If 25 FPS is playable, then you can probably manage with 22 FPS as well. And of course, a difference of 22 and 25 FPS is larger (13.6%) than 60 and 67 FPS (11.7%), plus if you run with VSYNC then 67 FPS isn't really important -- it's just the goal of getting over 60.
Personally, most games are not "playable" until at least 30 FPS for me -- or at least they're not very enjoyable at sub-30 FPS. 28/29 FPS, sure, you could live with that, but 25 is too far down the performance ladder.
Looks like you pretty much agree with the concept, just not the details. Had the example been going from 27 to 30 fps, would that have been more convincing?
I'd say more like 25 to 30 is what would be desirable to take a game from choppy to truly playable (though obviously there are exceptions). There's also the question of minimum frame rates -- if the game averages 35 FPS over a large portion of the game but there are areas (or battles) where it consistently drops to 20-25 FPS, then really you'd want to use settings that can get the minimum FPS up closer to 30.
The real problem is that I don't know that I've ever seen a driver update provide a huge increase in performance for an APU. The "up to 30% faster" sort of claims are often for the high-end GPUs, and even then a lot of the time it's more like 15% (e.g. half of what was claimed) with a few edge cases seeing larger gains.
Anyway, I'm curious to know if the drivers helped at all for people with a Kaveri APU in the games mentioned. Let me know! :-)
I don't see a single point in APU gaming AT ALL - at least, on desktops. You can get, say, a used HD 7950 for around $120 these days. If somebody wants to game, is it such a big deal? It's like a price of two AAA games, or, like, say, refueling a car fuel tank... APUs are for office PCs or, say, for kids or grandma's/grandpa's PCs.
I'd say that APU's are more for laptops and HTPC which are used for some gaming.
Still, the "get a used HD 7950 for around $120" claim is rather silly. First of all, it may be true in the US, but (contrary to popular opinion of US residents) not everyone lives in the US. Secondly, for some people $120 is a lot of money, and being able to play some games on a budget PC is preferable to not doing it because the next step up is out of your budget.
I said USED card from somebody else's hands, not a new card. I'm not in US either, I'm in Canada, for your reference - please stop making stupid conclusions without any information. So, if you go to Canadian online flea markets you can get, say, HD 7950 even for $100 and buy it in one day (e.g, I just bought very slightly used R9 290 for $250 two weeks ago, because the seller upgraded to GTX 970, which is cooler, quieter and much less power hungry). "For some people $120 is a lot of money" - then maybe it's better to earn a little money, then game (in this exact sequence, not the other way around, as you imply).
BTW, HD 7950 (Tahiti Pro) has WAY more firepower, than A10-7850K APU - 28 CU instead of just 8, plus higher GPU frequencies, plus 3 GB of dedicated GDDR5. It's a completely different device in terms of performance. Overall graphical performance of HD 7950 is 4 times bigger, if not more. So, gaming on APU is like riding a bike on a highway, when you can easily rent a car for relatively cheap.
I'm not actually trying to be arrogant. Let's put it another way- in terms of price to performance ratio. For $200 you can get new Core i5, and for $100-120 you can find used GTX 760 or HD 7950 or some other comparable discrete gfx card. DDR3 DRAM is the same for both platforms (LGA1150 and FM2+), motherboard cost can also be around the same. 500W PSU is enough for these cards, which is also not a big deal. So, all other things being essentially equal, it comes to $320 for Core i5 + gfx card vs $160 for AMD APU. However, you get 4 times more performance in graphics + way faster CPU everywhere for just additional $160 in case of i5 + dedicated gfx card. To me, the choice is obvious, as I said above - and that's why AMD now has less than 20% CPU (including APU) market share.
HA! The i5 with used 760 is what I originally went after. But we (wife and I) ran into three issues. First, the compact case we wanted didn't allow for the 760 (little larger than the PS4). Going with a case that could increased the cost. Second came down to price. Third came down to the memory we both wanted to use, DDR3-2400.
I didn't want a console, I've never really been a big fan. My wife didn't want a console because it could'nt do anything she wanted to on the work end of things. The middle ground was a compact PC with the budget of a console. That target was $400 - $500. The i5, vid card, and a compact case that fir that just didn't work. It put the price to high.
I installed 32gig of 2400 into a clients computer who did a lot of video/photo editing and the difference in performance for him was pretty amazing. But I had a hell of a time getting his board to recognize it (MSI board z87 1150). We finally got it to work, but it was a chore. I read several builds with the FM2+ chipset having no issues with the ram, and sure enough it was recognized without issue when we build our system. All in all we spent $500. Slightly more than the PS4 and much cheaper than ps4+computer and we saved a bit of space in front of our TV. And it performs GREAT for what it is and what it replaces.
@wolfemane Yes,support of different RAM on different motherboards is not really a sure thing, unfortunately.
In my case, however, I have 32 GB of DDR3-2400 RAM (2 identical 2x8 GB G.SKILL CL11 kits) in my aforementioned home desktop, and it works perfectly from the day I built the machine originally (back in June 2013, Z87 build also) with Core i7-4770 @ 3.7-3.9 GHz, and since September 2014 it continues to be rock solid with Core i7-4790K @ 4.3-4.5 GHz on the same M/B. I, however, have Gigabyte Z87X-UD3H M/B. Works like a charm with 32 GB of DDR3-2400. So, it is not LGA1150 Z87 platform itself which is glitchy in terms of memory support as a whole, it's just the particular M/B and/or RAM which caused these issues.
Yeah it turned out to be the board, had to get MSI on the phone and wound up with slightly different timings to get it to work.
Your system reminds me of my previous desktop. ASUS Sabertooth z77, i7-3770k @ 4.6ghz, 16gig GSkill trident X DDR3-2400, and twin sapphire tri-X OC 290's all on a duel radiator loop. Thing ran solid @ 4.6ghz and the memory worked right out of the package. Wih a little help from a certain OC site I had it at 5.1ghz with a custom fish tank chiller attached.
Unfortunetly I started getting weird black screens. Turns out both video cards had the non hynix (Elpida I believe) memory and was suspect to the problem. After several months of unsuccessful attempts to correct the problem and on the verge of RMA a friend offered a decent amount for the whole system. He didn't care about the vid problems. Couldn't refuse the offer and planned on a similar build to what you have, but decided on other things.
@wolfemane Oh! Then you are also a powerful PC enthusiast; I'm taking my words back :) I read about these possible Elpida VRAM R9 290(X) problems on the web. Luckily for me, my R9 290, despite being reference hot and noisy unit with "turbine" (centrifugal fan), has Hynix VRAM on it, according to GPU-Z and Hawaii Memory Info. Got it recently (around 3 weeks ago) and it seems to be very stable so far in all the loads I tried (no throttling even at slight overclocking and long 3D loads, but the "turbine" goes to 50+% speed and sounds like an airplane - headphones needed). BTW, I had two reference "turbine" HD 4870s in CrossFire in the past (back in 2008-2009) and also had no problems with them (heat and noise is "by design" there, it seems to be not a problem for stability).
Regarding Elpida memory on R9 290(X), people on the web are saying that maybe downclocking could help (but will hurt performance, of course).
That is one solution that has helped some. The best recommendation was using GPU tweak to set/limit 2D core and mem clocks to a fixed speed of 450/500. It decreased the number of black screens, but I still had the problem. The Elpida memory is very sensitive to voltage controls. The best way to put it is from our good friends over at overclock.net.
"This black screen from power transitions occurs because the memory clock jumps up to full speed (1250, etc), while the GPU voltage (which also controls the IMC--this is the problem) is STILL running at idle volts of 0.9xxx). That idle voltage is too low to handle the IMC, which is controlling the memory running at that speed (this is NOT due to defective memory modules btw), and thus the card black screens.
The fix, is to increase the GPU voltage, which ALSO increases the idle voltage to the IMC as well. (previous generation cards used separate voltages for idle clocks and 3D). This Is purely a bios/design problem. The IMC should NEVER be running at 0.9xx when the memory jumps to full speed, or, alternatively, there should be 'lesser' memory speed jumps, for when a light load is put on the GPU, so the memory doesn't jump from 150 MHz to 1250, but instead it could be 500 MHz for a very light load (like when the core goes from 300 to 450 MHz), etc."
I used both MSI AB and Sapphire Trixx to adjust my voltage higher but I still had the same issues (both programs I used when I was overclocking originally). I have a feeling that the black screens are due more to my previous overclocking. I was never really concerned about pushing my system (and I pushed it hard) because I never had any of the tail tale signs of problems (excessive heat, stuttering, visual degradation, etc..). At the time I could have pushed the cards just to hard (I was using seti@home in background mode @ 100% GPU usuage while at the same time running gaming/workload benchmarks. Felt it was a better use of the cards than straight up mining) but never saw or recognized the warnings until I was all done and returning the system back to normal everyday use. I was just starting to research writing my own BIOS for the cards to see about controlling all this upon start up instead of after boot when the offer was made. It would have been an interesting project.
Interesting; since we're talking black screens here, I'll tell my R9 290 story now; my specimen of R9 290 seems to be an OEM board for internal AMD office usage, which somehow made its way out of the office - it does not belong to any specific card vendor (Sapphire, XFX etc.), and it has just an AMD graphics sticker on the "turbine". Nothing else - not even "QC pass" on the back. The guy who sold it to me (just in antistatic "silver" plastic bag) said he got it from friends of friends (something like that) who work at AMD (this is all happening in Toronto, Canada, nearby to where AMD graphics (former ATI) division is). So, the card appears to be indistinguishable in terms of hardware from retail units (same AMD C671 marking on the PCB etc.), but initially it had very early and poor ...002... VBIOS (with even different GPU boost frequency state - 937 MHz instead of established 947 MHz). So, guess what - in idle 2D I got an unstable picture with horizontal ripple-like artifacts, followed by black screens, with it (because of poorly defined voltages for power tune in this ...002... VBIOS it had). After researching the issue, two days later I reflashed it with retail ...007... bios from techpowerup. And voila - since then it is completely stable so far (as I wrote here already), but only AFTER this VBIOS reflashing. It's all like a "detective story", which was specifically possible due to my location in Toronto, near former ATI office. Very unusual and uncommon experience, I suppose :)
Seriously? You can't see the point? Well here is a decent example.
I built an APU system because I wanted something between a console and a computer. That is all our budget allowed for. What we came up with was a perfect combination of both worlds. For damn near the same cost of a PS4 or a XB1 I built an APU system that games at higher quality settings than either console, and is a decent computer for all my families needs. My wife does photo editing in multiple screens along with standard online streaming and day to day use (internet, office, email, etc...).
AMD's APU's really have a nice niche for those that don't have a budget for both a computer and a console. APU systems combine the two at a good price point. I personally can't stand the console wars, but a home built system for $500 that plays games at console quality (and in some cases better) and can handle all aspects of day to day life and professional applications is a perfect fit.
...and you completely missed the point. I'm not discussing the benefits of an older graphics card over an APU. I get that, and I totally agree that a STAND ALONE graphics card is going to be more powerful than an APU. What I'm giving you is a perfect example of how APU's are viable. Both as a decent gaming system (compared to purchasing a console) and a good all around family work horse at a VERY reasonable price.
As for your cost comment.. . How do you go from someone stating $120 is to expensive for a graphics card to them not making enough money? Really? I've got my money dedicated to bigger things... house, retirement, college fund for my kids, school expenses and I'm by far not alone on this front. If I was single... ok... spending $500 on a graphics card wouldn't be a big deal. It's a choice of not wanting to spend that kind of money on a non needed item. The APU system works great for everything that I and my family want it to be used for, and as gaming goes its a great little system... for $500 (ok... you got me... I put the solid state in... so $600).
You gotta be able to look at the whole picture dude, not just your own narrow view.
Yes, I completely understand your point of view. You are just not a hardware enthusiast, like I am. It's all just a matter of personal priorities. BTW, I also have three dependents on me, wife and two very little kids, so don't think I'm a super freakin' single standalone geek - FYI, I'm 32, I have a PhD, work as a scientist and feed my young family. I just generally like fast PCs (and doing my work-related calculations sometimes on my home desktop) and fast graphics in particular. Best regards :)
Worst thing is I built a near identical system - and it has been a poor performer! The APU runs so hot in the small case that I got it undervolted and capped at 3.7 GHz, GPU at 1 GHz. Plays Planetside 2 very poorly (20-30 FPS at 1366x768 and low quality), an i7-4700MQ and its iGPU did only 10 FPS worse on the same settings (and 80W less)... One game, but damn I love that one game! The APU has heart, but it just cooks! No way games can go at 1080p and 30-40 fps, maybe 720p and 30-40 fps for vanilla Skyrim...
You've either got incredibly poor airflow through your case or you don't have the heat sink on correctly... or both. Having to under volt any chip at stock speeds because it overheats is a huge sign that the heat draw off the chip is failing and failing hard. I've played around with overclocking my 7950k on the stock heat sink with no fans drawing air in or out of the case and achieved very acceptable speeds under VERY acceptable temps. So... you might want to pull your system apart, clean off that heat sink, apply some new thermal paste and reapply it. 'Cause something is obviously wrong. And if you are positive you have your system properly cooled, RMA that chip. Good luck.
Not a criticism, but I always find it amusing when drivers support arbitrary hardware. (Unrelated but I also think its hilarious when minimum requirements on games or other software are inconsistent between Intel/AMD and AMD/Nvidia)
On the desktop they're only supporting GCN - which makes sense given the age of the architecture. The drivers for their VLIW5 and VLIW4 architectures are probably about as mature as they are going to get, minus specific changes to enable new games.
On the flip side, mobile support extends to the 7000M series where most of the SKUs are still VLIW5 - specifically Caicos and Turks.
Though this is a beta driver, it is probably a sign of things to come - that someday soon driver updates will be limited to GCN or newer hardware. From a business side I think this makes a lot of sense. I'm sure there's a lot of testing and validation that has to go on for these drivers to even reach a beta stage. By limiting themselves to GCN they save engineering resources yet still cover (almost) 3 years of hardware.
That in mind, I can see a lot of people getting up in arms should AMD drop support for the HD5000 series, let alone the HD6000 series, come the next WHQL release.
I bet, in that case, there will be tons of people bashing AMD for dropping support for their HD 5000 and HD 6000 series, because there is a large category of people who use their cards until they die due to some reason. Besides, Trinity & Richland APUs are also VLIW4, and these are still sold today. So, support for VLIW5 and VLIW4-based graphics should continue for a while, I think (despite the fact that I'm on GCN myself for a while (HD 7950 before, and now R9 290).
I meant, the cards "die" of course - due to, say, fan(s) mechanical failure or board electrical failure - nothing is eternal in this world, especially any electronics made in last 10-15 years.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
28 Comments
Back to Article
Pissedoffyouth - Tuesday, November 11, 2014 - link
Owning an APU makes you really love driver updates :)ddriver - Tuesday, November 11, 2014 - link
What do driver updates change for an APU? It is still low end hardware, a few % faster doesn't mean anything.talonz - Tuesday, November 11, 2014 - link
I'd argue the opposite. You're going to the difference between 22 and 25 fps a lot more than 60 to 67.nissefar - Tuesday, November 11, 2014 - link
Well, both 22 and 25 fps is unplayable, so dunno about that...TheinsanegamerN - Tuesday, November 11, 2014 - link
25FPS is not ideal, but its far from unplayable. 22 is playable for turn based games, but not for anything that requires reflexes.JarredWalton - Wednesday, November 12, 2014 - link
If 25 FPS is playable, then you can probably manage with 22 FPS as well. And of course, a difference of 22 and 25 FPS is larger (13.6%) than 60 and 67 FPS (11.7%), plus if you run with VSYNC then 67 FPS isn't really important -- it's just the goal of getting over 60.Personally, most games are not "playable" until at least 30 FPS for me -- or at least they're not very enjoyable at sub-30 FPS. 28/29 FPS, sure, you could live with that, but 25 is too far down the performance ladder.
ET - Wednesday, November 12, 2014 - link
Looks like you pretty much agree with the concept, just not the details. Had the example been going from 27 to 30 fps, would that have been more convincing?JarredWalton - Wednesday, November 12, 2014 - link
I'd say more like 25 to 30 is what would be desirable to take a game from choppy to truly playable (though obviously there are exceptions). There's also the question of minimum frame rates -- if the game averages 35 FPS over a large portion of the game but there are areas (or battles) where it consistently drops to 20-25 FPS, then really you'd want to use settings that can get the minimum FPS up closer to 30.The real problem is that I don't know that I've ever seen a driver update provide a huge increase in performance for an APU. The "up to 30% faster" sort of claims are often for the high-end GPUs, and even then a lot of the time it's more like 15% (e.g. half of what was claimed) with a few edge cases seeing larger gains.
Anyway, I'm curious to know if the drivers helped at all for people with a Kaveri APU in the games mentioned. Let me know! :-)
TiGr1982 - Wednesday, November 12, 2014 - link
I don't see a single point in APU gaming AT ALL - at least, on desktops. You can get, say, a used HD 7950 for around $120 these days. If somebody wants to game, is it such a big deal? It's like a price of two AAA games, or, like, say, refueling a car fuel tank...APUs are for office PCs or, say, for kids or grandma's/grandpa's PCs.
ET - Wednesday, November 12, 2014 - link
I'd say that APU's are more for laptops and HTPC which are used for some gaming.Still, the "get a used HD 7950 for around $120" claim is rather silly. First of all, it may be true in the US, but (contrary to popular opinion of US residents) not everyone lives in the US. Secondly, for some people $120 is a lot of money, and being able to play some games on a budget PC is preferable to not doing it because the next step up is out of your budget.
TiGr1982 - Wednesday, November 12, 2014 - link
@ETI said USED card from somebody else's hands, not a new card.
I'm not in US either, I'm in Canada, for your reference - please stop making stupid conclusions without any information. So, if you go to Canadian online flea markets you can get, say, HD 7950 even for $100 and buy it in one day (e.g, I just bought very slightly used R9 290 for $250 two weeks ago, because the seller upgraded to GTX 970, which is cooler, quieter and much less power hungry).
"For some people $120 is a lot of money" - then maybe it's better to earn a little money, then game (in this exact sequence, not the other way around, as you imply).
TiGr1982 - Wednesday, November 12, 2014 - link
BTW, HD 7950 (Tahiti Pro) has WAY more firepower, than A10-7850K APU - 28 CU instead of just 8, plus higher GPU frequencies, plus 3 GB of dedicated GDDR5. It's a completely different device in terms of performance. Overall graphical performance of HD 7950 is 4 times bigger, if not more.So, gaming on APU is like riding a bike on a highway, when you can easily rent a car for relatively cheap.
TiGr1982 - Wednesday, November 12, 2014 - link
I'm not actually trying to be arrogant. Let's put it another way- in terms of price to performance ratio. For $200 you can get new Core i5, and for $100-120 you can find used GTX 760 or HD 7950 or some other comparable discrete gfx card. DDR3 DRAM is the same for both platforms (LGA1150 and FM2+), motherboard cost can also be around the same. 500W PSU is enough for these cards, which is also not a big deal.So, all other things being essentially equal, it comes to $320 for Core i5 + gfx card vs $160 for AMD APU. However, you get 4 times more performance in graphics + way faster CPU everywhere for just additional $160 in case of i5 + dedicated gfx card. To me, the choice is obvious, as I said above - and that's why AMD now has less than 20% CPU (including APU) market share.
wolfemane - Wednesday, November 12, 2014 - link
HA! The i5 with used 760 is what I originally went after. But we (wife and I) ran into three issues. First, the compact case we wanted didn't allow for the 760 (little larger than the PS4). Going with a case that could increased the cost. Second came down to price. Third came down to the memory we both wanted to use, DDR3-2400.I didn't want a console, I've never really been a big fan. My wife didn't want a console because it could'nt do anything she wanted to on the work end of things. The middle ground was a compact PC with the budget of a console. That target was $400 - $500. The i5, vid card, and a compact case that fir that just didn't work. It put the price to high.
I installed 32gig of 2400 into a clients computer who did a lot of video/photo editing and the difference in performance for him was pretty amazing. But I had a hell of a time getting his board to recognize it (MSI board z87 1150). We finally got it to work, but it was a chore. I read several builds with the FM2+ chipset having no issues with the ram, and sure enough it was recognized without issue when we build our system. All in all we spent $500. Slightly more than the PS4 and much cheaper than ps4+computer and we saved a bit of space in front of our TV. And it performs GREAT for what it is and what it replaces.
TiGr1982 - Wednesday, November 12, 2014 - link
@wolfemaneYes,support of different RAM on different motherboards is not really a sure thing, unfortunately.
In my case, however, I have 32 GB of DDR3-2400 RAM (2 identical 2x8 GB G.SKILL CL11 kits) in my aforementioned home desktop, and it works perfectly from the day I built the machine originally (back in June 2013, Z87 build also) with Core i7-4770 @ 3.7-3.9 GHz, and since September 2014 it continues to be rock solid with Core i7-4790K @ 4.3-4.5 GHz on the same M/B. I, however, have Gigabyte Z87X-UD3H M/B. Works like a charm with 32 GB of DDR3-2400. So, it is not LGA1150 Z87 platform itself which is glitchy in terms of memory support as a whole, it's just the particular M/B and/or RAM which caused these issues.
wolfemane - Wednesday, November 12, 2014 - link
Yeah it turned out to be the board, had to get MSI on the phone and wound up with slightly different timings to get it to work.Your system reminds me of my previous desktop. ASUS Sabertooth z77, i7-3770k @ 4.6ghz, 16gig GSkill trident X DDR3-2400, and twin sapphire tri-X OC 290's all on a duel radiator loop. Thing ran solid @ 4.6ghz and the memory worked right out of the package. Wih a little help from a certain OC site I had it at 5.1ghz with a custom fish tank chiller attached.
Unfortunetly I started getting weird black screens. Turns out both video cards had the non hynix (Elpida I believe)
memory and was suspect to the problem. After several months of unsuccessful attempts to correct the problem and on the verge of RMA a friend offered a decent amount for the whole system. He didn't care about the vid problems. Couldn't refuse the offer and planned on a similar build to what you have, but decided on other things.
TiGr1982 - Wednesday, November 12, 2014 - link
@wolfemaneOh! Then you are also a powerful PC enthusiast; I'm taking my words back :)
I read about these possible Elpida VRAM R9 290(X) problems on the web. Luckily for me, my R9 290, despite being reference hot and noisy unit with "turbine" (centrifugal fan), has Hynix VRAM on it, according to GPU-Z and Hawaii Memory Info. Got it recently (around 3 weeks ago) and it seems to be very stable so far in all the loads I tried (no throttling even at slight overclocking and long 3D loads, but the "turbine" goes to 50+% speed and sounds like an airplane - headphones needed).
BTW, I had two reference "turbine" HD 4870s in CrossFire in the past (back in 2008-2009) and also had no problems with them (heat and noise is "by design" there, it seems to be not a problem for stability).
Regarding Elpida memory on R9 290(X), people on the web are saying that maybe downclocking could help (but will hurt performance, of course).
wolfemane - Wednesday, November 12, 2014 - link
That is one solution that has helped some. The best recommendation was using GPU tweak to set/limit 2D core and mem clocks to a fixed speed of 450/500. It decreased the number of black screens, but I still had the problem. The Elpida memory is very sensitive to voltage controls. The best way to put it is from our good friends over at overclock.net."This black screen from power transitions occurs because the memory clock jumps up to full speed (1250, etc), while the GPU voltage (which also controls the IMC--this is the problem) is STILL running at idle volts of 0.9xxx). That idle voltage is too low to handle the IMC, which is controlling the memory running at that speed (this is NOT due to defective memory modules btw), and thus the card black screens.
The fix, is to increase the GPU voltage, which ALSO increases the idle voltage to the IMC as well. (previous generation cards used separate voltages for idle clocks and 3D). This Is purely a bios/design problem. The IMC should NEVER be running at 0.9xx when the memory jumps to full speed, or, alternatively, there should be 'lesser' memory speed jumps, for when a light load is put on the GPU, so the memory doesn't jump from 150 MHz to 1250, but instead it could be 500 MHz for a very light load (like when the core goes from 300 to 450 MHz), etc."
I used both MSI AB and Sapphire Trixx to adjust my voltage higher but I still had the same issues (both programs I used when I was overclocking originally). I have a feeling that the black screens are due more to my previous overclocking. I was never really concerned about pushing my system (and I pushed it hard) because I never had any of the tail tale signs of problems (excessive heat, stuttering, visual degradation, etc..). At the time I could have pushed the cards just to hard (I was using seti@home in background mode @ 100% GPU usuage while at the same time running gaming/workload benchmarks. Felt it was a better use of the cards than straight up mining) but never saw or recognized the warnings until I was all done and returning the system back to normal everyday use. I was just starting to research writing my own BIOS for the cards to see about controlling all this upon start up instead of after boot when the offer was made. It would have been an interesting project.
TiGr1982 - Wednesday, November 12, 2014 - link
Interesting;since we're talking black screens here, I'll tell my R9 290 story now; my specimen of R9 290 seems to be an OEM board for internal AMD office usage, which somehow made its way out of the office - it does not belong to any specific card vendor (Sapphire, XFX etc.), and it has just an AMD graphics sticker on the "turbine". Nothing else - not even "QC pass" on the back. The guy who sold it to me (just in antistatic "silver" plastic bag) said he got it from friends of friends (something like that) who work at AMD (this is all happening in Toronto, Canada, nearby to where AMD graphics (former ATI) division is). So, the card appears to be indistinguishable in terms of hardware from retail units (same AMD C671 marking on the PCB etc.), but initially it had very early and poor ...002... VBIOS (with even different GPU boost frequency state - 937 MHz instead of established 947 MHz). So, guess what - in idle 2D I got an unstable picture with horizontal ripple-like artifacts, followed by black screens, with it (because of poorly defined voltages for power tune in this ...002... VBIOS it had). After researching the issue, two days later I reflashed it with retail ...007... bios from techpowerup. And voila - since then it is completely stable so far (as I wrote here already), but only AFTER this VBIOS reflashing.
It's all like a "detective story", which was specifically possible due to my location in Toronto, near former ATI office. Very unusual and uncommon experience, I suppose :)
wolfemane - Wednesday, November 12, 2014 - link
Seriously? You can't see the point? Well here is a decent example.I built an APU system because I wanted something between a console and a computer. That is all our budget allowed for. What we came up with was a perfect combination of both worlds. For damn near the same cost of a PS4 or a XB1 I built an APU system that games at higher quality settings than either console, and is a decent computer for all my families needs. My wife does photo editing in multiple screens along with standard online streaming and day to day use (internet, office, email, etc...).
AMD's APU's really have a nice niche for those that don't have a budget for both a computer and a console. APU systems combine the two at a good price point. I personally can't stand the console wars, but a home built system for $500 that plays games at console quality (and in some cases better) and can handle all aspects of day to day life and professional applications is a perfect fit.
miniITX - $50 - $70
7850k -$160
8gig DDR3-2400 - $80
PSU-$50
Case-$50
2tb Barracuda HD - $75 - $90
Total: $500
Throw in a 256gig solid state for the OS and that's another $75 - $100
System can run most games at console quality, 1080 @ 30-40fps. I can easily hit 60fps when I drop down to 720 with higher quality settings.
TiGr1982 - Wednesday, November 12, 2014 - link
@wolfemaneSee above.
wolfemane - Wednesday, November 12, 2014 - link
...and you completely missed the point. I'm not discussing the benefits of an older graphics card over an APU. I get that, and I totally agree that a STAND ALONE graphics card is going to be more powerful than an APU. What I'm giving you is a perfect example of how APU's are viable. Both as a decent gaming system (compared to purchasing a console) and a good all around family work horse at a VERY reasonable price.As for your cost comment.. . How do you go from someone stating $120 is to expensive for a graphics card to them not making enough money? Really? I've got my money dedicated to bigger things... house, retirement, college fund for my kids, school expenses and I'm by far not alone on this front. If I was single... ok... spending $500 on a graphics card wouldn't be a big deal. It's a choice of not wanting to spend that kind of money on a non needed item. The APU system works great for everything that I and my family want it to be used for, and as gaming goes its a great little system... for $500 (ok... you got me... I put the solid state in... so $600).
You gotta be able to look at the whole picture dude, not just your own narrow view.
TiGr1982 - Wednesday, November 12, 2014 - link
@wolfemaneYes, I completely understand your point of view. You are just not a hardware enthusiast, like I am. It's all just a matter of personal priorities. BTW, I also have three dependents on me, wife and two very little kids, so don't think I'm a super freakin' single standalone geek - FYI, I'm 32, I have a PhD, work as a scientist and feed my young family. I just generally like fast PCs (and doing my work-related calculations sometimes on my home desktop) and fast graphics in particular.
Best regards :)
genekellyjr - Monday, November 17, 2014 - link
Worst thing is I built a near identical system - and it has been a poor performer!The APU runs so hot in the small case that I got it undervolted and capped at 3.7 GHz, GPU at 1 GHz. Plays Planetside 2 very poorly (20-30 FPS at 1366x768 and low quality), an i7-4700MQ and its iGPU did only 10 FPS worse on the same settings (and 80W less)... One game, but damn I love that one game!
The APU has heart, but it just cooks! No way games can go at 1080p and 30-40 fps, maybe 720p and 30-40 fps for vanilla Skyrim...
wolfemane - Thursday, November 20, 2014 - link
You've either got incredibly poor airflow through your case or you don't have the heat sink on correctly... or both. Having to under volt any chip at stock speeds because it overheats is a huge sign that the heat draw off the chip is failing and failing hard. I've played around with overclocking my 7950k on the stock heat sink with no fans drawing air in or out of the case and achieved very acceptable speeds under VERY acceptable temps. So... you might want to pull your system apart, clean off that heat sink, apply some new thermal paste and reapply it. 'Cause something is obviously wrong. And if you are positive you have your system properly cooled, RMA that chip. Good luck.MrCommunistGen - Wednesday, November 12, 2014 - link
Not a criticism, but I always find it amusing when drivers support arbitrary hardware. (Unrelated but I also think its hilarious when minimum requirements on games or other software are inconsistent between Intel/AMD and AMD/Nvidia)On the desktop they're only supporting GCN - which makes sense given the age of the architecture. The drivers for their VLIW5 and VLIW4 architectures are probably about as mature as they are going to get, minus specific changes to enable new games.
On the flip side, mobile support extends to the 7000M series where most of the SKUs are still VLIW5 - specifically Caicos and Turks.
Though this is a beta driver, it is probably a sign of things to come - that someday soon driver updates will be limited to GCN or newer hardware. From a business side I think this makes a lot of sense. I'm sure there's a lot of testing and validation that has to go on for these drivers to even reach a beta stage. By limiting themselves to GCN they save engineering resources yet still cover (almost) 3 years of hardware.
That in mind, I can see a lot of people getting up in arms should AMD drop support for the HD5000 series, let alone the HD6000 series, come the next WHQL release.
TiGr1982 - Wednesday, November 12, 2014 - link
I bet, in that case, there will be tons of people bashing AMD for dropping support for their HD 5000 and HD 6000 series, because there is a large category of people who use their cards until they die due to some reason. Besides, Trinity & Richland APUs are also VLIW4, and these are still sold today. So, support for VLIW5 and VLIW4-based graphics should continue for a while, I think (despite the fact that I'm on GCN myself for a while (HD 7950 before, and now R9 290).TiGr1982 - Wednesday, November 12, 2014 - link
I meant, the cards "die" of course - due to, say, fan(s) mechanical failure or board electrical failure - nothing is eternal in this world, especially any electronics made in last 10-15 years.