AMD: Sea Islands R1100 (8*** series) Speculation/ Rumour Thread

It's capable, but supposedly more modern games make very little use of it. This isn't because of the API per se, I just mentioned that as a reference for when it was used much more.
Can you elaborate on that? The shaders have become more complex, but I don't see a major decline in the portion of special functions used in them.
 
If we were to ignore Tahiti, which sacrifices some performance/mm2 for extra GPGPU functionality and for being the first product on TSMC's 28nm process:

Pitcairn is 212mm2, not much higher than RV670's 192mm2.

Cape Verde is 123mm2, slightly lower than RV635's 132mm2.

And TSMC's 28nm process is now more mature and presumably cheaper than it was 12 months ago.
One may add Mars, which packs 6 CUs and a 128 Bit memory interface in an amazing 76mm^2. That looks already like quite an improvement compared to CapeVerde (okay, AMD has cut the VCE block, but I doubt it measures 20mm^2 so it can't explain the small size on its own).
 
That was a fairly special case in that the previous generation was a simple shrink and fix - in regards to cost and power efficiency - of the generation it succeeded. So performance remained stagnate, then, they used the shrink to once again increase the die size while at the same time further increasing efficiency.
There could be potential for a considerable improvement in density/efficiency with gen 2 of GCN. RV770 was basically gen 2 of VLIW5 and a big part of the improvements was much more performance per die area. You can also see this in RV670 vs RV730 which are pretty similar in performance like RV670 and R600.
 
RV770 and onward was also interesting in that there was an improvement in the tools over what was used on the chips that preceded it.
The density they had hoped to have for R600 was not realized until later.

An outsized improvement like that is uncommon, and we'd probably have more signs of density problems with current 28nm chips, which we've really not seen.
 
With the recent debacles at AMD, I wouldn't expect the morale to be any good and their best folks are probably stampeding for the nearest exit door. I wouldn't expect anything like RV770 in the near future.
 
to get this discussion booming again, im going to remind everyone of the awesomeness that was rv770.
RV770 wasn't that awesome. It was the R600, which was completely awful. Broken ROPs causing slow MSAA, ultra-wide bus, which took a lot of silicon, but couldn't be utilized because of the broken ROPs. Not to mention the leaky 80HS process, which caused horrible power consumption and lower than planned clock speed. Current situation is significantly different. Tahiti isn't broken, Pitcairn offers similar efficiency to GK104 (plus some advantages) - there isn't much room for improvements.
 
With the recent debacles at AMD, I wouldn't expect the morale to be any good and their best folks are probably stampeding for the nearest exit door. I wouldn't expect anything like RV770 in the near future.

Right, because the GPU's being launched in 2013 and 2014 will be designed after Project WIN began.

Oh.
 
Right, because the GPU's being launched in 2013 and 2014 will be designed after Project WIN began.

Oh.
This is rather inane. There is plenty of time for "environmental factors" to affect the performance of a product. How long have they had silicon back? Quite some time. It is certainly possible that layoffs took place before or during the post-tape out phase.

I don't believe that "morale" is a discussion-worthy topic in the first place, but if it did have a substantial effect on the final product(s), it is incredibly likely that HD 8000 would fall within the time frame for "morale" to have that potential.

I.e., I hold that the effects of "morale" on how a product performs has no place in serious discourse, as there is next to zero potential for scientifically backing any claims. Falsifiability is a critical component of a good hypothesis, and engineering "morale" claims are completely unfalsifiable. You'd have to literally conduct ridiculous psychological experiments to even paint such a correlation... however, if there were such a thing, there is enough of a time frame, post-tape out, for things to go wrong.
 
With the recent debacles at AMD, I wouldn't expect the morale to be any good and their best folks are probably stampeding for the nearest exit door. I wouldn't expect anything like RV770 in the near future.

Not sure where their running too, the unemployment line ? None of the tech giants are doing well, even Intel is taking a pounding. Mobility tech is where it is at, and none of the desktop players are 'players'..
 
Days are long gone when you walk out the door of a great job to another, no matter how good you are. For everyone that is considered elite there is two more waiting in line.
 
Days are long gone when you walk out the door of a great job to another, no matter how good you are. For everyone that is considered elite there is two more waiting in line.

My group currently has no budget, but if there was someone available who was truly "elite", we'd hire them.
Elite people rarely change jobs, and even the best teams have too few of them.
 
Days are long gone when you walk out the door of a great job to another, no matter how good you are. For everyone that is considered elite there is two more waiting in line.

they are not going to walk along the road waiting someone want them.. If you look the history, most of them have just gone to an other "group"..

Now, reading all about this " elites debacle" from AMD... i think we offtly forget the situation of AMD... I dont criticze them as many of them have got the talent and have do great things.
But at the same time, if they was on the elites of the A64 days, they was too there during thoses last 6-7years... its nice to read a good article on how this guys was so good on the A64 days..
But what have they done thoses last 6 years working for AMD ?

I have really mixed feeling about all of this.. is it a good thing they leave and let the place, or is a bad thing ?

Now, i hope by changing their function and job they will be able to use all the talent they are capable of.
 
Even if we assume "Group X" did A64 and in fact stayed together since, too, it doesn't mean they could pull out another "miracle chip" just like that - elite or not, you can't do miracles all the time.

And remember that big part of A64's success and halo wasn't because it was so good, but because Netburst was so bad
 
Even if we assume "Group X" did A64 and in fact stayed together since, too, it doesn't mean they could pull out another "miracle chip" just like that - elite or not, you can't do miracles all the time.

And remember that big part of A64's success and halo wasn't because it was so good, but because Netburst was so bad

Its what i said, they was there on the good days and there on the bad days ( saying A64 days is wrong, because, AMD was really on top well before that, on the Athlon XP times, Barton, even when they have show their first processors ( during a meeting of Intel launch of new products )...

But the guys who leave todays, or have been ejected, ..they have not do anything "miraculous" thoses last 7 years, so when i see peoples said: " wow look, the situation is dramatic", the best of the best are leaving AMD " ( I refer to the xbitlabs article ) ... im not so sure it is really the case. Maybe the best for AMD is to get a dramatically change of the person who have lead them thoses last 6-7 years. ( And again, i dont critize them as person, or there capacitiy, because i Know how it was complicate to make move the things inside AMD )
 
Last edited by a moderator:
Well yes but the gpu in question here has a 128bit bus anyway, so chopping off half the interface (along with half the bandwidth of course) just to then use gddr5 in clamshell mode seems a bit awkward.

You could look that way ... or you could see that is the AMD's way to maximize favorable conditions for OEM. Less routes 64 vs 128 means much smaller space needed for cramping up lot of bulk memory chips, and btw gddr5 already share x32 organization while ddr3 maximally has x16 bus width per chip. And ddr3 is so wastefully cheaper than same 1-2Gbit gddr5 chip. So it's like always dirty chip OEM options with four times more memory just for marketing sake. And per chip ddr3 consumes less especially if it's not needed huge amount of memory so it could be powered down. some HD8500M 2GB would look so appealing for bunch of new nerds next year. Much more than mediocre 512MB GDDR5 which is so Y2k9 number in Y2013
 
Days are long gone when you walk out the door of a great job to another, no matter how good you are. For everyone that is considered elite there is two more waiting in line.
silentguy is right. The market is really hot for anyone with experience in graphics or low power. I'm sure other areas are hot as well.
 
silentguy is right. The market is really hot for anyone with experience in graphics or low power. I'm sure other areas are hot as well.
Qualcomm in particular is in a good position to pick-up AMD employees. They have a Markham office not a Tim Horton's away from AMD's with plenty of open positions. Meanwhile it was recently pointed out to me that they also have an Orlando office with open job positions in graphics (AMD's Florida location houses the Read3D guys).

Edit: Real3D, not ArtX. My bad
 
Last edited by a moderator:
Back
Top