Showing Results: Most Recent
Pros: works well in gaming mode
Cons: 2d mode and idle prone to black screen
Other Thoughts: Follow up on my previous MOST HELPFUL review about this card titled : "A bit finicky, if you get black screen read this."
after about 3~4 months of ownership, the card is starting to black out again. previously lowering the clock speed for 2d mode did the trick. but it seems either the memory, or the memory controller is dying of a slow death or something else is going on, because i am starting to run into black screen more often now even with the lowered 2d clock speed of 800mhz core and 1250mhz memory. this is really sad, consider games run perfectly fine. it's while idling on desktop or watching a video.
tried swapping out the video card for my old trusted HD7950 and never see a single black screen.
i am going to attempt to RMA this card and see if Sapphire will send me back the original broken one, or someone's else's used card, or do what's right on sending me back a new functional replacement. i've heard Sapphire's support turn around is very slow, thus far i haven't received any response yet.
i shall update again.
Pros: Can handle 3440x1440 res at 38~45fps in Witcher 3 Ultra setting with hairworks tessellation at 8x. GTA v in the same resolution with mostly highest setting at low to mid 40s fps.
some of the best VRM cooling i've seen. Kudos to the heatsink design team who did a fabulous job in cooling this monster, and not just the core, the VRM stays as cool as the core temp as well.
Cons: the infamous black screen ah yes.
if your game runs fine, but encounter black screen when just sitting idle on the desktop. read on!.
Provided, if you have already ruled out the power issue, lack of power due to a weak power supply, driver problem, and other possible issues, read on.
Long story short, after long research and testing, here is MAINLY the root of the black screen problem: Memory speed VS Core voltage. in this generation of GCN 1.1, apparently the memory operating frequency is closely tied to the core voltage. you core speed can be at stock, but the moment you start raise the memory frequency, core voltage goes with it. hence the core and memory both share the same voltage. this is problematic. whether it's lack of testing on AMD's side, or vendor configuration. my conclusion is that, when the vendor starts to come out with overclocked models, their overclocked speed for core and memory, and their relative voltage setting is being tested under full load stress in a 3D environment, meaning the voltage is pumping 100% along with the overclocked core speed and overclocked memory frequency all working 100% (in this case 1020mhz core, and 1350mhz memory). THE PROBLEM BEGINS HERE. because vendors put out these overclocked cards, and bin/test/stress test them under the full clock speed full voltage environment, most of the time, the cards pass the test under the full stress. HOWEVER GCN 1.1's new fancy Dynamic clock modulation (basically a new generation of power saving feature, you can thank AMD for the headache), clock speeds for core and memory dynamically scales down accordingly to load, but worst, so does the voltage. So due to the new generation of fancy power saving feature and dynamic clock/voltage modulation. a Card can pass with flying color in gaming, but vendor fail to test the card in idle because in idle, the voltage drops dramatically!!. what you have is an overclocked memory at 1350, that shares the operating voltage with core. and when the load is low (idle), the core voltage drops dramatically, and because the core voltage is shared with memory. the overclocked memory is now running at an unstable state during idle environment because the core voltage dynamically drops to a crazy low level. this probably wouldn't cause so much problem if the vendors would realize this, and would test the products at an idle state rather than just throw the cards a couple gaming benchmarks and watch it pass. the problem is not full-load, it's idle, and the fact that at idle, voltage is dynamically scaled down dramatically at an overclocked memory speed frequency that's higher than stated 1250mhz.
Other Thoughts: Solution:
you can either raise the core voltage, how much is different in each person's case, but in most cases by +25mv should be enough, or +50mv should take care of it all.
OR you could down clock the memory speed to AMD designed 1250mhz.
again the problem is idle. so we need to ensure there is enough juice being pumped to the memory during idle, OR reduce the memory frequency at idle while keeping the voltage the same.
my suggestion for vendors, work with AMD to understand the architecture of the new generation of cards. don't just put out an overclocked model with only a brief test in full load. If you had understand the architecture, and the new power saving feature, which has dynamic voltage/clock built in, you would understand that the 290 behaves very differently in Idle vs Full-load.
NOT everyone will encounter this problem. if your particular card has a very strong memory, or very good memory controller, chances are there's tons of headroom there even with dynamic voltage down scale, there's enough headroom for it to not black screen. but if your particular card doesn't have a very strong memory, or memory controller, you will likely run into this type of issue.
Pros: fast chip when it works. lots of room for overclock; however it's inherently held back by its design.
Cons: believe it or not, i actually got a faulty chip the first time. the first chip didn't pass stress test and ran HOT. until this day, intel tech support will still deny the usage of TIM underneath the Heatspreader, which inherently limits what this chip's ultimate potential. also due to lack of competition, intel can charge a high price and there's nothing you can do about it, even tho this chip is already almost two years old tech, and only $10~$20 cheaper than its successor, at the very most if you're lucky.
Other Thoughts: the lack of competition has intel cutting corners. gone are the days when cores are properly soldered underneath the heatspreader. Intel started cutting corners by using cheap under performing thermal interface pastes to connect the cores under the heatspreaders, instead of soldering, while still maintaining the high price tag g to maximize their profit margin. the quality control also seems to have gone down over the years. i actually received a faulty chip the first time, and had to RMA. over the course of 15 years building computers, it was actually my very first faulty consumer chip. don't get me wrong, the i7 is fast, in terms of faster than its competition. but you're paying twice the money for them to cut corners in the manufacturing process (by using TIM instead of Solder). until this day, Intel tech support will still deny the usage of TIM for its consumer IB chip despite the fact that it has already been proven. i can only hope that the future will bring some much needed competition that'll give Intel run for its money.READ FULL REVIEW
Some manufacturers place restrictions on how details of their products may be communicated.