Page 106 of 158 FirstFirst ... 65696104105106107108116156 ... LastLast
Results 1,051 to 1,060 of 1571

Thread: Kepler performance

  1. #1051
    Senior Member
    Join Date
    Nov 2011
    Posts
    2,594
    Well, we can make a solid guess as to how many more 7970s are being moved compared to GTX 680s on newegg if we assume that the same percentage of users who buy an AMD write a review as the users who buy NV.

    7970 has 292 reviews in 103 days.

    GTX 680 has 185 reviews in about 30 days since the march 22nd release.

    So, 7970 is averaging 2.8 reviews a day and GTX 680 is averaging 6 reviews a day.

    The big assumption is that the same percentage of people who bought cards wrote reviews for them. If that's true, NV is selling GTX 680 significantly faster than AMD is selling 7970s.

    However, this doesn't account for 7950s. There are 164 7950 reviews in 81 days. That makes an average of 2 cards per day.

    If we consider 7970 and 7950 sales, AMD is still averaging 5 reviews per day (assuming 90 days). So, even with NV's awful availability, it looks like GTX 680 is selling faster than 7970.

    The biggest criticism of this method is that NV fans tend to be a lot more enthusiastic about their products so they're a lot more likely to write a review. Considering GTX 680 hasn't been around very long and it has this many reviews is impressive, as much as I don't like to admit it.

  2. #1052
    Senior Member
    Join Date
    Nov 2011
    Posts
    2,594
    Quote Originally Posted by __shared__ View Post
    Correct on perf/watt.



    Given the expected bandwidth increase I would expect about a 50% bump, but then again I'm not an expert when it comes to gaming performance. Maybe it's more driven by texture units and such which likely won't increase much for the 7B monster.

    Anyways the 680 is very bandwidth constrained, you came to that conclusion yourself in your first in-depth article on the topic.
    **personal comments removed** e a 50% increase in performance.

    Pitcairn is the GPGPU castrated AMD GPU. It has 2.8 billion transistors.

    Tahiti is the full GPGPU AMD GPU. it has 4.5 billion transistors.

    Tahiti is 60% bigger.

    http://www.bit-tech.net/hardware/201...0-2gb-review/5

    Looking at the BF3 benchmark, Tahiti is only 20% faster. Transistor count clearly doesn't scale linearly with performance.

    If this new big kepler is twice as big, you're only going to see 30-40% performance increase in games. When adding GPGPU transistors, you see a 30% of the performance gain compared to how many transistors you've added.

    So, if this new kepler is 100% bigger, you're going to see it be 30% faster. It's simple math. I'd expect it to be even lower than that, because NV castrated GPGPU on their game chip a lot harder than AMD did theirs. If I saw 20% increase in games I wouldn't be surprised. If I saw 10% increase I would, and if I saw 35%+ I would.

  3. #1053
    8-bit overflow
    Join Date
    Aug 2009
    Posts
    282
    Quote Originally Posted by sdlvx View Post
    Well, we can make a solid guess as to how many more 7970s are being moved compared to GTX 680s on newegg if we assume that the same percentage of users who buy an AMD write a review as the users who buy NV.

    7970 has 292 reviews in 103 days.

    GTX 680 has 185 reviews in about 30 days since the march 22nd release.

    So, 7970 is averaging 2.8 reviews a day and GTX 680 is averaging 6 reviews a day.

    The big assumption is that the same percentage of people who bought cards wrote reviews for them. If that's true, NV is selling GTX 680 significantly faster than AMD is selling 7970s.

    However, this doesn't account for 7950s. There are 164 7950 reviews in 81 days. That makes an average of 2 cards per day.

    If we consider 7970 and 7950 sales, AMD is still averaging 5 reviews per day (assuming 90 days). So, even with NV's awful availability, it looks like GTX 680 is selling faster than 7970.

    The biggest criticism of this method is that NV fans tend to be a lot more enthusiastic about their products so they're a lot more likely to write a review. Considering GTX 680 hasn't been around very long and it has this many reviews is impressive, as much as I don't like to admit it.
    afair, newegg allows anyone to post a review, and also, it shows which reviews are from ppl who actually bought the card... if ur counting all reviews and not only reviews from ppl who bought the card, u're only measuring fanboyism, not sales

  4. #1054
    Banned
    Join Date
    Jan 2012
    Posts
    185
    Quote Originally Posted by sdlvx View Post
    I'm sorry shared, you're nuts if you think you're going to see a 50% increase in performance.

    Pitcairn is the GPGPU castrated AMD GPU. It has 2.8 billion transistors.

    Tahiti is the full GPGPU AMD GPU. it has 4.5 billion transistors.

    Tahiti is 60% bigger.

    http://www.bit-tech.net/hardware/201...0-2gb-review/5

    Looking at the BF3 benchmark, Tahiti is only 20% faster. Transistor count clearly doesn't scale linearly with performance.

    If this new big kepler is twice as big, you're only going to see 30-40% performance increase in games. When adding GPGPU transistors, you see a 30% of the performance gain compared to how many transistors you've added.

    So, if this new kepler is 100% bigger, you're going to see it be 30% faster. It's simple math. I'd expect it to be even lower than that, because NV castrated GPGPU on their game chip a lot harder than AMD did theirs. If I saw 20% increase in games I wouldn't be surprised. If I saw 10% increase I would, and if I saw 35%+ I would.
    Your looking at die size while I'm mentioning the actual bottlenecks of the card, mainly bandwidth.

  5. #1055
    Senior Member
    Join Date
    Nov 2011
    Posts
    2,594
    Quote Originally Posted by __shared__ View Post
    Your looking at die size while I'm mentioning the actual bottlenecks of the card, mainly bandwidth.
    I don't see bandwidth as a problem on GTX 680.

    http://www.hardocp.com/article/2012/...card_review/10

    It performs just the same as 7970. Granted 7970 is the better card because it's at 925mhz and the GTX 680 is at 1056mhz+. If it were bandwidth limited, it would be a lot worse looking for GTX 680.

  6. #1056
    Nvidia used the same trick with Fermi, they basically loaded them all on the big etailers. They do it with Newegg and Scan every time.

    If you want the real numbers you need to look outside of their top choices. As I mentioned earlier, Ebuyer has no 680's at all.

    0 stock, 4 available for pre-order http://www.ebuyer.com/store/Componen...GTX-680-Series

    ~180 7970's in stock - http://www.ebuyer.com/search?q=7970&x=0&y=0

    ~ 115 7950's in stock - http://www.ebuyer.com/search?page=2&...vancy&limit=10

    Ebuyer is a big etailer but they aren't Nvidia's preferred one.

    I'm pretty sure that this will be repeated anywhere on sites that give stock totals.

  7. #1057
    Banned
    Join Date
    Jan 2012
    Posts
    185
    No the 680 would have simply beaten the 7970 handidly if it wasnt constrained by bandwidth.

    Stating the obvious issue here:
    Twice the theoretical performance of the 580 but at the same bandwidth.

    you're only going to see 30-40% performance increase in games.
    So you're willing to go as high as 40% and still call me insane for saying 50, seems like your bashing yourself here

  8. #1058
    Senior Member
    Join Date
    Sep 2009
    Posts
    3,026
    Quote Originally Posted by sdlvx View Post
    Well, we can make a solid guess as to how many more 7970s are being moved compared to GTX 680s on newegg if we assume that the same percentage of users who buy an AMD write a review as the users who buy NV.

    7970 has 292 reviews in 103 days.

    GTX 680 has 185 reviews in about 30 days since the march 22nd release.

    So, 7970 is averaging 2.8 reviews a day and GTX 680 is averaging 6 reviews a day.

    The big assumption is that the same percentage of people who bought cards wrote reviews for them. If that's true, NV is selling GTX 680 significantly faster than AMD is selling 7970s.

    However, this doesn't account for 7950s. There are 164 7950 reviews in 81 days. That makes an average of 2 cards per day.

    If we consider 7970 and 7950 sales, AMD is still averaging 5 reviews per day (assuming 90 days). So, even with NV's awful availability, it looks like GTX 680 is selling faster than 7970.

    The biggest criticism of this method is that NV fans tend to be a lot more enthusiastic about their products so they're a lot more likely to write a review. Considering GTX 680 hasn't been around very long and it has this many reviews is impressive, as much as I don't like to admit it.
    The newegg reviews can't really be considered a reliable measurement. Products get added and removed regularly affecting the math. We also can't assume that every single sale results in a review. The only measure that we can use would be minimum sales total for all product listed. In essence, there has to be at least one sale per review.
    When people tell me "You're going to regret that in the morning" I sleep in till noon because I'm a problem solver!

  9. #1059
    8-bit overflow
    Join Date
    Aug 2009
    Posts
    282
    Quote Originally Posted by distinctively View Post
    The newegg reviews can't really be considered a reliable measurement. Products get added and removed regularly affecting the math. We also can't assume that every single sale results in a review. The only measure that we can use would be minimum sales total for all product listed. In essence, there has to be at least one sale per review.
    I'll repeat again, review != sale... anyone can write a review... those who bought the card from newegg has a "verified owner" under their names, and you can clearly note there are few reviews without this notice

  10. #1060
    Quote Originally Posted by charlie View Post
    They can't get them to yield, it is obvious from the clocks and their statements. On top of that, the wafer shortage means they have to pick between big low yielding parts for a halo, or smaller better yielding cards to meet pre-existing commitments. It doesn't take a genius to figure out which one they picked.

    -Charlie
    What do you mean by the clocks? Are you saying they are having a hard time yielding 1006MHz GK104s? I don't know if I can believe that. I know a couple guys who have stock clocked 680s that overclock with final boost speeds over 1300MHz and stock Vs. Plus looking around the web and forums tells much the same story....most 680s can get gpu boost clocks 1250-1300MHZ easy on stock Vs when overclocked.

    I can buy that they are having a miserable time yielding a full 8gpc chip and in due time will see the GTX 670s floood market. It will be interesting to see if they can keep a harvested GK104 on store shelves. I'm now wondering if the double GK104 board may contain harvested 670 chips since the shortage of full spec GK104. In that case, AMD may be able to beat that since they should use full double Tahiti chips, albeit maybe not at full clock speed of the 7970.....

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •  
WordPress Appliance - Powered by TurnKey Linux