AMD's DirectX 10.1 entry and mid-range level GPUs: RV620 and RV635

Ahh, then you need rephrase and ask what you actually want to. It's mind blowing when people say "but I don't want to use the 6-pin connector" which little sense, maybe they're actually trying to say "I don't believe my power supply is up to the power demands" or such.

I don't know, in my first post I made it pretty clear that using a 6-pin connector was out of the question. Just roll with it, it's not like I'm some uninformed nub here...
 
1. PCIExpress 2.0 is 75W instead 150W
2. PCIExpress 2.0 is 16X instead 32X (Example Radeon HD3870)

What's the point of PCIE 2.0 anyway?

x32 is part of 1.x specs too, there's just no slots/cards for it, but 2.0 doubles the datarate per pin compared to 1.x
 
x32 is part of 1.x specs too, there's just no slots/cards for it, but 2.0 doubles the datarate per pin compared to 1.x

If I could recall correctly their is no performance gain PCIE2.0 vs. 1.x.
The only thing I may benefit of PCIE2.0 specs is increasing 75W to 150W in slot, but it seems it's absence right now.
 
If I could recall correctly their is no performance gain PCIE2.0 vs. 1.x.
The only thing I may benefit of PCIE2.0 specs is increasing 75W to 150W in slot, but it seems it's absence right now.
There may be no real-world gain right now in performance, but the data rate certainly doubles (though frankly I'd think there would be more benefits for the cheap 1x cards like sata controllers, but they aren't in sight yet - not to mention nvidias "new" 780 chipset of course only supports pcie 2.0 with the 16x slots).
I think there was (and still is) some confusion about the power limits, since even the pcisig page mentions changed "slot power limits". However, the only thing which seems to be new is the 8-pin connector, with pcie 1.0 cards are restricted to 150W (75W slot + 75W 6-pin connector), whereas pcie 2.0 cards are allowed to use 225W (75W slot + 150W 8-pin connector) or 300W (75W slot + 75W 6-pin connector + 150W 8-pin connector). Though apparently there are cards out there (HD2900XT) which already used the new connector with pcie 1.0...
This information would be in the pcie 2.0 CEM spec, but I haven't seen anyone quote from the final version of it yet :). But maybe increasing the slot power would have required changing the slot physically, which isn't a very hot idea (it has been tried before - remember AGP pro, which no cards supported ever?)
 
To increase the power spec you would be burdonening all motherboards and systems (by virtue of the PSU specs) that support an x16 lane with increased costs to support said power, when its only a minority graphics devices that actually require it (when you count IGP and add-in boards < 75W). This, somewhat, goes against the energy star push that is occuring in the PC industry right now.
 
There's really no need to recall incorrect facts. Just Google "PCIe specification", which results in this page.

For your convenience:

Yesterday I was driving my Ford Pickup truck on Freeway with speed Limit 60MPH, but Today I was driving my Ferrari Spider 360 on Freeway with same limit 60MPH as well. :(
 
There may be no real-world gain right now in performance, but the data rate certainly doubles (though frankly I'd think there would be more benefits for the cheap 1x cards like sata controllers, but they aren't in sight yet - not to mention nvidias "new" 780 chipset of course only supports pcie 2.0 with the 16x slots).
I think there was (and still is) some confusion about the power limits, since even the pcisig page mentions changed "slot power limits". However, the only thing which seems to be new is the 8-pin connector, with pcie 1.0 cards are restricted to 150W (75W slot + 75W 6-pin connector), whereas pcie 2.0 cards are allowed to use 225W (75W slot + 150W 8-pin connector) or 300W (75W slot + 75W 6-pin connector + 150W 8-pin connector). Though apparently there are cards out there (HD2900XT) which already used the new connector with pcie 1.0...
This information would be in the pcie 2.0 CEM spec, but I haven't seen anyone quote from the final version of it yet :). But maybe increasing the slot power would have required changing the slot physically, which isn't a very hot idea (it has been tried before - remember AGP pro, which no cards supported ever?)

Thanks. :)
 
HD3670 256mb ES review

Rv630 149mm²
Rv635 120mm²

Its a engineering sample but the card 55nm GPU has lower GPU clock speed (725mhz) than the hd2600xt GPU at 65nm (800mhz).
Expreview tested the card at hd2600xt 800/1600 clocks, its consume more power at idle/load than the 65nm version, this can be change if the retail card has a newer revision GPU, when not than than this shrink goes in the bad way.
Card support powerplay, 2d clock speed are 110mhz/405mhz.
 
Last edited by a moderator:
According to this slide, AMD isn't going to launch HD3670 - only HD3850:

02RV620.jpg
 
According to this slide, AMD isn't going to launch HD3670 - only HD3850:

02RV620.jpg

Rv620le = xx50, rv620pro = xx70, rv635pro = xx50, weird.
Well time will tell what coming, its won't surprise me after i read expreview rv635 review they skip the XT version.
 
Rv620le = xx50, rv620pro = xx70, rv635pro = xx50, weird.
Well time will tell what coming, its won't surprise me after i read expreview rv635 review they skip the XT version.

That slide is correct.

There is currently only 1 RV635 SKU and that's the HD3650. So that preview is falsely assuming that there will be a HD3670.

The HD3650 will come in 2 variants though. One with GDDR2 and one with GDDR3.
 
Back
Top