The Official G84/G86 Rumours & Speculation Thread

Arun

Unknown.
Moderator
Legend
Rumoured Data Points
- Both chips will be manufactured on TSMC's 80nm process.
- Presumably the G84 is on a more expensive process variant (GT vs G?).
- The die sizes are said to be 160mm2 and 120mm2 for G84 and G86, respectively.
- Both G84 and G86 have a 128-bit memory bus, but some SKUs might have a 64-bit one.
- Originally, both were rumoured to be aimed at the February/March timeframe.
- G84 is now rumoured to be introduced on April 17th, and G86 in May.
- It looks like G86 has been further delayed than G84, thus.
- The G84 corresponds to the GeForce 8600s, and the G86 to the GeForce 8400s.
- The GeForce 8500s (->both G84 and G86?) and GeForce 8300s (G86 SKU) also exist.
- The clocks for G84 are said to be 700MHz/2000MHz for the GTS model; $199-$249.
- The G84GT will apparently be clocked at 600MHz/1400MHz; will be sold at $149-$169.
- The GeForce 8500GT is expected to clock at 450MHz/800MHz while selling for $79-$99.
- The G84 and G86 cores are rumoured not to support AGP (not pin-compatible with G73?)

Extra Tidbits & Fun Facts
While the difference in die sizes is much smaller than in the current generation (G73: 127mm2, G72: 77mm2), it is quite similar to the one in the previous generation (NV43: 150mm2, NV44: 110mm2). This point of view certainly makes the rumours on that subject more reliable in my eyes. For anyone out there not familiar with die size comparaisons, it's a much more accurate measurement of costs (for a given process and process variant) than transistor counts.

The G84 and the G86 will be NVIDIA's first DX10 GPUs for notebooks, and in a recent conference call, NVIDIA's CFO has claimed they expect to be able to capture "70% of the market" by this time next year. It remains to be seen, however, whether NVIDIA is expecting the RV610 and RV630 to be ready for the Santa Rosa cycle, which it looks like they will be at this point.

Noteworthy Internet Rumours (from newest to oldest)
"NVIDIA Plans For GeForce 8800 Ultra" [VR-Zone]
"NVIDIA to Launch GeForce 8600 Series on April 17th" [VR-Zone]
"No AGP For GeForce 8 Series" [VR-Zone]
"DirectX 10, sub-$100 graphics cards take shape" [The Inquirer]
"NVIDIA G84 and G86 codenames out in the wild" [Beyond3D]
"G84 & G86 Info" [VR-Zone]

Thread Discussion Starting Points
- Does all of this rumoured information seem reliable to you? What, if anything, sounds fishy?
- Are the rumoured products looking good or bad from your point of view?
- What extra facts would you expect to be true for G84 and/or G86?
- What do you think this means for the rest of the NVIDIA line-up?

- No reliable information has been leaked on the number of ALUs/TMUs/ROPs/etc.
- So, what do you expect there? Do you think some architectural tweaks have been done?
- Do you believe the ratios of the different units have been changed? How so? How would it affect things?
 
If the plan for a 8800 Ultra exists, there's no reason not to assume a 8600 Ultra is also in the works, since i don't think the 8800 GTS (especially the 320MB variant) will last for long.

And any refresh will be either faster (widening the performance gap), or a further crippled G80/G80 refresh (Nvidia usually doesn't do that until very late in the products cycle, like the 7900GS, 6800XT, etc).
This would be consistent with such a move, although a 65nm shrink of the G84 is not out of the question until the end of 2007, early 2008.
 
Will the ALUs run at the same speed as the core or will they be significantly faster like with G80?

Either way, would it have any major effect on cost/yields or power consumption since these are intended for mainstream/budget and/or mobile?

Also, could the lack of support for AGP be because the architecture is just so different from previous designs that the bridge chip would need to be redesigned to be compatible (assuming it were even possible)?
 
Will the ALUs run at the same speed as the core or will they be significantly faster like with G80?
I think it's a fudamental characteristic of the architecture that the core and shader clocks are decoupled - it'll be interesting to see how the ratio evolves between the two for future G8x/G9x chips though, definitely.
 
You are assuming similar price points and manufacturing costs, though. If, say, the 8400GS was faster than the RV630, then things could get ugly.

So, you suggest that AMD has nothing in the pocket what can fight with the 8500GT/8600GT/8600GTS, only the RV670 coming in Q3 (+ the ATi delay™®©)? and why AMD call the card as the X2600 series, when its coming out against the 8400GS?

AMD last hope is the 8600GTS slower than the x1950PRO (chance for this is 1%), but the x1950pro not support DX10, from every angle the situation looks bad for AMD, when its true.
 
So, you suggest that AMD has nothing in the pocket what can fight with the 8500GT/8600GT/8600GTS, only the RV670 coming in Q3 (+ the ATi delay™®©)? and why AMD call the card as the X2600 series, when its coming out against the 8400GS?

AMD last hope is the 8600GTS slower than the x1950PRO, but the x1950pro not support DX10, from every angle the situation looks bad for AMD, when its true.

Not true. The saving grace is the die size. If RV630 is smaller and uses less power, which it undoubtedly is (perhaps up to or more than 1/2 if the 160mm2 and <80mm2 rumors are true), they have a lot of wiggle room for price cuts and 'x2' type-products.

Can't really say I'm surprised the G84 is faster. The competition is the Rv660 (if it exists) or RV670.
 
Not true. The saving grace is the die size. If RV630 is smaller and uses less power, which it undoubtedly is (perhaps up to or more than 1/2 if the 160mm2 and <80mm2 rumors are true), they have a lot of wiggle room for price cuts and 'x2' type-products.

Can't really say I'm surprised the G84 is faster. The competition is the Rv660 (if it exists) or RV670.

When it's not on the market for many months (years ?) to come, it's not "competition".
Future and vague promises don't pay the bills, i thought AMD/ATI had realized that by now with the "top-to-bottom R6xx family launch"...

It appears that, for the 3rd time in a row, Nvidia prepared itself for another clean sweep in the midrange market, while ATI is watching, doing nothing.
 
When it's not on the market for many months (years ?) to come, it's not "competition".
Future and vague promises don't pay the bills, i thought AMD/ATI had realized that by now with the "top-to-bottom R6xx family launch"...
It appears that, for the 3rd time in a row, Nvidia prepared itself for another clean sweep in the midrange market, while ATI is watching, doing nothing.

Maybe its just FUD (NV PR team working again ;)), but when is true, i don't understand how a company can make the same mistake 3 times in a row without any personal consequences, slowly looks like somene need to save the company because whats going on in the last 18 months its a route to the end.
 
When it's not on the market for many months (years ?) to come, it's not "competition".
Future and vague promises don't pay the bills, i thought AMD/ATI had realized that by now with the "top-to-bottom R6xx family launch"...

It appears that, for the 3rd time in a row, Nvidia prepared itself for another clean sweep in the midrange market, while ATI is watching, doing nothing.

Q3 is not years away. It could be 3 months after G84's launch, or up to 6 months knowing ATi. True being in the market first with a part is very important, no doubt, and nvidia constantly launches their parts on time, in a certain predictable order, and DAAMIT does not. 8800 vs R600 is the same story. This is coupled with parts like the 8800GTS 320, which certainly ATi would have to launch a GT or RV670 to compete with. That does matter, don't get me wrong. I'm just saying assuming 7600gt's life (13 months), there still is time for AMD to make a comeback even with a later launch, only if they screw up RV660/RV670 like they did RV560/RV570 do I see a unsurmountable win for nvidia like last gen.

I think people get this "top-to-bottom" idea wrong. ATi has been fairly consistent with releasing a high-end part, along with a low-end and super low-end part around the same time, with performance mainstream later along the same time with the refresh on the high-end. This happened with both the x800 and x1800 series. This is what I expect this time.

On the other hand, Nvidia releases the high-end first, and then performance mainstream and low-end later at the same time. This is what happened with the GF6 and GF7 and now GF8 products. Certainly their plan seems to work better, and it looks like their launch schedule is perma-ahead of ATi now.

Please don't get me wrong, this isn't good for AMD, and is good for nvidia at least in the short-term, but we can neither say they are doing nothing (as they clearly are) or that this is entirely unexpected. It seems nothing more than history repeating itself based on each companies history of how they choose to develop and launch their products.
 
More details:

http://www.hkepc.com/bbs/itnews.php?tid=752184

According to their sources, the G84 products (8600 GTS and 8600 GT) are stronger than the upcoming ATI RV630...

Doesn't have to be a problem. It all depends on pricing and if AMD can get RV670 out fast enough which is now scheduled for Q3.

Personally I don't think AMD intends for RV630 to go up against G84, seeing as we've been hearing about some nameshuffling (eg RV630 -> X2400 instead of X2600).
 
What's the reason for the shuffle?

I don't know if it's even true for sure, but if it is... it might be because of several reasons. One might be that RV630 isn't performing well enough to justify the X2600 name. Another one might be because RV670 (and maybe RV660) is closer than we think so they might need some space in their name-schemes (eg X2600-X2800) instead of squeezing it into the 50s range when they appear onto the market.

And if the fastest RV630 is a much slower performer than the fastest G84 like HKEPC is stating, changing the names from X2600 to X2400 isn't a bad idea, since a lot of people will be expecting similar performance of a GF8600 and Radeon X2600. And they'll just be disappointed again just like most people were expecting the X1600 to perform just as well as the GF7600.
 
  • Like
Reactions: Geo
Doesn't have to be a problem. It all depends on pricing and if AMD can get RV670 out fast enough which is now scheduled for Q3.
Well, you'd expect NVIDIA to have a new product to compete in the upper-midrange by then too, being positioned above G84 since the gap between G80 and G84 is pretty damn huge. The large amount of redundancy in the G80 GTS is helping close it a bit, but it cannot do miracles either.

Still, assuming RV610 costs much less than G86, NVIDIA has nothing to compete in that market segment except G72 and any potential 80nm or 65nm variants of that, which isn't exactly feature-competitive with UVD+DX10. We'll see. I'm really under the impression that NVIDIA doesn't care a lot about the super-low-end discrete DX10 market, for the very simple reason that management cares a lot about margins.

This might seem ridiculous, but from my point of view, NVIDIA seems to make certain decisions very much based on how the financial community would perceive them, rather than how much gross profit it would deliver. And considering analysts tend to focus a lot on margins nowadays as an indicator of how healthy a business is, it only makes sense for them to focus on these segments if they can keep respectable magins. In the end, this could help AMD to make some very nice profit there, though.
 
Doesn't have to be a problem. It all depends on pricing and if AMD can get RV670 out fast enough which is now scheduled for Q3.

Personally I don't think AMD intends for RV630 to go up against G84, seeing as we've been hearing about some nameshuffling (eg RV630 -> X2400 instead of X2600).

So they fuck*d (sorry) up again the mainstream-performance parts, and not have for NV performance cards any answer for months?
Or this is not mistake, its a strategy ? than its a really bad one.

Now i have one question left, why the 121/128watt TDP come when the card has bad performance?
65nm, slow performance, 121/128 watt TDP, when is true this will be the biggest blow in VGA history after the NV30.
 
Now i have one question left, why the 121/128watt TDP come when the card has bad performance?
Well, if the performance is worse than G84, and yet we are assuming R600 to be competitive with G80, then you must assume that the design is either less scalable or that the RV630 chip is quite (very?) small. As for the TDP, it'll be interesting to see if that rumour actually turns out true.
 
Back
Top