vertex_shader
Banned
3dmark default? games with AA? games with AF? games with AA/AF? games without AA/AF?
So, do we know anything about RV670 yet?
- Thread starter nicolasb
- Start date
3dmark default? games with AA? games with AF? games with AA/AF? games without AA/AF?
Or TMSC doesn't produce enough broken R600s.
Bouncing Zabaglione Bros.
Legend
So is the consensus that R680 was scrapped because (a) it suffered from the same problems as R600 (b) R700 is close enough to displace it, or (c) both (a) and (b)?
Be patient... Leaks will come soon..
10K+ is reachable though in 3DM06 on an FX62 with a RV670 at 600Mhz if what I'm hearing is true. Now try to imagne what the faster clocked RV670 will score on a faster CPU like a C2D or C2Q.
...
I criticism companys only when they deserve, if you say AMD/ATi not deserve any criticism in the last ~2 year (and mostly this year with r600 saga) thats ok for me, because i see the things different thats not mean i dislike something
I think that considering the massive internal change going on inside ATi for the last year, at least, it's nearly remarkable that R600 shipped at all. Much of the so-called "criticism" of ATi during the period preceding the merger announcement and since has been written from a "business as usual" standpoint as if ATi had not been bought by AMD. It's especially true of the period preceding the buyout announcement when few of us knew anything about the merger and were ready and eager to scoff at the merger rumors we were hearing (yours truly included.) The thing is that criticisms directed at the company during that period which do not include mention of the internal changes taking place inside ATi as a direct result of the merger are simply invalid. The ATi that will emerge at the end of the merger tunnel will, I think, bear scant resemblance to the ATi we have known since the company remade itself with the launch of R300 in 2002. The strategies of the new ATi will be much different from the strategies of the old, and just as the acquisition of ArtX fundamentally remade ATi from the inside, so I think will also be true for the end result of the ATi-AMD merger.
If I had to point to the single largest mistake ATi made post R300 and pre-AMD, I believe that it would be ATi's presumptive rush to hop on the PCIe bandwagon at the expense of its existing AGP-based customers--which at the time comprised 99% of that customer base, if not 100%. nVidia, otoh, was far more reasonable in its approach to the emerging standard of PCIe in that the company did not whole-hog leap its production capacities toward PCIe-only devices well before the PCIe standard reached the critical mass of public adoption necessary to provide a fertile market for discreet PCIe devices.
IIRC, unlike ATi, which couldn't wait to be the first to announce that it was shipping a "native" PCIe-only product line, I believe nVidia's first PCIe discreet products were AGP products bridged to PCIe pcbs--which as I said at the time was the correct approach. As well, from the standpoint of the company itself, nVidia's first substantial investment into PCIe product production came with SLI, preceding ATi's CrossFire initiatives, and in the process made the first practical use of the PCIe standard in terms of what the PCIe bus provided that the AGP bus could not--which was the ability to mount and run more than one discreet graphics device simultaneously. The press often played the significant difference between the two buses as one of bandwidth, but as nVidia knew very well, the bandwidth locally on their discreet products at the high end already far exceeded the system bandwidth provided by PCIe x16. And so nVidia's PCIe product push went toward the tangible PCIe advantage--which was the ability to mount SLI.
The irony to me in all of this is that although ATi reached its goal of being the first to announce and produce PCIe-native products, it was nVidia which obviously better understood the real differences between the two buses: namely, nVidia understood the economic fallacy of outrunning its markets with respect to PCIe adoption, and then nVidia provided the first, true, and practical use of the emerging PCIe bus standard--the new SLI (new as opposed to the "old" SLI coined by 3dfx.) [Note that this is not to express a personal opinion that values either SLI or Crossfire, because it isn't...
But it does show you the differences between the two companies in their approach to the emerging PCIe bus standard.]This is where I think ATi tripped up after having started so well out of the gate with R300. This is the kind of mistake that comes directly from the top. Apparently, someone highly placed inside ATi at the time reached the conclusion that what had sold R300 and after so well was not the technology itself but rather the marketing of that technology. Hence, it was decided that by sheer brute marketing force employed in a "We're first to do native PCIe products" campaign, that millions of people around the world would rush out and dump their AGP motherboards and AGP 3d cards and buy all of the other stuff required by the newer PCIe environments simply to be able run PCIe 3d natively.
Alas, such was not the case, and ATi has paid for that critical mistake ever since, imo. They wound up with lots of excess native PCIe products that nobody wanted because they weren't ready to move to PCIe, and they experienced a production shortfall of AGP products demanded by the people still running AGP, simply because they had overestimated the short-term demand for native PCIe graphics cards while underestimating the short-term demand for higher end AGP graphics cards.
This kind of mistake was surprising to see coming out of ATi, because ATi had been around even longer than nVidia, but nVidia grasped the essentials of the situation better at that time. Somebody at ATi learned first-hand the folly of outrunning your markets, and my guess is that the mistakes in judgment that led to that situation were made by somebody brought into the higher echelons of ATi from ArtX. I can well recall having to wait a long time to buy an X800 for my AGP system, as I had wanted to buy one much earlier but was unable to do so because I couldn't find one.
Basically, this is the kind of criticism and observation that I think is apt in relation to ATi. But most of the other criticisms I read pretend that ATi never merged with AMD, and do not take into account how that merger affected the internals of the company in terms of the time and energy the reorganization took away from product development--and is possibly still taking away from product development, although I believe the critical threshold has been passed and the new ATi is firming up nicely.
On a personal note, I have to say that my current AGP x1950 is the best 3d-card I have owned to date, certainly a lot better than the x800xt it replaced, but because of the mistake I've commented on here I was forced to wait far too long to buy either the x800xt AGP or the x1950 Pro AGP. Now I am wondering if the next release of the Catalysts will even support my x1950 Pro AGP card...which is the first time since 2002 that I have had to wonder about my currently functional AGP 3d card being supported by an ATi driver release, since apparently my product hasn't been supported by the Catalysts since the 7.7 CCC Catalyst release.
Ahead of the yet unknown competition? Wow, how cool is that?
That is why I said "rumor says". Considering that is much easier to get rumors from people closer to AMD/ATI, then from the other side, it could be true or false. Best guess is 3DMark without AA/AF, might even be comparing not comparable versions of the chip.
A chef has a roast in the oven in the back of a diner. This big roast takes 4 hours to cook.
2 hours in, somebody repaints the exterior sign at the front of the diner.
Does the roast explode in the oven because of that?
On the scale of things the AMD buyout did to R600's development, between "repainted the sign in front" and "Mongol hordes on steppe ponies burst into the engineering department, forcing Eric Dremers and the surviving engineers to fight them off in brutal guerilla warfare, using R600 engineering samples as ammunition and PCBs as body armor", I'm going to lean more towards the sign change.
The decisions that lead to R600's delays, the new design, the unified clocking of a high transistor part, the complex virtualization of the hardware, the transition to a new API, the selection of an 80nm process that was known to have leakage issues, all occurred long before they started shuffling executives and names on contracts in a department that was likely physically and organizationally removed from the design and implementation of R600.
You've most likely gotten this very backwards.
Before and during the merger, graphics work at ATI was likely little affected.
Now that ATI's cash reserves were swallowed by the buyout and the merged company is billions in debt, product development will be affected.
Considering that AMD will not sacrifice funding to the CPU lines, the cash-strapped company is going to chop the funds for GPU designers (that it hasn't appropriated for Fusion) first.
Could you explain what is bad about the use of DDR4 or point me to a reference to learn? I am not trying to argue; I just don't the advantage/disadvantage of the different types of VRAM.
Be patient... Leaks will come soon..
10K+ is reachable though in 3DM06 on an FX62 with a RV670 at 600Mhz if what I'm hearing is true. Now try to imagne what the faster clocked RV670 will score on a faster CPU like a C2D or C2Q.
I think that RV670 will be in High-End line. For instance, the RV670X2 ,RV680X2 .
vertex_shader
Banned
That comment i made months ago when i good remember, things changed a little bit
Gddr4 its not bad at all, the problem is still not have any significant advantage over gddr3 memory because mainstream/cheaper gddr4 ram still in the 0.8ns (2200mhz) zone with higher latency than gddr3 version.
Gddr4 using less power in paper but when we check the gddr4 version of hd2600xt not looks like in real its have any advantage, for the gddr4 version more complex PCB needed.
According to the rumors with rv670 the story looks different than with hd2600xt, because GDDR4 version core clock is higher too, and this GPU not a cut-cut-cut down version of r600, so can profit from the higher BW, time will tell :smile:
Be patient... Leaks will come soon..
10K+ is reachable though in 3DM06 on an FX62 with a RV670 at 600Mhz if what I'm hearing is true.
R600 rocks in the super funny game 3dmark-06 too, too bad it struggles when it comes too....what??
1. Cooking
2. Get rid of jaggies and still be able to play the damn game
3. Driving
Lets hope daamit delivers this time.
vertex_shader
Banned
Its not that great, because this can generate to high expectations/hype, and when NV silent launch something what kick the rv670 in the butt than only thing can come is the dissapointment over rv670.
vertex_shader
Banned
They are going to make substantial profit instead of fierce competition.
At least, both of them are making cheaper board for mainstream product.
