im hoping they have a good 3D solution, maybe an upgraded eye infinity which supports 3d?
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
ATI 6***series 2010 or 2011 confused
- Thread starter 3t3P
- Start date
im hoping they have a good 3D solution, maybe an upgraded eye infinity which supports 3d?
It has poor uptake because it's poor.
It's only getting used because nVidia either pay or bully devs in to using it.
Why do you think Ageia, the original owners of PhysX had to sell up to nVidia? The interest was extremely poor.
I just wonder when Nvidia will realise that devs and the gaming public dont have any interest in it and finally bin it. Seems like a massive waste of money and resourses when they could be working on something else.
I thought the 6xxx series was going to have some of the new architecture in it, it was just going to be on a 40nm part rather than a 28/32nm part due, then hopefully GF will be up and running and all ready to produce the entire architecture in 28nm as a respin.
Makes no sense if they release a 6 series thats based on the older architecture. Im thinking they will release 6 series at 40nm with some new architecture mixed with old then respin it on 28nm when the process is ready hopefully q1 2011.
Makes no sense if they release a 6 series thats based on the older architecture. Im thinking they will release 6 series at 40nm with some new architecture mixed with old then respin it on 28nm when the process is ready hopefully q1 2011.
Soldato
- Joined
- 28 May 2007
- Posts
- 10,222
Thats basically what the guys above said. The new 6 series southern islands is going to be some of cypress aka 5 series mixed in with some northern islands.
Ah sorry missed that one, makes sense for them to do that tbh but will they release it as a 6 series or a respun 5 series? I see have 3 options:
1) Do it as a respin on the 5 series and then when they give all features of the architecture and 28nm that will be the 6 series.
2) Name the Hybrid chips as 6 series and then respin them with all the features on Northern Islands and 28nm.
3) Name the Hybrid chips as 6 series and then name the real Northern Island chips on 28nm and all the features as 7 series :S
Im guessing some of it will come down to the marketing team to what will happen but will be interesting to see what they decide to do as in theory they are ready to go with Northern Islands its just the process which is holding things up at the moment.
Associate
- Joined
- 14 Jan 2008
- Posts
- 20
ATI have the Stream Technology and its Stream SDK (which is really OpenCL), and this is the ATIs answer to CUDA. OpenCL is better that CUDA as it is a crossplatform standard and will work on both ATI and NVIDIA as well as a whole load of other computing devices. And it seems ATI is putting down a lot of resources into this technology at the moment, which is brilliant I think.
As for physics I don't know what's comming. But as with NVIDIA where the Physx sits ontop of CUDA, any physics for the ATI cards would be implemented in OpenCL. This would already be better than Physx as it is again crossplatform and as any serious game maker would likely want to reach as many customers as possible. And as OpenCL is an open standard anyone can create a physics library so it doesn't necessarily have to be ATI that does this.
Furthermore OpenCL is cool!
Yet again tsmc have screwed everyone so now we have SI which is a contingency to tide ati over till they can transfer to GF's 28nm and bring out NI in it's full glory. I guess it also makes good sense as it will allow them to try out some stuff rather then fully being reliant on all new stuff in one release. Time will tell but for now at least ati seem to be on a roll and adapting a lot better then nvidia.
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 33,188
Arm chips are the ultra low power process, its not the same as the 28nm bulk, you can't make something as big and power hungry on the low power process.
They'll probably be sending designs to tape out late Q4/early Q1 but we probably won't see anything till Q2 at the earliest, and seeing as Nvidia have no shot at getting anything till Q3 out of TSMC at 28nm, well, it would be easier to stick with AMD's now "normal" schedual of making a new chip on 40nm for Sept/Oct this year, and moving to a slightly more mature 28nm process for the next gen in Q3 2011. Though it will depend, its HIGHLY unlikely Nvidia can squeeze any more shaders into any kind of improved Fermi at 40nm, its too big, theres just no where they can fit more shaders into, okay they might do a refresh with cache removed and fit in one extra cluster, though its more likely they'll do a cache free version and use the extra space for yield improving fixes rather than more shaders.
Which would mean very little more than current Fermi performance till TSMC can get a new process out, and TSMC suck balls, if they say Q2, I doubt we'll see anything before Q3, and going for a ultra high end Fermi 2, with twice the transistors on a new process they've not used and having shown zero ability to forsee and overcome problems in terms of manufacturing I think the chances of them taping out and using the first or second spin are marginal.
If Nvidia are unlikely to have anything significantly faster than a 512sp Fermi, before very late next year, or early 2012, I'm not sure where the rush for AMD to bring out the 7xxx series cards the second 28nm is available comes from.
Their 6xxx series will likely be a good 15-25% faster than the fastest Fermi, cheaper and available.
We'll just have to see, Global Foundries, ex AMD manufacturing is rightly regarded as one of , if not THE best fabs in the world but they've not mass produced GPU's, they've not run 28nm bulk yet and problems could happen, TSMC, well, if the current trend continues, their 28nm could be delayed even further as new problems crop up.
I'd actually not be even remotely surprised to see a 6770 on 28nm early in 2011, but no 7xxx series on 28nm till late in the year.
I'd also be pretty much astonished to see any more than a 45-50% bump over the 5xxx series in performance, with a half gen jump on the same process. They just don't have the die space, transistor/power headroom you normally get to double the transistor count. IF they were moving to 32nm they could pretty much double the transistor count, shader count, while being about the same physical size as now with about the same power output. On 40nm, the same doubling of transistors and shaders would pretty much double the die size, go up a good 60-70% in power usage and be incredibly difficult to make at 40nm.
I wouldn't be surprised to see as little as a 30% bump in performance, rather than a 4890 with 20% higher clocks and performance, we'll likely get similar clocks and a slightly new architecture with 30-35% extra shaders, transistor and size over current 5xxx series cards.
Don't expect miracles, and really don't expect anything "new gen" from Nvidia on 40nm.
They next year could be very slow and boring.
They'll probably be sending designs to tape out late Q4/early Q1 but we probably won't see anything till Q2 at the earliest, and seeing as Nvidia have no shot at getting anything till Q3 out of TSMC at 28nm, well, it would be easier to stick with AMD's now "normal" schedual of making a new chip on 40nm for Sept/Oct this year, and moving to a slightly more mature 28nm process for the next gen in Q3 2011. Though it will depend, its HIGHLY unlikely Nvidia can squeeze any more shaders into any kind of improved Fermi at 40nm, its too big, theres just no where they can fit more shaders into, okay they might do a refresh with cache removed and fit in one extra cluster, though its more likely they'll do a cache free version and use the extra space for yield improving fixes rather than more shaders.
Which would mean very little more than current Fermi performance till TSMC can get a new process out, and TSMC suck balls, if they say Q2, I doubt we'll see anything before Q3, and going for a ultra high end Fermi 2, with twice the transistors on a new process they've not used and having shown zero ability to forsee and overcome problems in terms of manufacturing I think the chances of them taping out and using the first or second spin are marginal.
If Nvidia are unlikely to have anything significantly faster than a 512sp Fermi, before very late next year, or early 2012, I'm not sure where the rush for AMD to bring out the 7xxx series cards the second 28nm is available comes from.
Their 6xxx series will likely be a good 15-25% faster than the fastest Fermi, cheaper and available.
We'll just have to see, Global Foundries, ex AMD manufacturing is rightly regarded as one of , if not THE best fabs in the world but they've not mass produced GPU's, they've not run 28nm bulk yet and problems could happen, TSMC, well, if the current trend continues, their 28nm could be delayed even further as new problems crop up.
I'd actually not be even remotely surprised to see a 6770 on 28nm early in 2011, but no 7xxx series on 28nm till late in the year.
I'd also be pretty much astonished to see any more than a 45-50% bump over the 5xxx series in performance, with a half gen jump on the same process. They just don't have the die space, transistor/power headroom you normally get to double the transistor count. IF they were moving to 32nm they could pretty much double the transistor count, shader count, while being about the same physical size as now with about the same power output. On 40nm, the same doubling of transistors and shaders would pretty much double the die size, go up a good 60-70% in power usage and be incredibly difficult to make at 40nm.
I wouldn't be surprised to see as little as a 30% bump in performance, rather than a 4890 with 20% higher clocks and performance, we'll likely get similar clocks and a slightly new architecture with 30-35% extra shaders, transistor and size over current 5xxx series cards.
Don't expect miracles, and really don't expect anything "new gen" from Nvidia on 40nm.
They next year could be very slow and boring.