AMD: Southern Islands (7*** series) Speculation/ Rumour Thread

Discussion in 'Architecture and Products' started by UniversalTruth, Dec 17, 2010.

  1. ECH

    ECH
    Regular

    Joined:
    May 24, 2007
    Messages:
    682
    Likes Received:
    7
    #4161 ECH, Nov 13, 2012
    Last edited by a moderator: Nov 13, 2012
  2. lanek

    Veteran

    Joined:
    Mar 7, 2012
    Messages:
    2,469
    Likes Received:
    315
    Location:
    Switzerland
    lol i have see that too.. funny .

    About AMD cards, well 600 series look to perform a bit better, but as you can even play at max setting with 2560x1600 resolution and full AA with any cards ( midrange as the 660 or 7850 ), no worry to have on this point..

    The game have some bug ( on some sp mission ) ( a bug with sight when you look down who decrease the fps /2 ) and some collision bug with both AMd or Nvidia cards, so it is not really easy to know if it is the game or can be driver related.

    I dont play it, so i cant tell about my experience, just report the few things i have see .
     
  3. caveman-jim

    Regular

    Joined:
    Sep 19, 2005
    Messages:
    305
    Likes Received:
    0
    Location:
    Austin, TX
    The blower style has been criticized for years by enthusiasts, people wait for the custom cooler designs for the dual and triple fans for noise and improved cooling. I'm glad to see that lesson is sinking in, hopefully the blower fans will go away across the entire product stack, if there is another one.

    AMD has a reputation of poor professional driver performance/stability/support, all the feedback I'm hearing is you need former AMD programmers on your team to get things working like they should... I think there is quite a few of them right now though.
     
  4. ECH

    ECH
    Regular

    Joined:
    May 24, 2007
    Messages:
    682
    Likes Received:
    7
    Thanks for the info.
     
  5. CarstenS

    Veteran Subscriber

    Joined:
    May 31, 2002
    Messages:
    4,798
    Likes Received:
    2,056
    Location:
    Germany
    That's why I wrote:
    „The one area where the FirePro shines the most - raw FLOPS that is - is already won by a healthy margin. And other reasons such as the ones hinted at by AlexV, wouldn't be disregarded by potential customers if only there was a 300 watt power envelope to the card.“ [IOW: What is it that makes the FirePro an attractive value proposition?]

    That may or may not be the case, we simply do not have the necessary data to prove or unprove it, but that could be one line of reasoning AMDs product team was following along.
     
  6. Gipsel

    Veteran

    Joined:
    Jan 4, 2010
    Messages:
    1,620
    Likes Received:
    264
    Location:
    Hamburg, Germany
    The point is: 375W makes it not only unattractive, it disqualifies it for a some(/most?) areas. 300W not so much as indicated by Intel's Knight Corner cards which come in 225W as well as 300W flavours.
    And as said already, one could not only boast the highest absolute flop numbers (SP as well as DP), the performance/Watt figures would be basically on the same level as with K10 and K20 (some in front, some behind, but generally comparable) with the added benefit that you get the blended advantages of both, the K10 (high SP performance, high SP perf/Watt, high memory bandwidth) and the ones from K20 (high DP performance, high DP perf/Watt, real ECC [also on the internal SRAMs, not only on the external memory]) in a single card. It would be simply more competitive against the Kepler Teslas.

    The reservations of AlexV regarding the software stack are orthogonal to the question if a company has the best hardware lineup they can reasonably achieve with the ASIC designs at their hands or not. The software either disqualifies them either way or the customer can work with it. That's a different construction site (I hope that German saying translate halfway to English, won't be a problem for Carsten though ;)).
     
    #4166 Gipsel, Nov 14, 2012
    Last edited by a moderator: Nov 14, 2012
  7. gongo

    Regular

    Joined:
    Jan 26, 2008
    Messages:
    582
    Likes Received:
    12
    No! I hope blower stays! works great for positive pressure exhaust....and in thermal reviews by behardware, blower has cooler surrounding areas of the motherboard.....

    Yes it is blowing loud on load and overclocked...but staying within TDP...it is acceptable noise....maybe AMD can optimise the design more...such as using EVGA high airflow brackets....a different type of bearings for the blower.....better dampening materials....larger blower....?

    The blower is around for years...Nvidia uses it too...blower stays! bro!
     
  8. Bouncing Zabaglione Bros.

    Legend

    Joined:
    Jun 24, 2003
    Messages:
    6,363
    Likes Received:
    82
    I like the blower too. Hot air is blown out of the case, instead of adding to the thermal load inside the case and making all your other fans have to work faster (and louder). Under normal gaming load, I've also found it to be quieter than many third party HSF designs.
     
  9. karlotta

    karlotta pifft
    Veteran

    Joined:
    Jun 7, 2003
    Messages:
    1,292
    Likes Received:
    10
    Location:
    oregon
    Blowers blow. I used to think that way, warm air out the case, Presure BS etc... Heck i even have a "silentcase" setup. But the twinfrozer etall are SOOO much better, cooling, noise and dust issues. I have no issue with a ~5c more temp in my case.. Blowers are lame shopvacs and in verywarm/hot ambeint rooms they just scream. Where the good aftermarket dual fans are still less sound than PSU, or some case fans.Never ever will buy a blower. All passive or a nice dual.
     
  10. Malo

    Malo Yak Mechanicum
    Legend Veteran Subscriber

    Joined:
    Feb 9, 2002
    Messages:
    7,042
    Likes Received:
    3,114
    Location:
    Pennsylvania
    I was the same way with blowers. My latest gpu is a msi 7870 twinfrozr and im so impressed with its cooling efficiency. My case amboents arent an issue either with it.
     
  11. Dave Baumann

    Dave Baumann Gamerscore Wh...
    Moderator Legend

    Joined:
    Jan 29, 2002
    Messages:
    14,079
    Likes Received:
    648
    Location:
    O Canada!
  12. CarstenS

    Veteran Subscriber

    Joined:
    May 31, 2002
    Messages:
    4,798
    Likes Received:
    2,056
    Location:
    Germany
    Ah, that's good to see. That seems to be the "problem system" which i mentioned here: http://forum.beyond3d.com/showpost.php?p=1679039&postcount=431 - seemingly it's stats were updated in the meantime and not only power consumption was added but also Rmax, which was at 200ish-something yesterday and reason for my worrying over there.

    I am not so sure - since most servers are not really plug'n'play like your average desktop PC but are qualified by the individual vendors. Granted, the norm seems to be to have parts with 225 watts, Nvidia breaking the upper ceiling as well as Intel sometimes.

    AMD already advertises S10000 as one solution to many problems - including that you basically can kill two birds with one stone.


    That'd be a different kettle of fish, right. And one I am not to well accustomed to dip my hands in. ;)
     
    #4172 CarstenS, Nov 14, 2012
    Last edited by a moderator: Nov 14, 2012
  13. Gipsel

    Veteran

    Joined:
    Jan 4, 2010
    Messages:
    1,620
    Likes Received:
    264
    Location:
    Hamburg, Germany
    Coincidentally, that is the same system as mentioned already in in the GK110 thread.
    But they corrected the performance up (421.2 TFLOP/s now, peak stays at 1098 TFLOP/s) and included the power usage (a modest 179 kW).
    As assumed over there, the number of FirePro cards is quite low (420). Who knows how they counted the cores (38400)? Is the given peak (~1,1 PFLOPs) even correct?
    The 420 FirePro cards have a peak of 620.9 TFLOP/s. As there is probably 1 card per node and the Xeon E5 supports 2 CPUs (16 cores) per node this would give us 6720 CPU cores @ 2 GHz with 8 FLOPs per cycle and core equaling only 107.5 TFLOP/s for the CPUs and 728.4 TFLOP/s for the whole system (and an efficiency of ~58%, what is believable). The core count would also be somehow close if one counts 56 cores (CUs) per FirePro and adds it to the CPU cores. But 30240 cores are still too short of the claimed 38400 to explain the difference with a few IO nodes. And the power consumption of 179.8 kW for 420 nodes each with a FirePro S10000 in it sounds believable (428W for each node, quite okay for a node containing a 375W TDP card plus dual 8 core CPUs plus some overhead for the network infrastructure; I wonder what would have happened with fully enabled Tahitis at 725 MHz and a slightly further reduced voltage :roll:), but not if one would add so many CPUs to account for the balance to the 1.1 TFLOP/s claimed peak.

    Anyway, with the recent update of the green500 list Titan isn't in the #1 spot as claimed (it's now 3rd). Second is this Xeon+FirePro cluster and the first spot takes a small intel only system with Knights Corner cards.

    Edit:
    Yes. But comparing it to K10 or K20 individually, it doesn't look that good anymore. So it's not compelling to someone, which has a clear preference for DP or SP performance and doesn't value the performance with the other precision too much. Imagine a 25% higher performance per Watt (at least from the spec side, it could be that the 375W of the S10000 are never really used and it just happens to be slightly over 300W in any realistic scenario) and how this would change the comparison. I said it before, one would sit just 3% behind in SP performance/Watt (with higher absolute performance) against K10 while being 25% in front comparing it to K20. With DP performance it's no competition against K10 and the performance/W would be just 5% behind a K20 (while providing higher absolute performance and bandwidth [just counting the upsides here]). So for people interested in SP performance it's a wash against K10, for DP against K20 and for people interested in some mix one would have had an actual advantage.
    Either way, a lower TDP couldn't have been a bad thing. ;)
     
    #4173 Gipsel, Nov 14, 2012
    Last edited by a moderator: Nov 14, 2012
  14. Dave Baumann

    Dave Baumann Gamerscore Wh...
    Moderator Legend

    Joined:
    Jan 29, 2002
    Messages:
    14,079
    Likes Received:
    648
    Location:
    O Canada!
    Evidently the actual power draw of S10000 in this platform is more like 330W.
     
  15. ams

    ams
    Regular

    Joined:
    Jul 14, 2012
    Messages:
    914
    Likes Received:
    0
    Did you notice that the Linpack performance/watt data for the number one "green" system with Xeon Phi does not match the actual data in the Top 500 page? According to the Top 500 website ( http://www.top500.org/system/177997 ), power consumption is 45.11 kW (vs. 44.89 kW on the green 500 website), and Linpack performance is 110.5 TFLOPS/s (vs. 112.2 TFLOPS/s on the green 500 website). So the actual Linpack performance/watt on the Top 500 website is 2.44957 GFLOPS/w (vs. 2.49944 GFLOPS/w on the green 500 website). Minor difference, but still notable. The number two "green" system with Firepro S10000 and the number three "green" system with K20X do match the data on the Top 500 website.

    On a side note, the Xeon Phi equipped supercomputer used by NASA for climate simulation has a Linpack performance/watt of 1.93553 GFLOPS/w, which is quite a bit behind the smaller Xeon Phi equipped supercomputer used by the National Institute for Computation Sciences/Univerisity of Tennessee:

    http://www.top500.org/system/177993

    Is this a sign that Linpack performance/watt for the Xeon + Xeon Phi system will actually go down as the supercomputer installation becomes larger and larger (because performance scaling diminishes when using more and more cores)?
     
    #4175 ams, Nov 14, 2012
    Last edited by a moderator: Nov 14, 2012
  16. Gipsel

    Veteran

    Joined:
    Jan 4, 2010
    Messages:
    1,620
    Likes Received:
    264
    Location:
    Hamburg, Germany
    I think, it is less during the benchmark run. I doubt you can subtract the CPUs TDP (95W) from the total power consumption per CPU+GPU pair (428W). ;)
    First, the 179 kW total power consumption are of course measured at the wall (not literally), which means with 90% power supply efficiency it would be already only ~385W per CPU+GPU pair. Then you have of course the RAM consuming a few Watts, the mainboard and not to forget the network infrastructure (Infiniband). I think it's safe to say that the actual consumption of a CPU + GPU pair was below 350W during the benchmark run (bounding the FirePro S10000 at or below 300W with that workload).

    Edit:
    And regarding the strange specs with the core counts and so on, the Frankfurt university gives more accurate data.
    So there are 210 dual CPU nodes, each containing two S10000 cards. That means one has just 3360 CPU cores delivering a theoretical peak of 53.76 TFLOP/s for a total theoretical peak of just 674.7 TFLOP/s (and not the 1.1 PFLOP/s as stated on the top500.org). That pins the efficiency at 62.4%.

    The whole thing fits in exactly ten 42U standard racks, each node takes 2U and would also allow to put four S9000 cards in it (or four K20 or any other card with a 225W TDP for that matter).
     
    #4176 Gipsel, Nov 14, 2012
    Last edited by a moderator: Nov 15, 2012
  17. ECH

    ECH
    Regular

    Joined:
    May 24, 2007
    Messages:
    682
    Likes Received:
    7
    Can anyone explain what this win7 update (Pre-Release) is?
    http://www.microsoft.com/en-us/download/details.aspx?id=35661
    At 1st glance it looks like dx11.1 but I don't think that's the case at all. I wondering if it will benefit the 7000 series or it's driver's in any way?
    The above link is dated the 5th. Im not sure if that's the right link or final version.
    Win7 SP1 platform update: http://support.microsoft.com/kb/2670838

    Edit:
    I also found this:
    http://www.neowin.net/news/microsoft-gives-in-adds-some-directx-111-features-to-windows-7
     
    #4177 ECH, Nov 15, 2012
    Last edited by a moderator: Nov 15, 2012
  18. Alexko

    Veteran Subscriber

    Joined:
    Aug 31, 2009
    Messages:
    4,496
    Likes Received:
    910
    I'm confused, the Green 500 website says "King Abdulaziz City for Science and Technology", which is in Saudi Arabia.
     
  19. Zaphod

    Zaphod Remember
    Veteran

    Joined:
    Aug 26, 2003
    Messages:
    2,204
    Likes Received:
    103
    The linked article states:
     
  20. ECH

    ECH
    Regular

    Joined:
    May 24, 2007
    Messages:
    682
    Likes Received:
    7
    #4180 ECH, Nov 15, 2012
    Last edited by a moderator: Nov 16, 2012
Loading...

Share This Page

  • About Us

    Beyond3D has been around for over a decade and prides itself on being the best place on the web for in-depth, technically-driven discussion and analysis of 3D graphics hardware. If you love pixels and transistors, you've come to the right place!

    Beyond3D is proudly published by GPU Tools Ltd.
Loading...