So, do we know anything about RV670 yet?

[trinibwoy]

When was the last time AMD shrunk their flagship and called it a RV sku?

When was the last time AMD's flagship had to be discounted because it could only compete with the competition's second tier parts? This wasn't exactly R580 vs G71 now was it? R600 was already in the "performance" bracket six months ago at launch. Why would you expect a new chip to be much slower? You mention high-clocks but that's irrelevant. What we're discussing here is performance levels. Whether RV670 got there through high-clocks or by retaining all of R600's units doesn't really matter. My overarching point is that the performance of the 8800GT isn't "too good to be true". The 6600GT did the same thing and the 7600 as well to some extent. It just looks like the second coming because the 8600 and 2600 lines were so pitiful.

This is the first and one that caught Nvidia offguard. However they managed to combat the situation albeit messing up their lineup.

The only thing messed up is the naming, not the performance. The $250-$300 part is exactly where it should be a year after the initial architecture launch. In your theory exactly what would would be done with all those low volume 96/112/128 shader parts after the 64 shader SKU volume was satisfied?

 

[Arun]

Let's say it's the same for G92. Even if 100% of G92 is built out of redudant blocks (which is ridiculous, of course)

Perhaps I'm just nitpicking here, but a discussion on redundancy is something I've wanted to have for quite some time: why would that be the case? This probably isn't the best thread for that but...

If 80% of your design is regular and easy to get redundancy out of, while 20% is irregular... If having zero redundancy in that 25% causes you to have 75% yields instead of ~99%, then doubling that 25% would result in a 1.25x bigger die but with much higher yields. However, the shocking part (okay, not so shocking since I picked the numbers!) is that the 1.25x bigger die is actually cheaper to manufacture, as 1.25 < 1.33 (=1/0.75).

 

[Lux_]

3890 or some such and no major improvements in R680 with R700 not launching until the 2nd Half next year.

I'd really like to see some kind of a roadmap from AMD when their platform launches: what is R680, what comes in Q1, spring and summer (R700 no earlier than July, I suppose). I'm not holding my breadth, though...

Regarding actual clocks - 800 vs 775 is 5%. If they give up that in immediate performance, they'd better win it back with game patches supporting DX10.1 and fast drivers (or at least cool techdemos...)!: No breathholding here, either...

 

[Jawed]

Unless anyone did actually see that page

Worked fine for me. It even had a picture of a box for a 3870 on it.

Jawed

 

[Arty]

When was the last time AMD's flagship had to be discounted because it could only compete with the competition's second tier parts? This wasn't exactly R580 vs G71 now was it? R600 was already in the "performance" bracket six months ago at launch. Why would you expect a new chip to be much slower? You mention high-clocks but that's irrelevant. What we're discussing here is performance levels. Whether RV670 got there through high-clocks or by retaining all of R600's units doesn't really matter. My overarching point is that the performance of the 8800GT isn't "too good to be true". The 6600GT did the same thing and the 7600 as well to some extent. It just looks like the second coming because the 8600 and 2600 lines were so pitiful.

True re: R600 not being competitive with the GTX. However how many people expected RV670 to be R600 on 55nm ..

And 8800GT with 96SP would be closer to looking like a 6600GT/7600GT. The way it is right now (112SPs), it is about 140% faster than the previous flagships.

The only thing messed up is the naming, not the performance. The $250-$300 part is exactly where it should be a year after the initial architecture launch. In your theory exactly what would would be done with all those low volume 96/112/128 shader parts after the 64 shader SKU volume was satisfied?

And people buy things based on brand names ..

Like I restated, after satisfying the 96SP sku, the 112/128 skus would be used in the GTS and GX2.

I think you should also check with Arun, even he believes that the 8800GT has a little more oomph to it than what Nvidia had originally planned ..

 

[Shtal]

Not at all suspicious that the image is the same width & only a couple of pixels different in height than these linked by Fudo?
Notable differences being use of HD2600Pro boxes in the B3D post + some slight typo differences.

The only explanation that RV670XT clocked @ 775MHz

A. For Quad Crossfire Setup (You need single slot cooler)
B. (Reduce cost) to fit in ~$200 dollars range.

But company like ASUS may use dual-slot solution for clocking RV670XT higher than 775MHz.

My guess estimate on 55nm you could clock RV670 @ ~875-900MHz range GPU and still it will run cooler than HD2900XT.
Edit: (Probably reserved against new Nvidia GTS-512) Probably be something like this "ATI Radeon HD3890 1GB"

My concerned that RV name for RV670 due to low cost for the GPU.

 

[CJ]

RV670 on the same lines would be 3/4 R600 or a 240SP sku.

Bingo.

But plans change...

When was the last time AMD shrunk their flagship and called it a RV sku? This is the first and one that caught Nvidia offguard. However they managed to combat the situation albeit messing up their lineup.

That's what I've heard too. NV got caught offguard when they found out that RV670 wasn't 75% of a R600 like originally planned.

 

[hoom]

Worked fine for me. It even had a picture of a box for a 3870 on it.

Damn So 775mhz

 

[SirPauly]

When was the last time AMD shrunk their flagship and called it a RV sku?

The ATI Radeon 7500?

 

[turtle]

The ATI Radeon 7500?

Actually, if we're nit-picking, I'd go with the x800xl.

Granted, 110nm was not ATi's friend.

ok...I guess granted not a RV chip, but it should'a been.

 

[Arty]

Actually, if we're nit-picking, I'd go with the x800xl.

Granted, 110nm was not ATi's friend.

ok...I guess granted not a RV chip, but it should'a been.

Wasnt that in the same price bracket as the current GTS?

 

[stevem]

The ATI Radeon 7500?

That's true, but the analogy won't hold until R700 is released.

 

[trinibwoy]

afaik, no need of special versions.

Actually it looks like the application has to explicitly create an Nvidia PerfHUD device for analysis. So far I've only gotten the UT3 demo to work.

(Sorry for the OT)

 

[chavvdarrr]

Actually it looks like the application has to explicitly create an Nvidia PerfHUD device for analysis. So far I've only gotten the UT3 demo to work.

(Sorry for the OT)

hmmm, I asked in ixbt forum how did they tested..

 

[turtle]

Wasnt that in the same price bracket as the current GTS?

Nope. $299. I call the it the same 'performance' bracket as 8800gt (>$200-$300) where-as the GTS' are >$300. I got mine from a reviewer for $200 right after he reviewed it since it couldn't take on the 6800gt, which was in the now-GTS bracket.

So many segments now-a-days. <50, 50-100, 100-200, 200-300, 300-400, 400+. I miss the three brackets consisting of:

1. Shit noone used for gaming, IE it drove a monitor. (9200/4200)
2. The mid-part everyone that wasn't made of money bought. (9500/4400)
3. The E-penis sector. (9700/4600)

At least then you knew what you were getting, and could depend on games being designed towards a smaller amount of hardware, and you could expect decent performance from a mid-range solution as most developers planned on gamers having that level of performance. I think the 8800gt (and perhaps 3870) thankfully will bring some of that back though, with games designed toward that level of performance and not above it (G80/R600) or below it (the craptacular G84/Rv630).

Apologies on random tangent.

 

[ShaidarHaran]

1. Shit noone used for gaming, IE it drove a monitor. (9200/4200)
2. The mid-part everyone that wasn't made of money bought. (9500/4400)
3. The E-penis sector. (9700/4600)

Way off with #1. The 4200 was one of the best bang-for-buck gaming cards of all-time. Perhaps you're thinking of an MX420/440?

 

[trinibwoy]

Way off with #1. The 4200 was one of the best bang-for-buck gaming cards of all-time. Perhaps you're thinking of an MX420/440?

For real. The 4200 and 9500 were probably two of the best cards ever brought to market.

 

[digitalwanderer]

/me gives props to the 4200!

The GF4 ti4200 was a freaking monster of a value performance card, I had 3.

 

[dizietsma]

The 4200 was generally faster than the 4400 I seem to recall as it came out later and had slightly better memory on it and slightly more overclockable gpu due to the more mature process.

Didn't the 4200 actually end up replacing the 4400 rather than being the low end?

The MX GF4 series was not so good It was beloved of PC sellers who could stick it cheaply in machines for low cost and millions were stuck in DX7 hell . I would hazard a guess a lot of those got replaced by GF 6600 AGP cards when nvidia released that.

What's the modern day equivalent of the GF4 MX range I wonder ?

 

[Kaotik]

The 4200 was generally faster than the 4400 I seem to recall as it came out later and had slightly better memory on it and slightly more overclockable gpu due to the more mature process.

Didn't the 4200 actually end up replacing the 4400 rather than being the low end?

The MX GF4 series was not so good It was beloved of PC sellers who could stick it cheaply in machines for low cost and millions were stuck in DX7 hell . I would hazard a guess a lot of those got replaced by GF 6600 AGP cards when nvidia released that.

What's the modern day equivalent of the GF4 MX range I wonder ?

Nope, the whole 4200-4600 line was out at once, 4200 wasn't any faster than 4400, it had both lower core and mem clocks.
4400 was "replaced" by 4800SE, and 4600 by 4800 later on, they were the "AGP 8x" versions of the said chips