MSI 280x Gaming 3G Safe Sustained Temperature

Question

Hi,
  I'm new to overclocking and underclocking but since I have 2 cards that are close together and am running them at 100% usage 24/7 I've been unable to use normal clock speeds and have the card very underclocked (ex, 800,740). My question is how hot can I run these on a 24/7 basis without damaging the cards?  I realize this may be a relative question but these are new cards and I expect a 5 year lifespan.  From other forums I've been optimizing with a maximum sustained temperature of 89C, but I am trying to attain maximum performance. To do that I'll need to know how hot I can let them run, non-stop for 5 years.
The crossfire bridge is not installed, and I'm using the Z87-GD65 board with an intel i5-4670 (1100 Watt power supply).
Cards: R9 280X Gaming 3G BF4 s/n 602-V277-38SB1401027047
 

Answer 1

Quote from: rritoch on 27-February-14, 16:18:40
Thank you, that is a good answer to my original question. These 280x's though would overheat within minutes, to be exact when I returned them the computer shop tested it and it took almost exactly 2 minutes to reach 90C at manufacturer settings. The 280x is rated at 1000mhz and being able to run them non-stop at 800mhz would be reasonable but I couldn't even run them non-stop at 600mhz without them overheating.  As for their intended purpose of gaming, including video and audio feeds with some voice recognition, object tracking, and object identification would probably be more than enough to create the same load that mining causes. I also intend to test neural networks and genetic algorithms on them, all things that could be included in a gameing engine for the purpose of enhancing difficulty based on player capability. If the cards can't handle a full load, than it's false advertising and the fully-loaded specs should be available to both programmers like myself, and consumers that are being mislead.
These cards are now in the hands of MSI, so I'll know more when I have normal 280x cards.  With this information, knowing the safe temperatures and expected operating ranges I should be able to utilize these cards, but stop lying to consumers. The specs on the box are clearly their peak performance, and not the performance they can actually operate at.
Operating temp: 75C-79C
Core Speed: 800mhz
The one spec you left out is the memory speed? Can I expect this also needs to be toned down 20%?
Memory Speed: 1200mhz ???
memory speed is fine! at 1200MHz (4.8GHz effective as its 1200MHz x4) constantly that the chips are rated at they should function 8-12years without issue unless there defective and Mining also will not push the Chips very hard only the GPU as its mainly just plain data that being crunched (would expect no more then a 512MB load out of 2GB on each card or below) so there unstressed anyway (VRAM is only going to be stressed by games or benchmarks that have huge Visual Data sets of textures designed to cause a heavy visual load to be calculated) so the VRAM frequancy is nothing as the chips are usually rated for the frequancy 24/7 for 10+ years at a typical clock speed so do not worry about that but if your still concerned drop it to 1000MHz (4GHz effective) but it will make no difference really overall!
Quote from: rritoch on 27-February-14, 16:18:40
I'm also curios if these actual "safe" operational conditions are available via OpenCL, or if it is going to report 1000mhz. If so than any application written to utilize GPU (such as matlab) is likely  going to burn up the card if the user doesn't know that it needs to be underclocked.
it will clock the GPU under load to its set High Frequency (dependant on what you set the speed at) under a heavy load as its not the Application doing so its the Cards VBIOS detecting a load and ramping up the Clocks to make the load be calculated faster as that is what people expect in a game to happen (Like I said before 24/7 operation over a long time frame is the realm of Professional cards that have been validated for a heavy work load not what these are which is Consumer Grade items that are many times cheaper and are disposable commodities generally as gamers usually upgrade them quite regular) so underclocking a consumer Grade item like your R9-280x is a good idea as your using them outside of there Design standard that is much lower a load then your throwing at the GPU and applications unless they are designed to alter the VBIOS Frequency limit then it will just use the default setting that is held in the driver table! <--- if your doing 24/7 calculation loading on a card of this level and you do not know to underclock and undervolt them then you are probably not meant to be doing so as you have too low a understanding of what its doing so really people who do that and they kill it are on there own as there doing something they have such little knowledge about that there asking for trouble! (you yourself seem to be fairly Knowledgeable to this which is why i'm not being very harsh with you because you do understand that if its overheating and getting near the limits to back it off to try and stop it breaking!)
TIP: when you get your new cards back from RMA let then sit under a Gaming load for a while (30-100% for 3-4 hours with 5 mins between each 30 mins max time frame loads "Maybe use a benchmark") to cure the TIM on the card to help its performance long term as you may have Baked the TIM to fast last time (dried it out too much) leading to the High temps by just throwing it into ultra high load mining constantly at it before it had had a chance to bed in and reach its correct thermal conduction rate Ruining it!