AMD To Launch Radeon HD4890 "RV790" In April

[SNiiPE_DoGG]

like I said programmable shaders = full dx11, or at least thats what I have been told....

[ownage]

Guys, what if this is really a "cover-up" for the ultimate 40nm refresh of the same chip generation, like the 2900--> 3870? I wouldnt be surprised if ATI has caught Nvidia off-guard at this. At 40nm, even 950 MHz seems possible. It would just really make sense to have more TMU's.

Debunked
OK, shameless translating continues.

RV790 is a product made from the 55nm GT process. The 55nm GT is the newer improved 55GC process.
40Nm would be very unlikely since ATI had lots of trouble with the RV740. Lots of leakage and clockspeed problems.
Most of those problems may be solved now, but its doubtful that ATI would have working samples by now, and fully working GPU's within a few weeks while they still have problems with RV740, which should be a less difficult design.

[Napoleonic]

40Nm would be very unlikely since ATI had lots of trouble with the RV740. Lots of leakage and clockspeed problems.

and they pushed the launch to april instead of may, don't sound that they have problem to me?

[LordEC911]

and they pushed the launch to april instead of may, don't sound that they have problem to me?

Well, since they are supposedly on the 2nd or 3rd revision, yes that would be a problem.

[seanx]

They could have adjusted the price accordingly.Also a cooler chip is what everyone wants,and what they need,I saw at least 100 pages of 4870 overheating threads,Which also hurt sales.

The pros simply have to outweigh the cons.
They could have made the RV770 'virtually' (As CJ described) bigger/larger so pin-counts and bonding pads wouldn't be a problem, it also would have result in a much cooler chip. The same chip, but just made bigger. The pros of a small economical die was probably more rewarding than a big 512bit chip.

well we could all possibly aggree then that the bigger die would've been better then , cooler and possibility for 512bit bandwidth ... yes?

This leads us to think that maybe Ati jumped the gun on its latest die shrink
simply to spite Nvidia ....? possible ? probable ? it seems so...

[Bo_Fox]

Heh.. all ATI needed to do was just turn up the fan speed by just a LITTLE bit. Most enthusiasts here are manually turning up the stock fan speed up to like 40 or 50 or as high as 60% for reasonably quiet operation at low temperatures. So, I've always read that those chips run quite cool at 50% fan speeds, and trust me, I've read countless threads..

[Bo_Fox]

Debunked
OK, shameless translating continues.

RV790 is a product made from the 55nm GT process. The 55nm GT is the newer improved 55GC process.
40Nm would be very unlikely since ATI had lots of trouble with the RV740. Lots of leakage and clockspeed problems.
Most of those problems may be solved now, but its doubtful that ATI would have working samples by now, and fully working GPU's within a few weeks while they still have problems with RV740, which should be a less difficult design.

Yeah, why does it take so long for ATI to make a chip that is just slightly larger than the RV740 on 40nm? I mean, just maybe 50% larger chip, but with a delay of 6 months?? Hopefully, ATI learned a lesson for not releasing the RV770 any sooner on 55nm (8 month gap after releasing the higher-clocked RV670 on 55nm). ATI must've been desperately waiting for GDDR5 memory after all, I guess, for more "attractive-ness" and cheaper 256-bit interface that requires less pins and slightly smaller die size "real-estate".

[Shintai]

Yeah, why does it take so long for ATI to make a chip that is just slightly larger than the RV740 on 40nm? I mean, just maybe 50% larger chip, but with a delay of 6 months?? Hopefully, ATI learned a lesson for not releasing the RV770 any sooner on 55nm (8 month gap after releasing the higher-clocked RV670 on 55nm). ATI must've been desperately waiting for GDDR5 memory after all, I guess, for more "attractive-ness" and cheaper 256-bit interface that requires less pins and slightly smaller die size "real-estate".

Because the 40nm aint as good as you think yet. I bet yield and quality from TSMC is still somewhat close to the trash on 40nm. They had problems so long and still struggles.

[v_rr]

Because the 40nm aint as good as you think yet. I bet yield and quality from TSMC is still somewhat close to the trash on 40nm. They had problems so long and still struggles.

ATI pushed back RV740 1 month sooner to April and release it to notebook and desktop ate same time, so you need a reality check.
If Nvidia is in deep trouble with 40nm is a hole different story.

[Shintai]

ATI pushed back RV740 1 month sooner to April and release it to notebook and desktop ate same time, so you need a reality check.
If Nvidia is in deep trouble with 40nm is a hole different story.

If 40nm was so good then RV790 would be using it.

[ownage]

It wouldn't surprise me if 40nm would be overkill for a slightly upgraded RV770.

RV790XT 850Mhz Engine, 975Mhz Memory (1GB GDDR5)
RV790Pro
RV740XT 700Mhz Engine, 800Mhz Memory (512MB GDDR5)
RV740XT 700Mhz Engine, 900Mhz Memory (1GB GDDR5)
RV740Pro 600Mhz Engine, 900Mhz Memory (1GB GDDR3)

Maybe on 40nm the RV790 could have it 900 or 950, but clockspeed alone probably isn't enough. On 40nm they would have add more muscle to the core. RV790 doesn't need to be much faster, and shouldn't because there has to be some room for the RV870 (40nm ) when its launched end this year (samples will be ready in Q2).

[MaDuLo]

ATI pushed back RV740 1 month sooner to April and release it to notebook and desktop ate same time, so you need a reality check.
If Nvidia is in deep trouble with 40nm is a hole different story.

don't quote shintai if you do i can read his replies

BTW, 4870 is designed for 55nm, right? so 40nm is only a half node away from it's orginal design. Would ik be possible to just make a 40nm dieshrink to cut costs? (like nvidia did with GTX280/260)
While on the other hand, a new 55nm chip (RV790) is created which is an optimized, better RV770?
just guessing though

[Kuroimaho]

I thought the same, if the 40nm would have good yields dumb shrinks would happen during this price war, and that we have yet to see such cards is the proof something isn't right at TSMC.

[largon]

4870 is designed for 55nm, right? so 40nm is only a half node away from it's orginal design.

55nm -> 40nm = full node shrink.
55nm -> 45nm = half node shrink.

[Shintai]

Not to confuse with 55 and 40nm is the actual halfnodes with optical shrinks of 65 and 45nm.

[MaDuLo]

55nm -> 40nm = full node shrink.
55nm -> 45nm = half node shrink.

makes no sense

65 is a full node
45/40 also is a full node

[largon]

"Full node" and "half node" are terms for shrinks, not for specific processes. 90nm, 65nm nor 45nm are nodes, not "full" nodes, just nodes. 80nm, 55nm and 40nm are also nodes, but they are also half node shrinks of the aforementioned nodes respectively.

SHRINKS are either full node shrinks or half node shrinks. Nodes are just... nodes.

65nm -> 45nm is a full node shrink because it skips 55nm.
65nm -> 55nm is a half node optical shrink.
55nm -> 45nm is a half node shrink, but not optical.
55nm -> 40nm is a full node shrink because it skips 45nm.

[STaRGaZeR]

What would you do without italics...

[Bo_Fox]

don't quote shintai if you do i can read his replies

BTW, 4870 is designed for 55nm, right? so 40nm is only a half node away from it's orginal design. Would ik be possible to just make a 40nm dieshrink to cut costs? (like nvidia did with GTX280/260)
While on the other hand, a new 55nm chip (RV790) is created which is an optimized, better RV770?
just guessing though

LOL, that coming from a guy who only has 26 posts here. That's mean anyways.. I dont see anything wrong with reading his posts.

Yeah, 40nm is a big step from 55nm. Let's draw out our calculators from our back pockets (as if we're cowboy nerds).. and calculate as quickly as possible:

55/40 = 137.5% difference
65/45 = 144% difference

The difference isnt as big, but it's close regardless. 55nm is already smaller than your Q6600, and 40nm is even smaller than the Corei7.

[LordEC911]

LOL, that coming from a guy who only has 26 posts here. That's mean anyways.. I dont see anything wrong with reading his posts.

Yeah, 40nm is a big step from 55nm. Let's draw out our calculators from our back pockets (as if we're cowboy nerds).. and calculate as quickly as possible:

55/40 = 137.5% difference
65/45 = 144% difference

The difference isnt as big, but it's close regardless. 55nm is already smaller than your Q6600, and 40nm is even smaller than the Corei7.

Math is slightly wrong... Plus why are you doing increases? We care about decreases.

With your math-
40/55=72.7%
45/65=69.2%

but in reality-
40^2/55^2= 52.9%
45^2/65^2=47.9%

[MrHydes]

i belive they are having great issues with 40nm process.

and yes, if they went to 40nm, it's the first time in history we'll see

GPU's more advanced technology than cpu's wich is amazing.

they might catch up Q4 with westmere 32

[Bo_Fox]

Math is slightly wrong... Plus why are you doing increases? We care about decreases.

With your math-
40/55=72.7%
45/65=69.2%

but in reality-
40^2/55^2= 52.9%
45^2/65^2=47.9%

Ahh, you win the nerd shoot-out contest!

[Budwise]

From the Inq

THERE HAVE BEEN a lot of rumours flying around about the upcoming ATI 4890 board, and few if any of them are true. The fevered speculation of a few has gotten almost to the point of silly, so lets set the record straight.


The 4890 is not a new ASIC, it may or may not be a new stepping of the R770, but it is not a new part. It is not 40nm either. On top of that, there is no R790 in the works at all. The new board is called the 4890, and it is simply a much higher clocked 4870.


It will be at least 20 per cent faster than the 4870, a real kick in the pants, but will take a bit more power. The added power will require five-phase voltage regulation, so there will be a new board layout, quite likely lining up with a new downcosted 4870 board as well.


End result, faster card. No new chip, no more shaders, nothing huge, just better. Anyone saying anything else is full of it. µ

[SNiiPE_DoGG]

From the Inq

hrrrrrrrm how can we know charlie knows anything more...

[eleeter]

From the Inq

But that article is written by Charlie, so it must be absolute crap right?