AMD: R8xx Speculation

How soon will Nvidia respond with GT300 to upcoming ATI-RV870 lineup GPUs

  • Within 1 or 2 weeks

    Votes: 1 0.6%
  • Within a month

    Votes: 5 3.2%
  • Within couple months

    Votes: 28 18.1%
  • Very late this year

    Votes: 52 33.5%
  • Not until next year

    Votes: 69 44.5%

  • Total voters
    155
  • Poll closed .
I think they are talking about the mobile chips in this post-

Yes, but how does that change anything I said? We already got 256bit membus as far as I can remember on last gen, didn't we? So 128bit for middlerange and 64 for lowend?
 
Yes, but how does that change anything I said? We already got 256bit membus as far as I can remember on last gen, didn't we? So 128bit for middlerange and 64 for lowend?
Pretty sure Broadway is at least identical from a logic point of view to Juniper, though maybe fabbed on low-leakage process (and Madison/Park correspond to Redwood/Cedar respectively), so no Cypress for notebooks. Hence Broadway being 128 bit.
 
Pretty sure Broadway is at least identical from a logic point of view to Juniper, though maybe fabbed on low-leakage process (and Madison/Park correspond to Redwood/Cedar respectively), so no Cypress for notebooks. Hence Broadway being 128 bit.

I don't see why it should be, Mobility HD4800 was RV770 based and powerwise RV770 and Cypress are really close
 
I don't see why it should be, Mobility HD4800 was RV770 based and powerwise RV770 and Cypress are really close
But there are only 3 notebooks chips there, and we know there are 4 desktop chips. So are they going to skip some other chip (maybe the lowest one if intel's igp is pretty ok)? That would sort of make sense I guess, though those monster chips in notebooks are pretty rare (and you pretty much can't find notebooks with amd highend graphic chips in them anyway, all gaming laptops use nvidia for some odd reason).
 
I see one of the lowend chips being cut as far more probable option than cutting high end; using Juniper would yield next to no performance increase on laptop space (only what they could achieve by higher clocks), especially since they list one of their main objectives to win the DX11 laptop transition.
Having entries for high end, middlerange and lowend is essential to reach that goal, IMO, rather than having (high) middlerange and two lowends
 
But there are only 3 notebooks chips there, and we know there are 4 desktop chips. So are they going to skip some other chip (maybe the lowest one if intel's igp is pretty ok)? That would sort of make sense I guess, though those monster chips in notebooks are pretty rare (and you pretty much can't find notebooks with amd highend graphic chips in them anyway, all gaming laptops use nvidia for some odd reason).

At the moment there are the MSI GT725/727/729 models, the ASUS W90Vp and now also Alienware M17x has the option to use the 4870 cards in single and CF combinations.

I have the MSI GT725 and it is a GREAT (and not so expensive) notebook for the gaming, and the 729 is even better. But there is the "laptop drivers issue": opposite to Nvidia, on the ATI side new drivers come from the laptop manufacturer, not from the GPU manufacturer. So bug fixes and CF profiles could take eons to reach the average customer, i.e. on the ASUS site the driver for the ASUS W90Vp is still the launch one after months and it was really buggy and based on the Catalyst 9.5 :rolleyes:
MSI is a little better, as there were a couple of updates, anyway I update the drivers with the drivermodder program from driver heaven but of course this is not the path any user should use. So this could be one of the reasons why we don't see so many Mobility 48xx parts in the market. Moreover, the 4860-4830 Mobility never saw the light (and I would ask AMD why, as they would have been the best mobility parts out there: supply constraint? Bugs?).
Of course also Nvidia ties with laptop builders could have helped, but AMD/ATI really have to change something on the laptop support side if they want to surpass Nvidia in the mobile high end, especially for multi GPU solutions.

On the other side, the 4670/4650 cards were really a success.
 
I see one of the lowend chips being cut as far more probable option than cutting high end; using Juniper would yield next to no performance increase on laptop space (only what they could achieve by higher clocks), especially since they list one of their main objectives to win the DX11 laptop transition.
Having entries for high end, middlerange and lowend is essential to reach that goal, IMO, rather than having (high) middlerange and two lowends
NV doesn't have a GTX-equivalent mobile GPU, and they probably won't have a highend DX11 GPU for the mobile market, either. Winning in the laptop space isn't solely about being the fastest, but also about having lower power consumption and outputting less heat. Speed-wise, Juniper beats the pants off NV's last gen mobile offerings, and can do so with much better power and thermal specs. They can make a truly mobile platform around it instead of just DTRs.
 
NV doesn't have GTX-equivalent mobile GPU, true, but Juniper isn't GTX-equivalent performance wise either (well, lowend GTX perhaps) - and wether we like it or not, DTRs are part of mobile world today, and there's a market for them too.
 
NV doesn't have GTX-equivalent mobile GPU, true, but Juniper isn't GTX-equivalent performance wise either (well, lowend GTX perhaps) - and wether we like it or not, DTRs are part of mobile world today, and there's a market for them too.
They can get CFX. Everybody's happy!:D
 
They can get CFX. Everybody's happy!:D

It wouldn't push the performance envelope practicly at all, since there's already RV770 CFX and G92 SLI options available (i'm not exactly sure how the G92s are clocked on mobile world so don't wether they're match for RV770s or not)
 
It wouldn't push the performance envelope practicly at all, since there's already RV770 CFX and G92 SLI options available (i'm not exactly sure how the G92s are clocked on mobile world so don't wether they're match for RV770s or not)
M9800GTX, has 112SPs and is clocked at 500/1250/1600mhz.
I don't know what I would do without wiki.
EDIT- M4870 is a full RV770 @ 625mhz.
There is also a mobile 4850x2 on the wiki page, though I cannot recall ever seeing it in a lappy, clocked at 550mhz.
 
Last edited by a moderator:
M9800GTX, has 112SPs and is clocked at 500/1250/1600mhz.
I don't know what I would do without wiki.
EDIT- M4870 is a full RV770 @ 625mhz.
There is also a mobile 4850x2 on the wiki page, though I cannot recall ever seeing it in a lappy, clocked at 550mhz.

Not really, 4870 Mobility has the possibility to go up to 625 (some said 680) MHz with high thermal dissipation available, but most reports I have seen out there said normal level is 550 MHz. 4850 is rated up to 550 MHz, but normal level is @500 MHz, I have one so I should know it. 4860 is a RV740 @ 650 MHz.
If 4870 was really @625MHz (4850 desktop level, even more with GDDr5), it would have been already the fastest Mobile single GPU out there, as the GTX280M is at the 9800GT desktop level.
Problem is, some producers have used these GPU with lower specs than these, probably for thermal reasons, as the ASUSW90V, i.e., has two "4870" with GDDR3 and 500 MHz core clock, that is, a 4850x2 Mobility labeled as 4870x2.
It would be interesting to know which clocks for core and memory the 4870 for the Alienware M17x are specified, and which core and memory the new parts will have. A Juniper at full speed shoud anyway be the fastest Mobile card.
 
Last edited by a moderator:
Cypress @ 550MHz should be possible as Mobile HD 5870 and AMD could say their mobile hig-end delivers more FLOPs than competitors desktop high-end. :LOL:
 
Fudzilla says that the 5970 launches in a couple of days!

he might be right! :D

btw, MOVE OVER FERMI
9r3lug.jpg

2irlmh2.jpg

2ilqn6.jpg

2qrzktd.jpg

o1k0o.jpg
 
Last edited by a moderator:
The power figures show it's just barely under the 300W limit.
The the specification for a certified PCI-E device indicates it shouldn't exceed that limit during normal operation under typical conditions.

If a user overrides normal operation by overclocking the card, it would be their choice to do so.
 
Back
Top