. Where are people coming off like $500 is a lot of money for a video card? I was paying that much for them in the 90's, so in today's terms its even less money now.

In the 90's you couldn't buy a decent whole PC for $500. It seems ridiculous to drop this kind of money on a graphics card these days, when you get so much performance for very little money. It seems to be much more about having the biggest e-peen than it does about have the performance you need to play games.

I don't know for certain, but I get the impression that the $500 video card market is a shrinking one. As it stands now, I don't know any of my friends that would even consider spending $200 on video card anymore except me and we are almost all computer nerds (sw devs/sw testers), but we are getting older now. Maybe twentysomethings are routinely dropping $500 on video cards. I don't know.

 

[LeninGHOLA]

If people have $500 and want to spend it, they will. I keep my video card budget to $300 tops for a single card usually, but only because I don't generally have the cash for more expensive cards and usually don't have a need. In most cases, I like to keep my PC builds under $1500, but if I had a larger throwaway income, I'd definitely spend more.

The market for $500 cards is very small, but it doesn't make it useless. I see people regularly spending thousands on jet-skis or sports cars. PC gaming isn't the most expensive hobby

As far as the 5870 not being able to put out maximum settings at 5760x1200 and above, yea more current games will need more horsepower. Hopefully the 2gb 6 port cards hit the market soon, as I think that will remedy the situation a good amount.

 

[YeuEmMaiMai]

This was a press release not a product launch so I'm not really sure who's going OMG since there's nothing to go OMG over at this point.

As hardware enthusitats we like to know what's coming down the pike and speculate and that's all we're really doing at this point.

Until there's something to buy obvisouly I'm not buying anything like everyone else for now. I'll see what nVidia has in the GF100. Now if I have to wait until JUNE to get the high end model, hmm, might be an issue for me.

where is, exatly this product you speak of?