http://forums.vr-zone.com/showthread.php?t=378647Quote:
It looks like that RV740 and RV790 will be produced in 40nm and that the launch will be in March 2009.
Printable View
http://forums.vr-zone.com/showthread.php?t=378647Quote:
It looks like that RV740 and RV790 will be produced in 40nm and that the launch will be in March 2009.
So did RV870 get delayed then?
The question is, whether RV790 is just a shrink or shrink with expand.
RV770 @ 40nm would push it down to around 190mm². Now, expand it some 40% and we'll have a chip that's the same size as RV770 but with 52TMUs and 220SPs (1100ALUs).
ati has been making two generations of cards off one chip for awhile now. hd2900 and 3870 are based off the same chip. 4870 and this which will be the 4900 will be based off the same chip. they are just fine tuning it adding a few more sps a die shrink and most likely a core and ram frequency boost. rv870 is set one year after rv770 and will be released in june 2009. nothing was delayed.
Lets just call it a stop gap just like with nvidia's gtx260 55nm and 280 55nm.
One interesting thing is that RV770 is very close to being pad-limited. Since pads don't scale as much as logic in a process shrink, one can assume that RV790 will likely feature more SPs or something else to prevent the chip from becoming pad limited.
However, they could also fix this by removing VDD/GND pads since the chip won't need as much power.
Then again, all this silence and odd timing could be yet another cunning serving of Sandbagging à la AMD.
Hmm, this is quite logical, since it can match nVidia toe to toe with their supposed upcoming GT212 cards, RV870 will be prepared to duel with GT300 family cards later on.
Didn't they say that the was going to release the upgraded versions to Xmas first?
well it would make sense. they said they would release a new card once a year and every 6 months after it is released they would make an upgraded one. 4870 was released at the end of june which would make sense with it at the end of December. so maybe that got delayed or who knows. maybe they are just saying 6 months after 4870x2 is released. which would be around march.
Lets not call it a stopgap. Ati has been pretty consistent with process switches. I think it goes back as far as the R500 series. X1950pro was a 80nm shrink from 90nm(I think) around Q4/Q1, then there was R600 80nm Around Q2/Q3, then RV670 around Q4/Q1, Rv770 around Q2/Q3, and now we have RV790/740 around Q4/Q1. Granted, all of these have depended on process technology availability. R800 will come probably around Q2/Q3. No stop-gap, just predictable consistent release schedules.
RV770 on 40 nm. would be closer to 120 mm^2. We are talking about a die shrink from 55 nm --> 45 nm --> 40 nm. here. That would require a 128-bit memory bus and would thus be a good candidate for what RV740 is going to be. If RV790 has a 256-bit memory bus than it probably around 200 mm^2, or closer to what RV770 is right now if power consumption is around the same as the current RV770 and they include the sideport.
Helmore,
Dunno how you got that 120mm² - it's way too low.
It would require a shrink worth of two full nodes. 55nm to 40nm is one single node (two half nodes, actually). One full node shrinks to about 75% of original.
On 40 nm, i think ATi can easily hit around 1600 SP 80 TMU 24 ROP specced chip with die size budget beetween 260-300 mm^2, if they can keep their success in designing a pretty dense transistorwise GPU chip.
Looks like we're in for yet another year of complete ATI dominance.
Too much?
Yep, my earlier figures are wrong. My late night algebra is sometimes off...
But these should be about right; a full node optical shrink from 55nm to 40nm reduces die size to ~55% of original.
256mm² @ 55nm -> 140mm² @ 40nm
edit:
Need 40nm process kgate/mm² figures...
:\
Your specs are too low or your diesize is way too large.
40nm's main advantage is that increased density. AMD/ATi already did a great job with RV670/RV770 density and I'm sure they will do it again.
R600 was ~420mm2 with 720m.
RV670 had UVD ondie, DX10.1 logic, power controllers.
dominance?
GTX260 costs the same as 4780, is just as fast in real games people actually play instead of crap like call of juarez (minimum framerates often get ignored), overclocks better, uses less power, has better drivers (ati is always releasing hotfixes, still no game profiles), better looking anisotropic filtering, is quieter and cooler. ATI just now got a north american partner with lifetime warranty.
Obviously Ati does it with a much smaller gpu and nvidia's margins on the GTX are crap but from a consumer point of view the GTX series is just as good in price/performance and imo better when you consider reliability and support.
I guess you are going to bring up the 4870x2 but I don't consider multi-gpu viable.
got this from techpowerup:
It will be a current-generation GPU built on the next generation 40nm manufacturing technology. There is a lot of speculation surrounding the RV790's specifications, with some of the more plausible ones hinting it has two additional SIMD clusters (960 SPs) and a total 48 texture memory units (TMUs). Both the RV740 and RV790 are slated for March, there's also a little indication of AMD using the occasion of CeBIT for its announcements and product launches.
full article can be found here: http://www.techpowerup.com/
and the source of this information is here: http://www.hardware-infos.com (german)
All the chips have pads/wires on the backside that are soldered to the boards they are attached to. As the chips get smaller, you can put only so many of these wires on the backside of the chip, limiting input/output/power. So, for RV770 with a 256-bit memory controller let's say you need 400 pins. On 55nm the size of the chip gives you enough space that you can fit all these 400 pins but if you keep the logic(transistor) count of the chip the same but shrink to 40nm, you will have much less physical space for these pins and may only be able to attach 300 pins. So this requires either a smaller memory controller or making the chip larger, ie adding more logic, to get enough surface area for all the pins.
I hope that all made sense :shrug:
TSMC or TFC?
LET THE WARS BEGIN! :ROTF:
It took Nvidia how many months to get the GTX 260 to be what it should have been from the very beginning? :rolleyes: Even if it had been such a great card from the beginning, it is far from complete, or any, Nvidia dominance for 2008 - thus it makes your point useless! ;)
End it. Now.
I can't believe that guy went and edited his post to "further own" me. After I explained the history of my comment...
I mean that, I literally can't. Because it didn't happen.
Well I'll take your word for it, I was just going off the post times.
In any case, the idea of any company having "dominance" is a bit laughable in itself. The tech industry shifts so often and so dramatically... You can never really predict the future with 100% accuracy.
40nm Nvidia / ATI should both be nice, but after 55nm Nvidia, I'm going to give ATI a bit more benefit of the doubt.
I often think of something else to say and then add it into my post without checking if the person I quoted has responded, it's a bad habit I know...
Ati won the AA war in 2008 and so far is still beating Nvidia in it. Just watch how fast GTX 295 crumbles in front of 4870x2 when you go for 8xAA or higher in modern games.
Hmm when's Intel coming out with their thingy again?
???5870x2???:rolleyes:
Well no, but when it comes to quad SLI/CF you could assume that.
Although in brand new graphically demanding games, both sets of cards fall apart at that res/AA... So it's almost moot.
I suppose you could say the X2 has an advantage in slightly older games at super high res/AA or something.
The B3D dudes have talked about it a long time before, IIRC during/before the G92 launch period.
It's hard to compare to ATI at that period because the 3850/70 were bad performers with AA on, and only AA would tax memory usage extensively enough to limit you to <1600x1200.
At raw speed tests it's harder to determine, but there are still cases where the 256MB on the 8800GT runs out first compared to the 3850.
http://images.anandtech.com/reviews/...2/vsATI-qw.png
There are also cases where the 8800GT 512MB flat-out dies off at 1920*1200 8x AA while the 3870 still holds on, although not very playable. I think this one can be referred to in the Computerbase tests.
Yea, I remember seeing some tests where they measured video memory usage in the same games at the same settings with the same OS and processes, and ATI used less memory.
Dunno why that's hard to believe. It seems logical that Nvidia and ATI would have their own memory management systems, and I highly doubt they would both be
100% equally efficient.
Anyway, RV790 is Performance 55nm process, not 40nm.
RV740 is the sole 40nm chip for now. I'm not sure if 790 still actually exists or is just a sandbag pipedream.
Even if that were definitive proof of what you describe, it's been over a year since those products were introduced, and there have been many new driver schemes with new memory optimizations on both sides... This isn't an all-true fact of life written in stone though you describe it as though today you know definitively what's happening with memory usage.
All I'm saying is there are so many variables here, and since there are also architectural limitations contributing, it's oversimplifying to sit here and write about how one brand uses more memory than another.
1ghz so easy vmod probably not required
vmoded @ 1.1ghz easy
bring it amd! :clap:
RV790 is 55nm. I don't care what you guys believe, I know this for a fact. Take it or leave it. It'll most likely be pin compatible with RV770 PCBs, with an extra phase on the PWM.
Notice the missing soldering pads in the middle right of the current OEM PCB: http://www.pcper.com/images/reviews/581/4870_naked.jpg
That will be filled.
^
Zing!
It's just a process change, from the general perf process to a high speed one.
Hey guys so tried some googling now I'm confused, the RV790 & RV740 which one replaces what I'm guessing the 4870s new one is the rv740?
RV740 = 40nm, replacement for 4830 and most likely 4850 (but will probably be slightly slower than 4850 with AA turned on because it only has 8 ROPs)
128bit DDR5 = same memory bandwidth as 4850, cheaper PCB, $100 USD MSRP
RV790 = 55nm, replacement for 4870, same 256bit DDR5 but probably with faster chips, should be using the same 4870 PCB but with an extra PWM phase. Rumors about 960SP and 48TMU are probably true, so expect up to 20% boost in FPS for shader intensive games. Pricing unknown, probably still cheaper than the 55nm GTX260.
Perimeter? I know the memory controller itself is located at the edge of the chip, but the contact pads on the bottom of the chip don't necessarily have to be at the edge of the chip AFAIK. So to me it seems more of a matter of having enough surface area to be able to place enough contact pads on the bottom of the chip.
Also, if you are able to lower the chip's power draw, then you can also reduce the amount of VDD and GND pads, which already take up more than half of the used pads at the moment and thus use a smaller chip that way as well.
Maybe you should also add a couple of those CFX 'fingers', then that photochop will really start to fly across the interwebz :p:.
Well, that would definatly give them the ability to boost the clockspeed. Though, I really am curious as to the performance that you would get from a 4870 at say, 900mhz.
Since it would only be a slight process change, it also makes sense for it to be pin-compatible, and destined for current 4870 PCB's, with an extra phase.
What I really want to know though, is if the tweaked the design other than going for the high-performance proccess node variant.
I also wonder if AMD will have this as a hard launch, and how much of a price premium it will be over current 4870's.
Sorry:D whit the foto,s fun in the street
hopes of something news,from ATI.
Basically, chips have pads or pins, which connect them to the outside world. These pads are used for VDD/GND (power), as well as I/O (for example, connecting to GDDR and PCI-E, as well as JTAG, scan chain, etc test pins). These are either connected to the perimeter of the chip or on the entire bottom of the chip on a separate layer. Therefore, if we need 100 pads for example, then the minimum size of the chip is equal to 100*pad area. If your logic area is less than that, then you are pad limited.
Being pad limited is something very good for any complex chip design, since that means you can add "free" logic :)
From what the speculation is, a RV740 > 4830 > 9600GT, right? I'm thinking about a small upgrade .. : >
depending on the clock and if its gddr3 or 5 it will be some ware from the 4830 up to or slightly higher than the 4850. and it should retail relatively low. and sapphire told mne to rma my dead fan 3870 since they only make single slots facing the front of the case and that will be the same price point as the 3870 so they could rma and give me one. so thats what im looking forward to if its gddr5
Did Sapphire really tell you about the upcoming RV740? I never expected a company like that to do such things as it will have effect on their current sales as some might wait for the new product. They also basically confirmed that RV740's launch is not so far away and also where it should fit in the current line-up.
My guess would be that RV740 will be called the 4700 series of chips and RV790 will be sold as the 4900 series and the current 4800 will be scrapped. Just my prediction though. My prediction for is that RV740 will be a modified RV770 on a 40 nm. process, it will have 8 SIMD cores, 4 ROP clusters with each 4 ROPs and a 32-bit memory controller, no Sideport and it should be smaller than 120 mm^2. As for RV790, well I don't know, but I do think it will be made on a 40 nm. process and it will have more SIMDs than RV770.
Read my post: http://www.xtremesystems.org/forums/...3&postcount=56
You are wrong.
Although I have to say that you still can make a pin-compatible GPU even across generations.
6600GT/7600GT was one of them.
The "You are wrong" part was completely un-neccessary. :shakes:
Not that I don't believe you, but how do you know for ceartain? You still haven't given any proof, other than your own words. No need for a 'direct link' kinda thing, but any insight would make your words more credible. :up:
RV790 needs more units to be faster than RV770 if the retail RV790 will end up w/ the same clks the current RV790 ES chips have.
No way. :ROTF:
It's a $100+ product, but + means 20-30USD.
Think of the bright side! Even less power consumption, less heat, cheaper (vs 4850 MSRP), OCs more (if it has a 6-pin so you have enough power)
Most nVidia loyalists would be all over such a product already. :D
Afaik the core size was around RV730/HD4670, so the manufacturing cost should be somewhere near to RV730. Of course the 40 nm manufacturing process is more expensive as in per wafer + lower yields due to immature process. But that should be fixed in few months. Also I doubt that AMD necessarily needs to cut the price anyway, the biggest competitor for that card is AMD itself with HD4830/HD4850. ;)
Oh, completely forgot the GDDR5. Then I completely agree that the sub $100 is totally laughable. Silly me. :shrug:
Edit:http://vr-zone.com/articles/ati-rv74....html?doc=6334
I was sure I did not get the < $100 from my hat! Still quite doubtfull due to yields, more expensive process per wafer and GDDR5. Though, GDDR5 prices shouldn't be that high nowadays, and it makes the PCB simplier. So, the final bet: $110-$130 on launch, eventually falls sub $89. :)
I am expecting at least 3 different models out of RV740.
The highend with a six pin and will compete with HD4850, a slightly lower clocked one to be close to 4830, and then the very lowend possibly with some parts disabled to take the place of 4670. Heck they could even do another lowend part off of it if they had to. That really just depends on how RV830/810 is shaping up.
750 MHz. it seems:
http://www.hardware-infos.com/news.php?news=2658
I expect this card to rule the 85-110 US$ segment, with good profitability per unit.
Whats replacing the 4870x2? Are they going to bring it a big card? Or just update what they got?
A 40nm RV740 with a 6pin for additional power might be a fun, cheap card to abuse with volts.... imo :). Not expecting miracles being the first card on the new process, but there could be potential!
Exactly. The wafer price should drop rather fast and as the yields improve week by week, the margins grow rapidly.
Wafer size/die size,a.k.a. more dies per wafer increases. Also the usable wafer area increases since there are less redutant dies on the edges of the wafer.
should we except a return of the ringbus for RV740 or the RV870??
I don't see why they'd go back seeing as how successful ATI's current architecture is
Yeah, it might be one of the articles on rage3d, but they went with a distributed method which also allowed them up internal cache and other things, which is why performance on the RV770 went beyond just increasing units over the RV670 - as well as allowing so much to be fit within a small area
http://www.hardware-infos.com/news.php?news=2711
According to hardware-info RV790 has now been tested at 850Mhz on the core with a voltage bump.
750/900 MHz @ 1,24 Volt
750/975 MHz @ 1,24 Volt
850/975 MHz @ 1,30 Volt
english version: http://www.hardware-infos.com/news.p...2711&sprache=1
If they would launch new stuff in march, then they would show it at Cebit.
ZDNET (badly translated) says that AMD cancelled there presence at Cebit this year. So I doubt it they will launch anything soon.