Search results

  1. R

    5670 Review

    Folowing up 5670, looks like the low end part is arriving soon as well. From arabhardware.net forums some photos and quick benchmarks of the Radeon 5450(Cedar): http://www.arabhardware.net/forum/showthread.php?t=153562 (Probably should have started a new thread for this but anyway)
  2. R

    5670 Review

    The chip also has a DDR3 version the 5570, at the moment i can only find a chinese review that has been up for awhile here: http://vga.it168.com/a2010/0107/834/000000834487.shtml Provided the OEM supplies the required video outs think this should also support eyefinity and should come in...
  3. R

    5670 Review

    AnandTech Hardwarezone have their 5670 review up early(don't think the others are due till the 14th): http://www.hardwarezone.com/articles/view.php?cid=3&id=3109 Looks quite a bit above the 4670, perhaps +50% in places, also just above the GT240, sometimes even breathing down the neck of...
  4. R

    AMD HD5xxx sales figures: 300k HD58xx, 500k HD57xx

    I cant imagine that they would want to do that much, as it pushes up inventory and increases chances of mispredicting future order quantities as would have to estimate too far into the future. There should be enough capacity on line now that can handle reasonable order quantities. The...
  5. R

    AMD HD5xxx sales figures: 300k HD58xx, 500k HD57xx

    From TSMC : I assume the last part should read "40nm output will exceed 20,000 wafers per month by December of 2009" When considering relative amounts of wafers produced, make sure you include all the 40nm products: Nvidia : GT218, GT216, GT215, Fermi AMD : RV740, Cypress...
  6. R

    Charlie Demerjian: Nvidia roadmaps turn up, Ugly and devoid of hope

    I don't think the socket licensing fees will drop that fast. From intel's perspective their chipset business is now only really only selling a southbridge, they dont want the income for that division to plummet dramatically so they appear to be charging alot more for PCH + high socket fees to...
  7. R

    Charlie Demerjian: Nvidia roadmaps turn up, Ugly and devoid of hope

    It not only the above causing problems, there are technical obstacles as well as indicated by this article: Intel's triumvirate of licensing, high socket fees and low bandwidth have scorched the ground for third party chipsets.
  8. R

    Geforce GTX 295 previews

    nm
  9. R

    Geforce GTX 295 previews

    If i was in the market for this card i would take that price. Will be very limited availability at first. Otherwise suggest hiberating in the basement till at least late january when should be more available. $499 for 2 55nm GT200s seems quite a good deal really, the GT280 is still close to...
  10. R

    GTX 280 cards coming down?

    The recently released tesla fx card with a 55nm gt200 has a 512 bit memory bus suggesting that they managed to preserve that during the shrink. My preferred explanation is that the 295 board is too difficult to design with 2 x 512 bit memory interfaces and they have saved some board layers...
  11. R

    Where the hell is my GTX 290

    It's turkish i think. So GTX260+ mid january and GTX285 + 260GX2 late jan early feb. Your step up should be fine. I think the theory behind the later 285 release is that not many chips of the wafer are perfect and they will need time to bin enough of these to put together a release.
  12. R

    Nvidia stockpiles 55nm parts for a massive assult on ATI

    Expreview has put up a picture of zotacs new GTX260 board for the 55nm GT200s: http://en.expreview.com/2008/1...nm-geforce-gtx260.html The P654 10 layer board replaces the original P651 14 layer board. New power module, and memory has been moved all to one side of the board using samsung...
  13. R

    Nvidia stockpiles 55nm parts for a massive assult on ATI

    I have been thinking this, and having real difficulty explaining it at all... Look here the expreview article: http://en.expreview.com/2008/1...5nm-gt200-exposed.html Suposedly 55nm Gt200 product date week 33 2008 (ie august 11-17) They were producing 65nm parts only a few weeks before...
  14. R

    Nvidia stockpiles 55nm parts for a massive assult on ATI

    It is generally quoted that a single 300mm 65nm die from TSMC costs around $5000. TSMC has lots of customers so i guess over time the info has leaked out somehow. 55nm is more expensive but hopefully not too much. So for a back of the napkin calculation: Total die area = pi * (300/2)^2 =...
  15. R

    Dual GT200 samples in December

    Also the 280 would have to design a board with 2x512bit memory interfaces compared to 2x448bit memory for the 260. All the dual cards so far 7950, 3870, 9800, 4870 appear to have been 256 bit mem interfaces jumping to 448bit in roughly the same area would be hard enough let alone 512. The...
  16. R

    Dual GT200 samples in December

    I might be mistaken but i think Zap was referring to a new board design for the 55nm 9800GTX+. His statement is kind of ambiguous though...if the 55nm GTX260 board now only has 1 power connector that implies quite dramatic power savings. Apparently the 9800gtx had a 12 layer board and the...
  17. R

    55nm GT200 Picture

    More from expreview: http://en.expreview.com/2008/1...as-geforce-gtx295.html "Dual-GPU designed GeForce GTX260 GX2 graphic card will be officially named as GeForce GTX295. Apparently, GeForce GTX295 is coming to regain its dominance of performance which has been grabbed by AMD Radeon...
  18. R

    55nm GT200 Picture

    A die shot of a 55nm GT200: http://en.expreview.com/2008/1...5nm-gt200-exposed.html "The first of new GTX260 graphic cards will hit the market in January next year. The number of its stream processors will maintain as 216, and its memory frequency is unchanged as well. The reference and...
  19. R

    No GTX 270/290

    Not easily :( I think to be 100% sure have to pull off the HSF and scrape off the thermal compound to reveal the sku on the actual chip. Then also measure the size of the chip with a ruler to double check. For instance on the 65nm 9800s is printed G92-270-A1 and on the 55nm G92-280-B1...
  20. R

    No GTX 270/290

    Yeah, they wouldnt do it otherwise. Requires high margins though. For instance hypothetically say they sold you a GTX260 for $250 and you later stepped up to a GTX280 at $350. You have paid them a net total of $350 and they have a second hand 260 that can be resold or used for warranty...
  21. R

    No GTX 270/290

    Well looking at their other cards: G92 -> G92b 9800gt is both 65nm and 55nm G94 -> G94b 9600gt is both 65nm and 55nm G96 -> G96b 9500 is both 65nm and 55nm Given above it wouldn't be improbable that that the same thing occured for the GT200 series cards as well. Has the advantage when...
  22. R

    When are the new Nvidia cards coming out?

    I am not sure of the truth of the above rumors. The power saving i think will be less, perhaps 20%, they will try for more core clock up to 700mhz maybe. The second article is also slightly wrong, the tesla cx that came before the fx was also a 55nm chip, for some reason nobody much noticed...
  23. R

    FX5800: Most Powerful Graphics Card in History!

    Except according to techreport it has a higher fill rate 52b texels/s compared to the GTX280 at 48.2b texels/s Suggests maybe a higher clock? ie 52/48.2 * 600 = 647mhz
  24. R

    FX5800: Most Powerful Graphics Card in History!

    Er the second, the Quadro CX was also supposedly a GT206. Interestingly the specs on the nvidia product page dont show the memory interface for the FX. The CX has a 384bit memory interface(looking today, that has been removed from the nvidia specs sheet for some reason). Assuming they...
  25. R

    AMD plans G200b counterattack

    The first link is from 2007, things have changed since then, not sure who is using TSMCs 45nm process, everything i had heard from their clients was on 40nm. At the bottom of the second link is a reference to a digitimes article on TSMCs 40nm process: "Taiwan Semiconductor Manufacturing...
  26. R

    AMD plans G200b counterattack

    That date seems way too aggressive! The wafer supplier for tsmc only announced it is taking orders for production levels of 45nm wafers in march next year. Once tsmc has the wafers the foundry takes at least 30days(from memory, it might actually be closer to 60) and then another 30days to...
  27. R

    Power and Innovation to Drive High-End GPUs in 2009

    According to a vr-zone the first ati chip shrunk to 40nm hasnt shown much reduction in power consumption. So for a RV770 shrink 256mm2 @ 55nm -> ~ 140mm2 @ 40nm. If the worst case happens and there is no or little power savings that will be around say 140W meaning the chip is emitting heat...
  28. R

    nVidia profit down 74% in Q3

    I was wondering about that to, then i found the earnings call transcript which talks about how they sold $75m of MCP79 chipset presumably to apple and others at quite a high margin. Their workstation stuff seems to have gone fairly well, and the sony licensing revenue also made things good...
  29. R

    ATI RV870

    Sorry for the delay Mr Azn, have been away! Here is a labelled die shot of the RV770 from Rage3d Very roughly using a ruler the shaders in total occupy about 25% of the total die area so: 25% * 256mm * (1280 - 960)/800 = 25.6mm2 for the extra 320 shaders. Therefore total die size for...
  30. R

    ATI RV870

    Ok to do this need to make an assumption: That ATI's first 40nm chip will be a straight die shrink. They wont alter anything that cant be avoided. Looking at ATI's current designs we have from most recent the RV730, RV770 and RV710. They are all at 55nm, so shrinking to 40nm: RV730 146mm2...
  31. R

    ATI RV870

    Does not the 3870 have higher clocks and more bandwidth(GDDR4) than the 4850? Maybe the 3850 is a better compare? From the same site: Techreport 4830 Review There seems some improvement in the 4850 from the 3850 say 25% for that test. Yes its not double obviously. Back in July, Nordic...
  32. R

    Q3 2008 JPR marketshare results

    Notebook sales up 40% in one quarter is not radical?...soon your kids will start getting beaten up at school if they don't have a netbook(they mightn't have much computing power, but it sure hurts to get hit by one) Good timing for nvidia with their integrated notebook chipset. FWIW...
  33. R

    ATI RV870

    What i meant to say is doubling the ROP power would be useful if the chip was gddr5 only, but a waste of transistors if they are planning mainly to sell it with gddr3. ie the gddr3 would be the bottleneck They did it with the RV770, they are double power compared to RV670(see rage3d link...
  34. R

    ATI RV870

    Ok, working quickly from from the wikipedia page for memory bandwidth: RV670 3850 GDDR3 53 3870 GDDR3 57.6 3870 GDDR4 72 RV770 (which had ROPs with doubled power according to Rage3d) 4830 GDDR3 57.6 4850 GDDR3 63.6 4870 GDDR5 115.2 According to Nordic Hardware...
  35. R

    4850x2

    That product description is a 4870x2(and is priced like one). Over the last few months the australian dollar has crashed from about US95c down to US65c, so today the card is about US$600. For australia, if you want to look for cards/prices the best pricing engine is static ice. There is no...
  36. R

    ATI RV870

    What i am trying to say is that the memory interface is the defining characteristic, the rop count is just a secondary thing. Years of 16 rops/256bit is an artefact of the gddr3 interface and the speed it runs at. The RV770 memory controller added compatability for gddr5 but does not appear...
  37. R

    ATI RV870

    Currently the RV770 uses the same ROP behind the memory controller for both gddr3 and gddr5. As gddr5 provides roughly twice the data as gddr3 the ROP cant be tuned perfectly to both(unless it runs a double the clock for gddr5 case) it must be specialized for one or the other surely? I am...
  38. R

    ATI RV870

    Are you sure the ROP count is directly tied to the memory controllers? I thought they kept it at 16 for DDR3 cause DDR3 was the limiting factor, with DDR5 running at double speed they could double the ROPs to 32 if they wanted without increasing the size of the memory interface. This would...
  39. R

    ATI RV870

    I dont think a 384bit memory bus is viable unless the chip is alot bigger and more expensive ie 400mm2 or something. GDDR5 i think requires more pins than DDR3 as well at the smaller process feeding all the power in is going to be difficult. The sideport thingy also requires a bunch of pins as...
  40. R

    Not all 4830s the same (some have less shaders)

    I dont think that will be possible, its disabled on the asic. The story about the fault apparently is a test 4850 bios used to emulate the 4830 got onto the reviewers card it disabled 2 clusters from the bios which didnt line up with the 2 clusters disabled on the chip, leaving one net extra...