PDA

View Full Version : NVIDIA GF100 Previews


Pages : 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 [22] 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52

shadow001
03-12-10, 01:01 PM
And they essentials did the same thing R600/670 > R770 and it didn't give them any real performance advantage. It sure looked good in the synthetic benches, but did **** in real world compared to Nvidia. Here we are again with ATI doing the same as before and as before chip to chip(ie single gpu vs single gpu) it isn't working to their advantage again in actual gaming performance. Some thing to concider concerning Fermi. The 8 or 12 ROPs(depending on the card) over 32, aren't really being used for texturing but rather for AA performance. If AA isn't being used, they essentially sit idle. As Razor1 suggested, go read up on the white papers for Fermi.

There is also the fat each generation ATI has had 3x the shaders, but never had a shader based performance lead except for the synthetic based benches. efficentcy is a bitch aint it?

yes it doesn't improve on the raw numbers but its just like ATi's 8x AA performance edge last gen, although the gt200's had higher bandwidth, they weren't able to use as efficient as ATi's counterparts. For the most part ever single part of Fermi has been rebuilt, this isn't a simple extrapolate numbers for last gen situation, until we start seeing some synthetic tests, or real world benches to assume performance based on the chip stats, is pretty much pointless.

That really doesn't matter much, most guys that know about the numbers, know to research reviews to make their purchases, most people just care about end performance anyhow. If it was a numbers game, AMD should be selling head over heals since the r600 with the sheer number of shader units on their GPU's.




Tessellation is only part of Fermi's performance advantage, the others are still unknown, ATi has pushed hard on tessellation too, which seems to be a mistake now since Fermi will have an advantage here.



So far what I have seen nV's focus on their GPU's tend to be more balanced and focused on today's games. Lets have a look at the r520, it is a good chip but it was unbalanced with great AA performance but lacked shader performance against the 7800, then they went to the r580, huge shader performance but we didn't see any games that needed that till next gen cards came out, and that performance was not really seen either. This is where nV tends to have a more consumer centric philosophy when designing their GPU's, ATi at the time tried to market their cards as more advanced and capable of doing more, and they were the truth, but at the end reviews weren't able to show the extent of how much better they were in future applications, since the games out at the time couldn't take advantage of the r580 shader core.



Synthetics push the cards on certain portions, that's the point of synthetics. I'm not talking about game engines like unigine or 3dmark, I'm talking about individual tests, that stress the shader units depending on shader length, texture look ups, vertex texture fetches things like that.




Once you add AA in there, now you are polluting the test, you are now stressing 2 parts of the card (well actually more since you doing alot more Vantage, its acutally a game engine simulation), how are you going to determine what part is being bottlenecked without a starting point? You have a identify what you are looking for and then proceed from that point. You can't just take this benchmark and this is what it is doing without a reference point. More interested in individual shader tests, fillrate tests, etc. Use those as a starting point then go from there.



You have to break the performance to see where the bottlenecks lie. There are many synthetic tests out there, ixbtlabs and computerbase.de are review sites that uses their own synthetic tests, but there are others too that use like shadermark.


Well here's something strange,and it concerns the supposedly strong points of Fermi here(Tesselation),looks at the heaven demo results on that tech report article right here,which goes back to january 20th,which the chart was supplied by Nvidia themselves,and showing the performance difference between Cypress and Fermi.


http://www.techreport.com/r.x/gf100-arch/unigine-slide.png



Now note the latest image in the video posted by an Nvidia marketing director also running the heaven demo just last week,comparing the HD5870 again with the GTX480:


http://i765.photobucket.com/albums/xx298/Superfly101_02/Capture5.jpg



So what happened here,did the HD5870 gain performance on tesselation?....Did fermi not hit the clocks it was supposed to,because it seems to me the gap is quite a bit smaller overall.....Just something to think about when all 3 of you said it's not about the raw numbers,it's about how efficient it is.


And this is the strong point for Fermi,tesselation.

Razor1
03-12-10, 01:06 PM
The problem is the other parts of the benchmark what are the bottleneck's? it could be pixel fillrates, which both cards should be close on that, I don't think its bandwidth since there is no AA.

This is why we won't know what is going on till we actually see other tests, synthetics, ect.

Its impossible to base a conclusion on something we have incomplete data on. There are parts of that demo loop where the 5870 gets close even with tessellation (the last two spikes) and parts like at the end where there isn't much tessellation where Fermi pulls ahead.

When ever playing a full game engine like this, there are going to parts of the scene that might favor one card and parts that favor another card. Even within the same frame bottlenecks can shift.

shadow001
03-12-10, 01:11 PM
The problem is the other parts of the benchmark what are the bottleneck's? it could be pixel fillrates, which both cards should be close on that, I don't think its bandwidth since there is no AA.

This is why we won't know what is going on till we actually see other tests, synthetics, ect.

Its impossible to base a conclusion on something we have incomplete data on. There are parts of that demo loop where the 5870 gets close even with tessellation (the last two spikes) and parts like at the end where there isn't much tessellation where Fermi pulls ahead.

When ever playing a full game engine like this, there are going to parts of the scene that might favor one card and parts that favor another card. Even within the same frame bottlenecks can shift.


At no point in the first picture was the HD5870 even close to fermi,and we see a much tighter race in the entire benchmark run second picture,and this is the GTX480 card,not even the GTX470 one.


The only difference between both pictures is that the top one specifically says it's a 60 sec snapshot within the Heaven benchmark,while the entire benchmark run is much longer than 60 secs,so it seems to me they isolated a specific part of the benchmark to show tesselation in the best case scenario only,not the overall picture,because heaven forbid(pun intended),to show the entire run and seeing ATI's card quite close in many parts.


It's called very selective marketing material.

Razor1
03-12-10, 01:17 PM
Both sides have had missed beforehand. Suggesting that Nvidia only had 2 mis-steps is a little disingenious when you are bringing up every potential issue for AMD above.

THE biggest issue for AMD has been the 2900 launch/performance/availability. Beyond that, they have been up and down a bit in terms of schedule and delivery but performance has been decent. The 2900 and related family launches were terrible performers given the scope of the hardware and this was not really rectified till the 4800 series debuted.

Its always a comparative analysis between the competing companies, nV has only screwed up twice when competing against ATi. We can bring in nV1 but ATi was no where at the time.

The thing about AMD's model for this and the previous generation is they are looking at scalable solutions that can be increased/decreased in complexity/performance, relatively easily. Nvidia's previous and upcoming model does not allow this flexibility, but they still typically deliver top to bottom market coverage.

What makes you so sure of that, nV's GPU models have been very flexible, if anything more flexible in certain situations then ATi's solutions when scaling down. Scaling down doesn't mean using the same chip across the midrange and top end.


This doesn't mean one is better than the other but AMD's appears to be working better over the past 2 years.

True


Most of the items have already been discussed before but, I think your discussion of nomenclature is telling because over the past 2 or 3 years, Nvidia has been terrible, to the point of being blatantly misleading in their naming convention. Just look at all the retail names the G92 chip has received.

Branding doesn't go into the individual product names, its a holistic approach of all products of a company, in this case, everything from hardware, software, initiatives, design of events etc, etc, the whole shebang. nV hasn't gotten to the point of like Intel with Intel inside or Nike with its swoosh, those have become iconic. But you get the picture of what I'm talking about now?

Razor1
03-12-10, 01:21 PM
At no point in the first picture was the HD5870 even close to fermi,and we see a much tighter race in the entire benchmark run second picture,and this is the GTX480 card,not even the GTX470 one.

The only difference between both pictures is that the top one specifically says it's a 60 sec snapshot within the Heaven benchmark,while the entire benchmark run is much longer than 60 secs,so it seems to me they isolated a specific part of the benchmark to show tesselation in the best case scenario only,not the overall picture,because heaven forbid(pun intended),to show the entire run and seeing ATI's card quite close.

the first picture was a 60 second snap shot of the full bench. Mind you if you look closely you can see where that segment is and also you will see from that 60 sec seg to the full bench you can see that the performance has increased in the full segment.

So what does that matter, they used it as a piece to show how well tessellation performs on the Fermi architecture. That was the only point they were trying to make with those presentations. Did you see them talk or present anything else? Also keep in mind the Unigine Dx 11 path was made on ATi hardware. So without really knowing what is going on you can't say this one benchmark is going to show us whats going to happen in other games.

shadow001
03-12-10, 01:33 PM
the first picture was a 60 second snap shot of the full bench. Mind you if you look closely you can see where that segment is and also you will see from that 60 sec seg to the full bench you can see that the performance has increased in the full segment.

So what does that matter, they used it as a piece to show how well tessellation performs on the Fermi architecture. That was the only point they were trying to make with those presentations. Did you see them talk or present anything else? Also keep in mind the Unigine Dx 11 path was made on ATi hardware. So without really knowing what is going on you can't say this one benchmark is going to show us whats going to happen in other games.

Well,since you've stated that real world situations is what really matters in the end,i though it would be interesting to show that the only situation where Fermi could ever be 60% faster than Cypress,even in tesselation which is it's strong point,is in a specific part of a benchmark that lasts 4x longer(about 260 secs in fact) than the 60 sec graph shown in the first picture.


It's called misleading information at the very least,order to show a product at it's best,in a synthetic benchmark....Some could call it a lie even,and and that isn't a very real world scenario now is it?

Razor1
03-12-10, 01:40 PM
Well,since you've stated that real world situations is what really matters in the end,i though it would be interesting to show that the only situation where Fermi could ever be 60% faster than Cypress,even in tesselation which is it's strong point,is in a specific part of a benchmark that lasts 4x longer(about 260 secs in fact) than the 60 sec graph shown in the first picture.


It's called misleading information at the very least,order to show a product at it's best,in a synthetic benchmark....Some could call it a lie even,and and that isn't a very real world scenario now is it?

how is it a lie when its true, they even state its a 60 sec snap shot of certain part of the demo, dragon and cobblestone walkway? Pretty specific on what they stated. Really grabbing at straws.

shadow001
03-12-10, 01:48 PM
how is it a lie when its true, they even state its a 60 sec snap shot of certain part of the demo, dragon and cobblestone walkway? Pretty specific on what they stated. Really grabbing at straws.


We don't play the 60 sec parts of a game where the highest FPS is displayed do we?....We play the entire game,it's really that simple....Real world scenarios.


In any case,we'll see in 2 weeks time when reviews are released.

Razor1
03-12-10, 01:52 PM
We don't play the 60 sec parts of a game where the highest FPS is displayed do we?....We play the entire game,it's really that simple....Real world scenarios.


In any case,we'll see in 2 weeks time when reviews are released.

You are taking things out of context, that slide was shown in the editor's day deep dive event when they were talking about tessellation, there was no mention that Fermi will perform in other situations when they showed that slide.

shadow001
03-12-10, 02:03 PM
You are taking things out of context, that slide was shown in the editor's day deep dive event when they were talking about tessellation, there was no mention that Fermi will perform in other situations when they showed that slide.



Nvidia's own CEO stated a few months ago that Fermi will be up to 60% faster,and i can dig up the interview where he stated that btw...


Looks like we found the only situation where it actually happens,in a portion of a benchmark,lasting about 1/4 of the entire test,and specifically stress testing the strongest point of fermi.


That's like a developer specifically writing a piece of software where insane amounts of shading power are consumed and where Cypress would shine the best,but actual games are no where near using using those crazy amounts of shading anyhow,so it's a theoretical test at best,and more importantly,not representative of actual shipping games.


And if you go with the argument...Well in the future it might be,that should also include the simple fact that both ATI and Nvidia wil have much faster hardware by then,and no one will care about Fermi and Cypress anyhow,as they're old technology anyhow.

Razor1
03-12-10, 02:13 PM
Nvidia's own CEO stated a few months ago that Fermi will be up to 60% faster,and i can dig up the interview where he stated that btw...


Please do because I have not heard anything like that, Jensen rarely talks about performance numbers of unreleased products, if at all.

Looks like we found the only situation where it actually happens,in a portion of a benchmark,lasting about 1/4 of the entire test,and specifically stress testing the strongest point of fermi.


You don't know that yet, you seem to take things very black an white but guess what there is a lot of grey in the middle that we haven't seen yet :)


That's like a developer specifically writing a piece of software where insane amounts of shading power are consumed and where Cypress would shine the best,but actual games are no where near using using those crazy amounts of shading anyhow,so it's a theoretical test at best,and more importantly,not representative of actual shipping games.

Arg? where did that come from, dude I can really go into alot more depth if you like, but this is getting to be a very stupid conversation from one side. Sorry but I don't care to talk to someone that seems to be so focused on one thing and one thing only. You are the one here stating that Unigine is the end all be all that nV used to show Fermi's performance, guess what its an unreleased product, lets see what they show when it is released.

And if you go with the argument...Well in the future it might be,that should also include the simple fact that both ATI and Nvidia wil have much faster hardware by then,and no one will care about Fermi and Cypress anyhow,as they're old technology anyhow.

I'm not saying that at all, I'm saying you seem to have Fermi up your butt :) (take it as a joke) and can't seem to understand what we have seen so far, might not be what actually comes out! Do you remember the g80, many people thought it was a 2x the g70 with some modifications, until oh 3 weeks before launch well guess what happened? It was a shocker, so far outside of very specifically limited information, we don't know anything about Fermi's actual performance. You are basing your assumption that Fermi is only marginally faster then the HD5870 on only one benchmark, that nV has released to show tessellation performance. And you are talking it as black and white.

Iruwen
03-12-10, 02:20 PM
http://i765.photobucket.com/albums/xx298/Superfly101_02/Capture5.jpg

Nvidia's own CEO stated a few months ago th

at Fermi will be up to 60% faster,and i can dig up the interview where he stated that btw...

Obviously this was understatement, since in that screenshot I can see clearly that Fermi is almost up to 100% faster.

XMAN52373
03-12-10, 02:21 PM
Well here's something strange,and it concerns the supposedly strong points of Fermi here(Tesselation),looks at the heaven demo results on that tech report article right here,which goes back to january 20th,which the chart was supplied by Nvidia themselves,and showing the performance difference between Cypress and Fermi.


http://www.techreport.com/r.x/gf100-arch/unigine-slide.png



Now note the latest image in the video posted by an Nvidia marketing director also running the heaven demo just last week,comparing the HD5870 again with the GTX480:


http://i765.photobucket.com/albums/xx298/Superfly101_02/Capture5.jpg



So what happened here,did the HD5870 gain performance on tesselation?....Did fermi not hit the clocks it was supposed to,because it seems to me the gap is quite a bit smaller overall.....Just something to think about when all 3 of you said it's not about the raw numbers,it's about how efficient it is.


And this is the strong point for Fermi,tesselation.

You are either dense or dumb or both. If you compare both pics you posted, anyone with half an active brain cell could tell what segment of the second picture the first picture comes from. Pretty much right out of the middle which is where tesselation in that demo is its heaviest.

Iruwen
03-12-10, 02:26 PM
Btw., I don't think that triple monitor setups have a special relevance in the gaming market apart from some flight sim enthusiasts until frameless displays become common. Even our developers only use two monitors, although they have Eyefinity cards. The relatively low number and high cost of displays with display port also probably is a problem.

shadow001
03-12-10, 02:48 PM
Please do because I have not heard anything like that, Jensen rarely talks about performance numbers of unreleased products, if at all.


This is a fun comment when he did the initial presentation....Setting yourself up for eating your words much?


Fermi to me feels like this: Start with an SR-71 Blackbird for speed, add two parts Prius for fuel economy, add 100 gallons of Moonshine for kick, three parts light saber for lethality, DNA from Megan Fox for sex appeal, fold space like in Dune so you can get close to a dark star with enough gravity to smash it all into a finger nail sized chip, hit it with millions of lightning bolts lines of driver code written by GPU-Jedi-Masters, and then bake until ready.[


http://blogs.nvidia.com/gtc/fermi/


Though i will look for the exact 60% comment later today,got stuff to do.




You don't know that yet, you seem to take things very black an white but guess what there is a lot of grey in the middle that we haven't seen yet :)


Final reviews will reveal all,and if in practical terms on real world games,it's barely faster than the HD5870,doesn't beat the HD5970,uses more power and is alot more expensive,prepare for a world of bad press from the major hardware review sites.


So yes,i am very black and white in that sense i guess.


Arg? where did that come from, dude I can really go into alot more depth if you like, but this is getting to be a very stupid conversation from one side. Sorry but I don't care to talk to someone that seems to be so focused on one thing and one thing only. You are the one here stating that Unigine is the end all be all that nV used to show Fermi's performance, guess what its an unreleased product, lets see what they show when it is released.


I never stated that,but Nvidia seems to lean in that benchmark quite heavily in a controled manner that benefits Fermi,that much is obvious,and is not representative of real world scenarios with actual shipping games....That was my point.


I'm not saying that at all, I'm saying you seem to have Fermi up your butt :) (take it as a joke) and can't seem to understand what we have seen so far, might not be what actually comes out! Do you remember the g80, many people thought it was a 2x the g70 with some modifications, until oh 3 weeks before launch well guess what happened? It was a shocker, so far outside of very specifically limited information, we don't know anything about Fermi's actual performance. You are basing your assumption that Fermi is only marginally faster then the HD5870 on only one benchmark, that nV has released to show tessellation performance. And you are talking it as black and white.


We'll see when reviews are released in 2 weeks i guess.

shadow001
03-12-10, 02:50 PM
You are either dense or dumb or both. If you compare both pics you posted, anyone with half an active brain cell could tell what segment of the second picture the first picture comes from. Pretty much right out of the middle which is where tesselation in that demo is its heaviest.


Middle section is twice as long in terms of time frame,more like closer to 120 secs long in the second picture,rather than the 60 secs shown in the first picture....;)


There's that annoying peak in the middle where both cypress and fermi are almost even.

shadow001
03-12-10, 02:57 PM
Btw., I don't think that triple monitor setups have a special relevance in the gaming market apart from some flight sim enthusiasts until frameless displays become common. Even our developers only use two monitors, although they have Eyefinity cards. The relatively low number and high cost of displays with display port also probably is a problem.


No need for a display port LCD,as there's active mini display to DVI converters from sapphire that cost about 70~80$ and hook that up to a standard LCD with a regular DVI port.


I'm having a blast with mine in battlefield bad company 2,especially when flying the helicopter and viewing a very large portion of the map without having to even to move the helicopter much.....That is,when the EA server is actually working.

Razor1
03-12-10, 03:22 PM
This is a fun comment when he did the initial presentation....Setting yourself up for eating your words much?

http://blogs.nvidia.com/gtc/fermi/

Though i will look for the exact 60% comment later today,got stuff to do.

I'm guessing you will have to look very hard ;)

Final reviews will reveal all,and if in practical terms on real world games,it's barely faster than the HD5870,doesn't beat the HD5970,uses more power and is alot more expensive,prepare for a world of bad press from the major hardware review sites.

Outside of Charlie (who started off with piss poor tessellation performance, to broken down performance due to bad chip design but tessellation will work fine, from loosing to the HD5870 badly, to now it will perform around the HD5870), no else has really said that, some people speculated based on clocks, but again, we all know clocks don't really mean much because that's based on the design of the chip


So yes,i am very black and white in that sense i guess.

Well that's why I say, it "can" be very different then you think

I never stated that,but Nvidia seems to lean in that benchmark quite heavily in a controled manner that benefits Fermi,that much is obvious,and is not representative of real world scenarios with actual shipping games....That was my point.

Oh yes you did and have been saying that in almost every single one of your posts, now you are changing you statements with the last two posts and putting nV marketing in there, if you stated it that way in the beginning, I would have responded, its marketing that's their job, of course they will!

Iruwen
03-12-10, 03:29 PM
http://www.semiaccurate.com/2010/03/12/semiaccurate-wrong-about-nvidia-480gtx-power-use/

Well... although he was actually right, he still is a moron :p

Now it all depends on price and real world performance and power consumption (considering Nvidia usually provides the actual maximum TDP and ATI the average TDP iirc, which makes a ~20W difference).

/e: looks like parts of the article may actually not be true at all, again... sigh.

/e²: I wasn't correct about the TDP, actual consumption is much higher with Furmark for Nvidia too.

scubes
03-12-10, 03:42 PM
guys really this ****ty demo doesnt i my eyes doesnt prove a thing to me the thing i want to see the most is GAMES if this card doesnt run crysis @ 60 fps at 1920 1200 they can shove it were the sun doesnt shine im not interested in this tech demo its game,s i wanna see. for instance can you play this crappy DEMO no can you play games YES.ok so fermi might be good at tessie but wen its all said and done i just want my games to run at 60 fps with AA CRACKED UP AND AF CRANKED UP ALSO..

Razor1
03-12-10, 03:44 PM
Well... although he was actually right, he still is a moron :p


ah that sentence reminded me of dirty rotten scoundrels "he is a moron" :D

Iruwen
03-12-10, 03:54 PM
ah that sentence reminded me of dirty rotten scoundrels "he is a moron" :D

Fun fact of the week: the movie's german title is "Zwei hinreißend verdorbene Schurken" (would probably translate back into "Two ravishingly profligate Scoundrels" or something like that). German movie titles almost always suck :|

Razor1
03-12-10, 04:02 PM
LOL

shadow001
03-12-10, 04:50 PM
My bad,i though it was Nvidia's CEO who said it,but this particular article only states the information comes from Nvidia,and not naming anyone in particular,and with regards to real world gaming performance advantage that Fermi should have over Cypress


Enough with the technical gobbledygook. What will cards based on the GF100 perform like in games? Unfortunately, we don't really know. Nvidia's early benchmarks has it performing up to twice as fast as the Radeon HD 5870 (ATI's fastest single-chip DX11 graphics card) in some tests. Those are usually directed geometry-heavy benchmarks, though. In real games, it looks like performance will be anywhere from 20-50% faster, depending on the game and settings. Nvidia promises dramatically better performance with 8x anti-aliasing modes this time around, and a much lower performance hit over the 4x MSAA modes (which would put them nicely in line with ATI's latest GPUs). A new 32x coverage sampling anti-aliasing mode could be the new high quality mark, and might be fast enough to be truly useable, but we'll have to wait for our testing to bear that out.


20 to 50% faster in real games,depending on settings is it?....It'll be interesting to find out once reviews come out.


http://www.techworld.com.au/article/332893/nvidia_releases_details_gf100_chip


Oh and razor1,i see that you removed the earlier quote i made when Nvidias CEO made the intro for Fermi,you know,this dumb thing:


Fermi to me feels like this: Start with an SR-71 Blackbird for speed, add two parts Prius for fuel economy, add 100 gallons of Moonshine for kick, three parts light saber for lethality, DNA from Megan Fox for sex appeal, fold space like in Dune so you can get close to a dark star with enough gravity to smash it all into a finger nail sized chip, hit it with millions of lightning bolts lines of driver code written by GPU-Jedi-Masters, and then bake until ready.


That claim is like shooting yourself in the foot territory if fermi doesn't live up to it's performance claims,you know that right?....Even more so when what he was holding in his hand was a mockup,and not the real thing,you know?....That statement was made 6 months ago,so enough with the PR bull**** and spill the real performance numbers,as all the talk and no action is getting annoying to say the least.

Razor1
03-12-10, 05:12 PM
My bad,i though it was Nvidia's CEO who said it,but this particular article only states the information comes from Nvidia,and not naming anyone in particular,and with regards to real world gaming performance advantage that Fermi should have over Cypress

20 to 50% faster in real games,depending on settings is it?....It'll be interesting to find out once reviews come out.


Its a guess by the author, so I don't know how we can take that seriously since at the time, cards weren't sent out well outside of developers :)


Oh and razor1,i see that you removed the earlier quote i made when Nvidias CEO made the intro for Fermi,you know,this dumb thing:

That claim is like shooting yourself in the foot territory if fermi doesn't live up to it's performance claims,you know that right?....Even more so when what he was holding in his hand was a mockup,and not the real thing,you know?....That statement was made 6 months ago,so enough with the PR bull**** and spill the real performance numbers,as all the talk and no action is getting annoying to say the least.


I do expect stuff like that from the CEO of a company that is making the product, why not?

But seriously though, can't hinge on every word, how bout this from AMD and the r600 being the fastest card they have every produced, so many people twisted that around and stated its going to be the fastest on the market period.


“The R600 will be [absolutely] the fastest DirectX 9 chip that we had ever built,” said Richard Huddy, the head of ATI Technologies’ software developers relations department, at a press conference in London, UK.

http://www.xbitlabs.com/news/video/display/20060525104243.html

in one sentence they downplayed Dx10 and gave us no real answers!

That's PR for ya.....

Both companies do it, I don't really pay attention to those things. But if you want to, nV has been saying Fermi is going to the fastest GPU, now they might be talking about Dx11 only situations who knows.