So, do we know anything about RV670 yet?

[Ailuros]

I see Cooln'Quiet as not wasting electricity having your system blasting away full tit all the time rather than for cooling/fan noise & so keep such things on.

I don't think I'm that concerned about power consumption, otherwise I wouldn't have a G80 sitting next to it.

Full tit is relative; with CnQ enabled and it's own heatsink/fan the CPU temperature reduced by barely 1 or 2C at times, with miraculous peaks exceeding 75C under load.

Replacing the heatsink with a obnoxiously large fanless heatsink not only yielded 0 noise, but also peak temperatures that hardly exceed 60C even during heatwaves.

While I accept the waste of CPU cycles, I wonder wether those preserve power consumption more than a by ~15C lower core temperature during full load.

 

[nicolasb]

Latest from The Inquirer:

http://www.theinquirer.net/gb/inquirer/news/2007/11/05/amd-hd3800-pricing-shock

Edited highlights:

The interesting part is... the Radeons HD 3850, 3870 and 3870 X2 (R680).

The R680 is the only part planned for next year (January), but we would not be surprised if it appears as a Yuletide present to the hardware community.

The HD 3850 will exist primarily as a 256MB part, while the 3870 will bring 512MB to the table. Some 3850 parts may come with 512MB and some 3870 parts with 1GB of memory -for 20 and 50 USD/EUR above current asking price.

Pricing for both launch parts is very aggressive, and AMD claims it is all due to 55nm process working like a clockwork and achieving excellent yields. Radeon HD 3850 should retail for between $149 to 179 (USD), while 3870 should be between 200 to 230. This is brilliant pricing, going way below 8800GT.

The real truth is that AMD started running into clocking walls (with a selected heatsink-fan combo) and again decided to crash the prices instead of gunning for a better cooling setup that will enable those 800 MHz+ clocks we saw earlier in development.

...

This is what will sell as Radeon HD 3850:
55nm RV670PRO GPU at 668 MHz (why not 666?)
256MB GDDR3 memory at 828 MHz DDR (1.66 GT(/s)
10.69 GPixel/s fillrate
52.99 GB/s memory bandwidth
320 Unified Stream Processors
24x Custom Filter Anti-Aliasing support
...

As far as higher performing part, the 3870 is considered, specs are following:
55nm RV670XT GPU clocked to 775 MHz
512MB GDDR4 memory clocked at 1.2 GHz DDR (2.4 GT/s)
12.40 GPixel/s fill-rate
76.80 GB/s memory bandwidth
320 Unified Stream Processors
24x Custom Filter Anti-Aliasing support

 

[Lux_]

Edit: if true, then IMHO it's really remarkable that the number of stream processors is same across the "whole" lineup from 3850 to R680.

Well, I really hope that the cards will be compatible with some great thirdpary coolers (for example this or this) and the overclocks are limited only by temperature...

 

[turtle]

Edit: if true, then IMHO it's really remarkable that the number of stream processors is same across the "whole" lineup from 3850 to R680.

Well, I really hope that the cards will be compatible with some great thirdpary coolers (for example this or this) and the overclocks are limited only by temperature...

Eh, you never know. We may see a 3830 yet...You know somewhere along the line they will have to get rid of parts that are not 100% functional, ie with a 240 shader sku. I think they are wise to not even mention the possibility of this sku until much later (Q1?), as the price on it would surely be very low ($100?) and their flagship (for the time-being) is already going to be cheap, and they don't want to lose the potentially higher sales.

As for cooling...I too hope 3rd party coolers make a big difference. For once, I don't doubt Thinq's conjecture about the cooler not being able to cut it for higher clocks. ATi's stock cooler's have been known to be pieces of shit with the cores running pretty hot at default. Looking back at RV530/560, for example, it's not surprising considering rv670's die size it smack dab between the two, and the cooler is probably similar performance. I also grant them that AMD is probably stuck on using that single-slot cooler for both SKUs so they can pimp quadfire.

Still, $149-200 is pretty nice. I hope [H] reviews at least the 3850 (and heck, the 3870) against the 8600GTS for the general pc public (for whom only [H] seems to matter), because if (big if) they actually launch at those prices, that's the competition $-wise imho until the 8800gt 256mb and/or prices drop on the GT. Yeah, it's only $40-100 difference between the 3870 and 8800gt, but it's going to be $60-100+ difference between the 3850 512mb and the GT at the current rate, and even a greater amount vs. the 256mb product. How much of a playability difference do you really expect between the two 512mb RV670 products when all is said and done (overclocking)? Even still for $150, a 3850 256mb might be worth it to some, if they were going to grab a 8600gts.

$149-$179 is going to be a great entrance price to that level of performance, even if it is killed by the 8800gt.

Granted, it would be nice if Nvidia can match these prices sooner rather than later, as it surely seems they have the superior product. Perhaps once Flextronics gets it's parts out we'll start to see this?

 

[Geo]

Hi

Since we have this thread about RV670; so I made poll vote, I hope you guys don't mind.

Thanks.

best regards,

Shtal.
http://snappoll.com/poll/230631.php

Oh damn, it's too bad we can't do polls here with the forum software. Oh, wait, we can.

 

[Geo]

$149-$179 is going to be a great entrance price to that level of performance, even if it is killed by the 8800gt.

I have yet to see an 8800GT for less than $269. When ATI tried to expand the midrange price points with X800XL, their success was middlin' at best.

 

[ShaidarHaran]

I have yet to see an 8800GT for less than $269. When ATI tried to expand the midrange price points with X800XL, their success was middlin' at best.

I dunno, the X800 XL (and subsequently the GTO) sold quite well, at least at retail. Not too many OEM wins though, so the profitability of these parts is still in question.

 

[Geo]

I dunno, the X800 XL (and subsequently the GTO) sold quite well, at least at retail. Not too many OEM wins though, so the profitability of these parts is still in question.

It sold well by enthusiast standards perhaps, but not by midrange standards. Above $250 you have room to move the market around a bit by fiddling with price/performance, but then it's a much smaller market. That $150-$200ish price range market is much more stubborn, and a lot bigger --at the macro level they are after the best performance at their price point, not the best price point for their performance (given individuals may vary, of course).

 

[satein]

I think Jawed will be please to hear this news from the INQ
AMD prepares DX10.1 part for AGP computers

THE REMAINING AGP UPGRADE market will be all AMD's, it seems.

After HIS launched 2400 and 2600 parts for the AGP 4x/8x interface, we were set for a swan-song for the connector which has hosted our graphics cards in one form or another since 1997.

But AMD has decided to address the needs of this dying market with a DirectX 10.1 part as well. Manufacturers only made the switch to PCI Express this year, with PCI E parts taking over from AGP ones.

With Nvidia screwing up the GeForce 8 series (G8x chips cannot work with BR02, Nvidia's own PCIe-2-AGP bridge chip), the market was left open for AMD's X1950 AGP and different models from 2000 series. The 2900XT was just delivered in small volumes to encourage AIB vendors to develop a special PCB with an AGP connector. Also, power consumption was such that you would need two 8-pin connectors in oder to compensate for only 35W delivered by an AGP connector (PCIe 1.0a x16 can deliver 75W). But, 2400&2600 will not go into the history as the last AGP parts.

Since RV670 has advanced power features, the design of AGP board with this chip was not a pipe dream. 35W from the motherboard and a single 6-pin, 75W deliverable connector is all that's needed to power this 90-110W part.
...

 

[ShaidarHaran]

Why can't AGP just die already? PCI-e has been out since mid-2004. PCI-e2 mobos & cards are already out. Die! Die AGP!

How much life can one reasonably expect to get out of an AGP system at this point? I would say you're rather limited. At best you can hope for an X2 on an NF3 board, or a C2D on an 875 board (one of the few boards that allow this). Given the cost for a decent DDR2+PCIe system though, why keep throwing money into an old system?

 

[kyleb]

Because people like to walk into a store and find a faster videocard on the shelf that fits in their current computer.

 

[AlphaWolf]

I'm sure agp would die, if pci-e offered the average user a tangible advantage.

 

[ShaidarHaran]

I'm sure agp would die, if pci-e offered the average user a tangible advantage.

Oh, you mean like the ability to run significantly faster graphics cards?

 

[satein]

Why can't AGP just die already? PCI-e has been out since mid-2004. PCI-e2 mobos & cards are already out. Die! Die AGP!

How much life can one reasonably expect to get out of an AGP system at this point? I would say you're rather limited. At best you can hope for an X2 on an NF3 board, or a C2D on an 875 board (one of the few boards that allow this). Given the cost for a decent DDR2+PCIe system though, why keep throwing money into an old system?

LoL

Probably, the reason behind this is that there are planty of AMD systems with the AGP motherboard and AMD won't want to kill them off to put the user to get a new system that might not be the AMD CPU

Another reason is that on the good old workstation system with the AGP slot, people tend to upgrade a good graphic to prolong the usability of their system too.

 

[Sound_Card]

I thought that the accelerated grapahics port was made by ATi.

 

[AlphaWolf]

Oh, you mean like the ability to run significantly faster graphics cards?

I said average user. They aren't buying 8800 ultras or running anything in crossfire or sli, nor would they give a damn to do so. They just want the thing to work.

 

[Geo]

For many users it means no external power, and that seems like an advantage to me.

 

[AlphaWolf]

For many users it means no external power, and that seems like an advantage to me.

Because you don't have to plug in a 4 or 6 pin connector? How does that impact them at all? That's certainly not a reason for rushing out and buying a new motherboard.

 

[ShaidarHaran]

I said average user. They aren't buying 8800 ultras or running anything in crossfire or sli, nor would they give a damn to do so. They just want the thing to work.

8800 GT is only available in PCI-e. The fastest AGP card you can get on the NV side is an 7900 GS OC (basically a GT on AGP), and on the AMD side it's an X1950 XTX (until the 3850 AGP materializes).

 

[ShaidarHaran]

For many users it means no external power, and that seems like an advantage to me.

Exactly. I have a Prescott rig at home that will only accept a 775 P4 (not even Pentium D). But at least it's DDR2 & PCI-e. I was able to replace the pathetic X300 SE with an X1650 XT which didn't require a PCI-e power connector for a very healthy upgrade.