AMD CPU speculation... and expert conjecture

[8350rocks]

The 35W difference is mostly due to being a K series. If you don't overclock then the power consumption will be very close to the 5700.

not even close. i was talking about load consumption at stock, which is at least 35w higher i.e. without any overclock. when overclocked, a10-5800k can use as much as 30w+(than stock) on average, higher on load. not suitable for sff.

It uses more a bit more power because is a larger card with more RAM. It run so hot as Nvidia but is more quiet. The 7990 is faster than the 690 and Titan. In some game that I have seen the 7990 gives up to 200% increase in performance over both!!!!

You have a faster card, less noisy, much more RAM, better dual, and bundle promo.

it uses more power because of more vram and 'card size'? then why isn't titan(250w) using 375w? it has same 6gb of vram (on the pcb) just like 7990. this isn't the same as radeon 6670/7750 with ddr3 vs gddr5. you need to get more information on gpu, board, vram tdp and board power limits.
it's less noisy because of the cooler design (3 axial fans), the components don't produce any less power. it's actually a bit worse because the coolers dump all the hot air inside the case, unlike gtx 690, which blow part of the heated air out of the case.
7990 definitely has the raw performance and potential. but the card design and software support is really lacking(long way to launch) and amd's launch timing is terrible. try this for angles: nvidia launched their gtx690 (almost a year) before they made fcat available for reviewers and before frame time analyses was as widespread as it is now.
actually titan(6gb) has the most amount of vram available to a high perf. consumer gpu, not 7990. 7990 has 3gb per gpu. it says right there in the first page of the review. doesn't seem like you read the review(s), only looked at the longer 'bars' in fps charts..... :lol:
if you need more info, check out graphics cards section of the forum.

Titan and the GTX690 while quieter, also run quite a bit hotter...or did you read Tom's review? Under load the GTX 690 can reach temps of 90-100C and it would require additional cooling to be overclocked beyond the factory boost.

Additionally AMD's HD 7990 has 3 intake fans and vents alot of heat out the rear of the card.

Titan gets destroyed by the HD 7990 too btw...that's why nvidia is coming out with the Titan Ultra. Additionally, you don't get 7-8 games with Titan, so you're paying more for less.

Titan makes no sense, and the GTX 690 might be the quietest card, but it runs the hottest against the vendor cards that were factory overclocked, etc. Though the Powercolor card was the power hungriest, the HIS 7970X2 in the test was clearly the best performing card...Tom's, in unusual fashion, showed a preference for the nvidia card despite the fact that it's cooling was subpar and performance was 3rd best. They made a huge issue out of coil whine as well.

 

[8350rocks]

well it is easier to find a nail in heystack rather than fx-6300 in market
so currently i have no price info about fx6300 :/

though to give an idea, fx-6100 is available for $120-130
fx-8150 for ~$200
fx-8350 for ~$220
and for a board like ASUS M5A97 R2.0 ~$125 is the cost

i had that impression already... just needed to verify it. i'll add: the few places that carried fx6300 (possibly for a very limited amount of time, or preorder), they were priced higher than fx8000 cpus. geddit? higher, as in between fx8350 and core i5 3570k. you could get a core i5 3550 for less than fx8350 with a cheap b75 or z77 mobo (eg. asrock z77 pro4) and handily out-game a non existent fx6300. so much for price/performance. :lol: speaks volumes about amd's execution.

That's a supply issue...when they're 1/17th the size of intel, does it surprise you at all that they don't make as many chips?

 

[kettu]

@gamerk316: To answer the question about FPUs, in Steamroller, the FPUs will be beefed up as well...each separate FPU will be able to do 1 additional DP MAD FP calculation...which means effectively a 50% increase in FP capability...(going from 2 DP MAD FP calculations per unit to 3 per unit is a substantial % increase over PD). So, steamroller is a large increase across the board...hence the 30% figure.

I haven't heard that before. Got a link, I'm interested to read more about it.

The Keynote speech was delivered about 30 days prior to the FX piledriver release, but after trinity launch, and after the paper launch of the FX series...I would be fairly certain he was talking about PD, since that technology was already out or arriving. For example, the Richland paper launch has already occurred, and we still do not see them released yet.

That's a difficult call. It could be either way but I think it doesn't matter in the big picture. As far as I'm aware the front-end throughput difference clock-for-clock between BD and PD is minor at best. It's basically few tweaks here and there across the entire module that gave a nice performance boost for PD. A major contributor being the improved power efficiency allowing higher clockspeeds.

My understanding is that this 30% throughput increase is just refering to the improved fetch hardware and the addition of the second decode unit. (I can't remember off-hand if there were some other improvements there.) And that this increase is realised in situations where the module is fully loaded. How that translates to real world performance is another matter. In cases where the cores were starved I think the gains will be significant. I wouldn't be surprised if it was pretty close to 30%.

On top of that the actual core has it's own tweaks that will improve performance in other cases. But my guess is that BD/PD cores are being atleast somewhat starved in most multi-threaded situations. So in strictly single-threaded situations the performance improvement won't be that awesome. Is 7% - 10% the norm these days between CPU generations? But in most multi-threaded realworld cases it's going to be better than that.

This kind of a multi-threaded performance gains are a big deal for AMD. They're going to be much more competitive in the high-margin server market. It's also a pretty big deal in consumer markets with games becoming more and more threaded. I don't see how Intel can compete on gaming performance with their multiplier-locked dual-core line-up against Steamroller quad-cores. Now there will be lightly-threaded cases where i3 will be ahead of the steamroller but the problem I see is while SR's single-threaded performance is "good enough" I don't think same can be said for multi-threaded performance of i3. My guess is were going to see an unlocked i3 next-generation. That's the only way I can see them still being relevant. The i5 series will definately be a contender and with multi-threaded games I don't see unlocked i3's eating that much sales against i5 because it will be superior. Rattling sabres with the six-core SR.

As for the FX-6300 availability:
It can be had in Finland for ~130-140 euros. Roughly ~10-20 more than an i3 and ~30 less than FX-8320

 

[8350rocks]

Steamroller articles:

http://www.xbitlabs.com/news/cpu/display/20130331080217_AMD_We_Are_On_Track_With_Steamroller_Micro_Architecture_in_2013.html

http://www.bit-tech.net/news/hardware/2012/08/29/amd-steamroller/1

http://en.wikipedia.org/wiki/Bulldozer_(microarchitecture)

http://www.brightsideofnews.com/news/2013/3/6/analysis-amd-kaveri-apu-and-steamroller-core-architectural-enhancements-unveiled.aspx

Those articles should cover the amount of information available at this point in time...the last one discusses FPU enhancements in some detail...I couldn't find the link to the article that gave a thorough engineering level breakdown of the improvements, but those "layman terms" articles should give the gist of it.

 

[kettu]

Thanks. If you happen to stumble upon the more thorough article please share.

Edit:
What I got out of them is that AMD is dropping one MMX pipe from the FPU and reducing FP pipelinestages from 4 to 3 reducing instruction latency. Also various undisclosed tweaks.

 

[de5_Roy]

Titan and the GTX690 while quieter, also run quite a bit hotter...or did you read Tom's review? Under load the GTX 690 can reach temps of 90-100C and it would require additional cooling to be overclocked beyond the factory boost.
Additionally AMD's HD 7990 has 3 intake fans and vents alot of heat out the rear of the card.

one fan in titan and gtx690 vs 3 in 7990, with open air cooler design to boot. cooling the card would obviously be better than blower and central axial types. what's not obvious is that the cooler choice will affect heat buildup inside the case and case choice- which was my concern. to repeat - gtx690's cooler blows part of the hot air out of the case, titan's blows most of it out, 7990's dumps all of it inside the case.
just in case, casually trying to dismiss by 'like, most/all cases totally have side vents so what-everr..' - won't change anything.
from toms' review:

There is a price to be paid for this sort of design, though. Three fans, side-by-side, are only effective if they aren’t doing battle with each other. And that means channeling air vertically, rather than horizontally. At the end of the day, then, the slotted rear I/O panel that typically helps exhaust hot air from most graphics cards is almost non-functional. Instead, waste heat from both GPUs is jettisoned out the top of the card, right into your case.

now... i don't know what you consider 'rear' of the card.... may be you meant the 'other' rear...wherever that is.....

Titan gets destroyed by the HD 7990 too btw...that's why nvidia is coming out with the Titan Ultra.

:lol: no amount of 'destroying'(!) can take away the ease of using single, power gpu and the sheer possibility of combining with may be another two. then there's it's easy fitment in smaller cases worrying much less about heat buildup. since these are all $1k cards being discussed, these kinds of factors matter quite a bit. titan being outperformed by 7990 is not really anything earth shattering - titan's configured to underperform gtx690. titan's purpose is just different.
additionally, 7990 suffers badly from amd's driver support. amd's trying so hard to make 7990 look like it's worth the $1k yet they neglect to provide good drivers for it. i noticed 7990 getting hilariously 'destroyed' by [strike]titan[/strike] 7990(prototype drv) in crysis 3 (titan outperformed and kept up in some others), especially in practical fps. it should not have happened because 7990 has more powerful gpus, yet it did happen. 7990 woulda looked worse if amd hadn't provided optimizations for hitman, skyrim and others and the prototype driver.
may be in a few months or later this year, 7990 will seem worth it, but the driver delay is inexcusable.
titan 'ultra' is showing amd that nvidia can have multiple halo gfx cards(with better drivers) while amd is struggling to make one seem worthwhile. marketing.
tbh, i think you entirely missed the reason why nvidia is suddenly making so many high end cards available. the real reason is so simple- improved yields and/or stockpiling, and amd's clean sweep with the consoles, especially with ps4. imo from todays benches, seems like nvidia could care less about 7990 since it doesnt have that vital software component available yet.

Additionally, you don't get 7-8 games with Titan, so you're paying more for less.

heh. you can bundle all the games in the worlds for free, it won't count towards calculating price/performance of the card. if 7990 was $200 or more cheaper, may be woulda mattered. may be. in the end, you end up paying a lot more for quite less - with 7990.

Titan makes no sense, and the GTX 690 might be the quietest card, but it runs the hottest against the vendor cards that were factory overclocked, etc. Though the Powercolor card was the power hungriest, the HIS 7970X2 in the test was clearly the best performing card...Tom's, in unusual fashion, showed a preference for the nvidia card despite the fact that it's cooling was subpar and performance was 3rd best. They made a huge issue out of coil whine as well.

titan makes no sense... to people who don't understand the purpose of halo products and simple marketing tactics; i'll just leave it at there. his' card never made it to the market, they withdrew. his is selling amd's reference 7990 instead of their own '7970x2'. coil whine is annoying and downright wrong for a highest end, halo card which should have top notch design. if it was 200-300 bucks cheaper, no one woulda cared.
.... most of these are addressed in toms' review at least, others touch on these in varying degrees. i sense lack of attention.

That's a supply issue...when they're 1/17th the size of intel,

size means nothing compared to execution. they have almost all other new cpus available but... just not the one that has the best chance for selling.

does it surprise you at all that they don't make as many chips?

they don't make 'as many chips'? they flooded markets with llano apus and zambezi cpus that no one cared to buy soon after trinity and vishera came out. they make plenty of chips alright.

 

[8350rocks]

one fan in titan and gtx690 vs 3 in 7990, with open air cooler design to boot. cooling the card would obviously be better than blower and central axial types. what's not obvious is that the cooler choice will affect heat buildup inside the case and case choice- which was my concern. to repeat - gtx690's cooler blows part of the hot air out of the case, titan's blows most of it out, 7990's dumps all of it inside the case.
just in case, casually trying to dismiss by 'like, most/all cases totally have side vents so what-everr..' - won't change anything.
from toms' review:

The card has a slotted design with exhaust vents on the rear of the card, where are you coming up with the design doesn't exhaust heat out of the rear of the card? The design schematics indicate that is not proper function. Perhaps the engineering sample was not 100% ready yet?

now... i don't know what you consider 'rear' of the card.... may be you meant the 'other' rear...wherever that is.....

It's exactly where the design schematics indicated it should be...

:lol: no amount of 'destroying'(!) can take away the ease of using single, power gpu and the sheer possibility of combining with may be another two. then there's it's easy fitment in smaller cases worrying much less about heat buildup. since these are all $1k cards being discussed, these kinds of factors matter quite a bit. titan being outperformed by 7990 is not really anything earth shattering - titan's configured to underperform gtx690. titan's purpose is just different.
additionally, 7990 suffers badly from amd's driver support. amd's trying so hard to make 7990 look like it's worth the $1k yet they neglect to provide good drivers for it. i noticed 7990 getting hilariously 'destroyed' by [strike]titan[/strike] 7990(prototype drv) in crysis 3 (titan outperformed and kept up in some others), especially in practical fps. it should not have happened because 7990 has more powerful gpus, yet it did happen. 7990 woulda looked worse if amd hadn't provided optimizations for hitman, skyrim and others and the prototype driver.

HD 7990 drivers are in beta, how are they deficient? By the time the product is launched it should be fine...by the way...the Crossfire issues have been fixed now, and I have every reason to believe that the next set of drivers will be fantastic with Kudori being responsible for that.

may be in a few months or later this year, 7990 will seem worth it, but the driver delay is inexcusable.
titan 'ultra' is showing amd that nvidia can have multiple halo gfx cards(with better drivers) while amd is struggling to make one seem worthwhile. marketing.
tbh, i think you entirely missed the reason why nvidia is suddenly making so many high end cards available. the real reason is so simple- improved yields and/or stockpiling, and amd's clean sweep with the consoles, especially with ps4. imo from todays benches, seems like nvidia could care less about 7990 since it doesnt have that vital software component available yet.

The drivers issue will be fixed in short order.

heh. you can bundle all the games in the worlds for free, it won't count towards calculating price/performance of the card. if 7990 was $200 or more cheaper, may be woulda mattered. may be. in the end, you end up paying a lot more for quite less - with 7990.

No, what matters is it outperforms the GTX690 and GTX Titan, so you're getting more for the same money. If Titan was about $600 then it would make sense...especially considering the HD 7970 blows it away, and the GHz edition will nearly keep up with the GTX 690 in many applications.

titan makes no sense... to people who don't understand the purpose of halo products and simple marketing tactics; i'll just leave it at there. his' card never made it to the market, they withdrew. his is selling amd's reference 7990 instead of their own '7970x2'. coil whine is annoying and downright wrong for a highest end, halo card which should have top notch design. if it was 200-300 bucks cheaper, no one woulda cared.
.... most of these are addressed in toms' review at least, others touch on these in varying degrees. i sense lack of attention.

Coil whine is a product of the massive cooling required to keep such a high end card cool. Would you rather have something inside your case running at 90C or 70C? Considering that most people will be running high end cooling systems that own these cards, and, will not be using an open benchmark setup like Tom's review did, the noise is acceptable. If you're running water cooling on your system, you won't even hear the fan in most cases over the massive push/pull fans on your radiator. So what's your point? This is not a product for a low end gaming machine...this card goes into a $2k computer build, and under those circumstances, you could use a high end sound dampening case to reduce any noise anyway.

they don't make 'as many chips'? they flooded markets with llano apus and zambezi cpus that no one cared to buy soon after trinity and vishera came out. they make plenty of chips alright.

Considering they possess 1/4 of the market share, I would say intel produces 3x more chips than AMD. Their biggest market presence is in the US, no doubt they have supply issues overseas, and that's something they could actively work toward remedying, but haven't addressed other issues first which are more important.

 

[gamerk316]

OFF TOPIC:

The HD7990 represents enthusiast level, and trust me when I say an enthusiast building a 7990 based system is not going to give a continental about power and noise it uses and there are reasons for that;

1) Surveys suggest that most owners of these cards will use some form of water cooling which has a much higher accoustic level than the card, to compound this they are likely to be watercooling the GPU's so sound and the extra power is hardly a feasible criteria.

2) AMD got very serious with its beta drivers, considering it's beta and more improvements are to come, Crossfire is no longer a legitimate arguement.

3) The extra power corrolates into 90FPS in BF3 at enthusiast level resolutions, this mark was sizeably above the GTX690 and made the Titan look like childs play, so for that added power corrolated into substantial performance and now AMD all but has a card to control the $1000 mark by a comfortable margin.

Nvidia won the last generation, AMD has won this won. To make the nut even sweeter is the fact that AMD's never settle bundle with the HD7990 includes 7 top end titles in the cost so even at $1000 factor in 7 games with it which easily comes to $50-60 per game is a very good bargain imho.

well it is easier to find a nail in heystack rather than fx-6300 in market
so currently i have no price info about fx6300 :/

though to give an idea, fx-6100 is available for $120-130
fx-8150 for ~$200
fx-8350 for ~$220
and for a board like ASUS M5A97 R2.0 ~$125 is the cost

i had that impression already... just needed to verify it. i'll add: the few places that carried fx6300 (possibly for a very limited amount of time, or preorder), they were priced higher than fx8000 cpus. geddit? higher, as in between fx8350 and core i5 3570k. you could get a core i5 3550 for less than fx8350 with a cheap b75 or z77 mobo (eg. asrock z77 pro4) and handily out-game a non existent fx6300. so much for price/performance. :lol: speaks volumes about amd's execution.

That's a supply issue...when they're 1/17th the size of intel, does it surprise you at all that they don't make as many chips?

AMD doesn't make chips; they sold their fab unit, remember?

 

[juanrga]

Jaguar SOCs are being sold in the embedded market. Prices range from $49 to $72.

They should show up on boards like this any day now. That platform was always more geared towards embedded with BGA only parts.

http://www.newegg.com/Product/Product.aspx?Item=N82E16813131843

That C60 is one of my favourite fanless mobos.

 

[8350rocks]

AMD doesn't make chips; they sold their fab unit, remember?

Well, yes, they spun off GloFo, but they are still technically supplying chips, and control supply of their product.

 

[de5_Roy]

The card has a slotted design with exhaust vents on the rear of the card, where are you coming up with the design doesn't exhaust heat out of the rear of the card? The design schematics indicate that is not proper function. Perhaps the engineering sample was not 100% ready yet?

It's exactly where the design schematics indicated it should be...

lame, baseless excuses. amd or anyone in the right mind won't give below-standard products to reviewers. they are well known for supplying reviewers with [strike]golden[/strike] cherry picked, best samples that deliver the best performance in order to look good in the benchmarks and tests.
i said the 7990's cooling design dumps heat inside the case. the vent in 7990 only dissipates minor part of the hot air from the closest fan, not from the other two further inside.
btw, fanboy linear mathematics (FL math(tm)) like '1 fan out of 3 is fiiine' does not apply here.

HD 7990 drivers are in beta, how are they deficient? By the time the product is launched it should be fine...by the way...the Crossfire issues have been fixed now, and I have every reason to believe that the next set of drivers will be fantastic with Kudori being responsible for that.

The drivers issue will be fixed in short order.

they're not just deficient, the sheer delay when this premium 7990 card is launching within weeks is bad. amd themselves have said that the prototype driver, the one that actually improves multi-gpu performance and experience, will come out in 2h 2013, could be as early as june or as late as october. either way it's way too long. per-game optimizations (in catalyst profiles) won't carry them for long, they only look passable in benchmark reviews, not so much in real life..

No, what matters is it outperforms the GTX690 and GTX Titan, so you're getting more for the same money. If Titan was about $600 then it would make sense...especially considering the HD 7970 blows it away, and the GHz edition will nearly keep up with the GTX 690 in many applications.

7990 trades blows with gtx 690 and only outperforms titan when the driver isn't making sure it can't. and performance/price does favor gtx690, titan only falls short in 1600p, if tpu's analysis is of any indication. even if i think tpu favors nvidia, 7990's perf/price against amd's own 2x 7970 is much worse. you really pay much [strike]less[/strike] more with 7990 and get less. the things that should justify 7990's extra price, don't deliver. worse, counting drivers.
titan shoulda been $600 or less.. even, only if there was a similarly competitive card from amd. your claim about 7970 beating/keeping up with titan and gtx690 is only right for niche gpgpu applications, that's nowhere near the general case. i noticed that you didn't use the word 'games'.

Coil whine is a product of the massive cooling required to keep such a high end card cool. Would you rather have something inside your case running at 90C or 70C? Considering that most people will be running high end cooling systems that own these cards, and, will not be using an open benchmark setup like Tom's review did, the noise is acceptable. If you're running water cooling on your system, you won't even hear the fan in most cases over the massive push/pull fans on your radiator. So what's your point? This is not a product for a low end gaming machine...this card goes into a $2k computer build, and under those circumstances, you could use a high end sound dampening case to reduce any noise anyway.

you need to read the section where toms' addresses this noise issue. seems you didn't understand it. i apologize for using 'coil whine' directly from your post due to negligence.
it's not because of 'massive cooling', the reason 'massive cooling' is required because the gpus produce more heat while trading blows with gtx690.
bringing up user-customized cooling, other cooling apparatus making more noise, 'high end sound dampening' and so on - make amd's cooling seem even worse. :lol: those are lame excuses anyway.
my point is simple - highest end products should be very, very good, especially in hardware. they should represent the best (of the best), the most innovation, the essence of a halo product. amd underdelivered (worse in some cases) in all of them. it's undoubtedly much better than previous 6990 and it does have the raw performance, but the whole package is sub-par for the price and for the competition. if it was cheaper, no question woulda be raised.

Considering they possess 1/4 of the market share, I would say intel produces 3x more chips than AMD. Their biggest market presence is in the US, no doubt they have supply issues overseas, and that's something they could actively work toward remedying, but haven't addressed other issues first which are more important.

lawl, it's the opposite of the 'amd has significant share (>30%, iirc) in desktop cpus' pitch! just to make amd's poor execution look better. there are no other issues more important than selling cpus - the right ones, as many as possible(regardless of yields), in potential markets. no one cares about intel when the fault is amd's own.
edit: fixed the brainfart and other errors.

 

[de5_Roy]

Inside the PlayStation 4 With Mark Cerny
http://www.gamasutra.com/view/feature/191007/inside_the_playstation_4_with_mark_.php?page=1
unified architecture is the name of the ...wait for it... game,geddit?

 

[8350rocks]

So PS4 is heavily HSA oriented and the APU is really going to act like a CPU/GPGPU combination capable of running approximately 96 simultaneous threads based on the pipelines and capabilities of the 8 core jaguar.

It seems no PC will be able to keep up with that for some time to come. SR is only going to be able to run 44 simultaneous threads, and tier 1 GPUs (HD 7950+) allow running 4 compute threads at once...for a total of 48.

 

[The Q6660 Inside]

Hmm, I doubt that the PS4 will truly have the advantage, but I had my doubts with the PS3's RSX and its 128-bit bus since the 7900GTX it was based on had 2x the bandwidth, but of course does not run the games at anywhere near as good quality as the PS3's RSX+Cell (largely thanks to optimized code and not sheer power). However, I do think that PS4/Durango ports should optimize SR and PDs 2 /4 unused cores in modern titles, thus possibly gaining ground on Team Blue, but I would think that Haswell would still have better IPC.

 

[griptwister]

Once SteamRoller comes around, we shall see who has the better IPC

 

[juanrga]

The 35W difference is mostly due to being a K series. If you don't overclock then the power consumption will be very close to the 5700.

not even close. i was talking about load consumption at stock, which is at least 35w higher i.e. without any overclock. when overclocked, a10-5800k can use as much as 30w+(than stock) on average, higher on load. not suitable for sff.

It uses more a bit more power because is a larger card with more RAM. It run so hot as Nvidia but is more quiet. The 7990 is faster than the 690 and Titan. In some game that I have seen the 7990 gives up to 200% increase in performance over both!!!!

You have a faster card, less noisy, much more RAM, better dual, and bundle promo.

it uses more power because of more vram and 'card size'? then why isn't titan(250w) using 375w? it has same 6gb of vram (on the pcb) just like 7990. this isn't the same as radeon 6670/7750 with ddr3 vs gddr5. you need to get more information on gpu, board, vram tdp and board power limits.
it's less noisy because of the cooler design (3 axial fans), the components don't produce any less power. it's actually a bit worse because the coolers dump all the hot air inside the case, unlike gtx 690, which blow part of the heated air out of the case.
7990 definitely has the raw performance and potential. but the card design and software support is really lacking(long way to launch) and amd's launch timing is terrible. try this for angles: nvidia launched their gtx690 (almost a year) before they made fcat available for reviewers and before frame time analyses was as widespread as it is now.
actually titan(6gb) has the most amount of vram available to a high perf. consumer gpu, not 7990. 7990 has 3gb per gpu. it says right there in the first page of the review. doesn't seem like you read the review(s), only looked at the longer 'bars' in fps charts..... :lol:
if you need more info, check out graphics cards section of the forum.

As said the 35W TDP difference is considering that one is a K series. If you don't overclock the power consumptions are similar. The difference is due to slightly higher stock clocks.

About the HD 7990 vs GTX 690 vs GTX titan, a resume of what said:

the HD 7990 is faster than the GTX 690 in most tests that we ran, but not in all of them.

[...]

Even more surprising is the card's lack of noise and heat, [...] We'd say all in all it's probably just a bit quieter than the GTX 690, and gets about as hot.

http://www.maximumpc.com/article/videocards/amd_radeon_hd_7990_first_look

As the results on the preceding pages have shown, the Radeon HD 7990 puts up some impressive frame rates, outpacing the similarly priced GeForce GTX 690 and GTX Titan more often than not.

[...]

Power consumption under load was relatively high for the Radeon HD 7990, but we were expecting that to be the case considering the card is packing 8.6B transistors and 6GB of memory.

[...]

With that said, the Radeon HD 7990 is immensely powerful and it runs surprisingly cool and quiet, considering its performance.

[...]

The AMD Radeon HD 7990 is a beast.

http://hothardware.com/Reviews/AMD-Radeon-HD-7990-Review-The-Silent-Beast/?page=13

 

[The Q6660 Inside]

Once SteamRoller comes around, we shall see who has the better IPC

Cannot wait for some more P̶e̶n̶t̶i̶u̶m̶ ̶4̶ ̶P̶r̶e̶s̶s̶c̶o̶t̶ ̶v̶s̶ ̶A̶t̶h̶l̶o̶n̶ ̶6̶4̶ ̶W̶i̶n̶c̶h̶e̶s̶t̶e̶r̶ ̶ Intel i5 Haswell vs AMD FX Steamrolller action!

 

[de5_Roy]

As said the 35W TDP difference is considering that one is a K series. If you don't overclock the power consumptions are similar. The difference is due to slightly higher stock clocks.
http://www.vortez.net/articles_file/19865_pwrl.png

... you coulda at least posted the review link. anyway, interesting info. in the review, it says right at the top of the page that motherboard models affect power draw. the similar peak draw seems unusual because no one else seem to have measured the same numbers. if correct, it makes a10 5700 as unsuitable as 5800k for sff. i think 5700 and 5800k's similar consumption can be explained. xbitlabs noticed (and amd confirmed) that when certain artificial loads are put on the apus, protection kicks in and throttles clockrate under base rate as long as turbo core is enabled. 5800k coulda been throttling harder under load. i could be wrong about this though.
here's a hwmonitor screenshot from guru3d's a10 5700 review:
http://www.guru3d.com/articles_pages/amd_a10_5700k_review_apu,8.html
you can see the apu's power use @stock in the power section - not similar to a10 5800k in any way.
http://www.guru3d.com/articles_pages/amd_a10_5700k_review_apu,1.html
there's a significant difference between 5700 and 5800k's power use.
http://www.goldfries.com/images/hwreviews/2013/amd-a10/hwmonitor.jpg
this one is from a lesser known review site (my web filter doesn't rate it highly so you've been warned.), but it shows a10 5800k's power use @stock. also keep in mind that they use a gigabyte board while guru3d used asus iirc and vortez used asrock.

About the HD 7990 vs GTX 690 vs GTX titan, a resume of what said:

Spoiler

the HD 7990 is faster than the GTX 690 in most tests that we ran, but not in all of them.

[...]

Even more surprising is the card's lack of noise and heat, [...] We'd say all in all it's probably just a bit quieter than the GTX 690, and gets about as hot.

http://www.maximumpc.com/article/videocards/amd_radeon_hd_7990_first_look

Spoiler

As the results on the preceding pages have shown, the Radeon HD 7990 puts up some impressive frame rates, outpacing the similarly priced GeForce GTX 690 and GTX Titan more often than not.

[...]

Power consumption under load was relatively high for the Radeon HD 7990, but we were expecting that to be the case considering the card is packing 8.6B transistors and 6GB of memory.

[...]

With that said, the Radeon HD 7990 is immensely powerful and it runs surprisingly cool and quiet, considering its performance.

[...]

The AMD Radeon HD 7990 is a beast.

http://hothardware.com/Reviews/AMD-Radeon-HD-7990-Review-The-Silent-Beast/?page=13

HH tested frame time and their conclusion is similar to others. mpc... just slapped things together.thanks for the synopses, changes nothing though.

 

[gamerk316]

It seems no PC will be able to keep up with that for some time to come. SR is only going to be able to run 44 simultaneous threads, and tier 1 GPUs (HD 7950+) allow running 4 compute threads at once...for a total of 48.

Wut?

Explain how an 8 core processor can run 44 threads at a time. I REALLY want to hear this explanation.

 

[8350rocks]

It seems no PC will be able to keep up with that for some time to come. SR is only going to be able to run 44 simultaneous threads, and tier 1 GPUs (HD 7950+) allow running 4 compute threads at once...for a total of 48.

Wut?

Explain how an 8 core processor can run 44 threads at a time. I REALLY want to hear this explanation.

Easily...

Each core is capable of 4 threads at once (4x8=32)
Each FPU is capable of running 3 threads at once (3x4=12)
Total:44

Those are engineered maximums...they cannot be exceeded. Now, they might not hit 44 at once without supremely good software utilization, but the capability is there.

By contrast, the i7-3770k can run 36 threads at once:
4 per real core and 2 per HT process: (6x4=24)
3 FP calculations per core (2 core/1 HT): (3x4=12)
Total: 36

The i7-4770k is rumored to be able to run 40 threads at once. though I am not sure of the breakdown.

 

[mayankleoboy1]

^ source ?

 

[truegenius]

^ i need links to that explanation

 

[gamerk316]

Each core is capable of 4 threads at once

Someone doesn't have a clue how CPU's work, apparently.

One set of registers = One thread on the core. 8 cores = 8 threads.

So yeah, I think you're talking out of your behind just a little bit...

 

[8350rocks]

Each core is capable of 4 threads at once

Someone doesn't have a clue how CPU's work, apparently.

One set of registers = One thread on the core. 8 cores = 8 threads.

So yeah, I think you're talking out of your behind just a little bit...

It's CPU multi-tasking...this is EE 101 stuff guys... Back when you had a single core CPU...(some of you may not have been old enough to remember...not sure of ages here)...Your CPU is capable of multi-tasking. This means that the maximum number of threads a CPU can run at once are engineered maximums before the pipelines fill up. There is an acceptable limit to the amount of clocktime the engineers originally designed to allow to be split up under heavy load.

By concurrently, I don't mean 100% from each core to each thread. I am talking about one thread receiving closer to 60% clocktime and the other 3 splitting up approximately 40% or so...(not sure on the % split there...could be 50/50 or 70/30 etc. depending on design limits)...so while work is being done on background threads in each core, the foreground thread...(or highest prioritized thread)...is receiving the most resources until it is completed, then the CPU moves on to the next thread in order of priority, and continues working on the same threads it was while picking up one more in the open pipeline. If one thread is completed while running in the background, that pipeline hands the next thread to the scheduler, and the process is begun all over again.

This was the old school way they made your CPU able to run multiple programs at once...and why you noticed your single core CPU bogging down after more than 2-4 processes were open at once...do you guys think they lost 15 year old technology all of the sudden?

It won't execute them all at 100%, in fact, at those loads, no single thread would get 100% resources from any core/FPU...the % would be quite a bit less than that. Though rather than processing one single thread per core, you're doing more work in the same time frame...and enhancing the speed with which you can complete large voluminous tasks.

For example...on the i5-3570k...(without HT obviously)...you could have something like these programs running all at once:

Internet Explorer (1+ threads)
Windows Media Player (streaming) (1+ threads)
MS Word (1+ threads)
iTunes (downloading) (1+ threads)
Skyrim (1+ threads)

Now, by your logic, you could only run 4 of those or fewer...but your PC doesn't stop you and say..."Oh, no...you can't do that sir! We don't have 5 cores!" does it? No...clearly not. So what happens is, the programs scheduled with highest priority receive more resources than the background processes.

I am astonished that you guys don't already know this, or have forgotten about it. This is simple distribution of resources.

 

[gamerk316]

It's CPU multi-tasking...this is EE 101 stuff guys... Back when you had a single core CPU...(some of you may not have been old enough to remember...not sure of ages here)...Your CPU is capable of multi-tasking. This means that the maximum number of threads a CPU can run at once are engineered maximums before the pipelines fill up. There is an acceptable limit to the amount of clocktime the engineers originally designed to allow to be split up under heavy load.

Multithreading != Multiprocessing.

Multithreading: Using instruction level parallization to keep a single CPU pipeline moving at an optimal rate.
Multiprocessing: Using multiple CPU's or CPU cores to run multiple threads at the same time.

SMT is capable of running multiple threads on a single CPU, but aside from Intel Hyperthreading, isn't done on X86 CPU's. You are limited to run one thread context at a time per set of registers. Basically, the OS scheduler sends out any threads that are "ready to run" (non-blocked), and the CPU thread scheduler will scheduler appropriately. Under a Windows OS, the highest priority threads run, until they either finish, or block (cache miss, for example). Priorities get adjusted by the OS the next time the threads go out to the CPU.

For example...on the i5-3570k...(without HT obviously)...you could have something like these programs running all at once:

Internet Explorer (1+ threads)
Windows Media Player (streaming) (1+ threads)
MS Word (1+ threads)
iTunes (downloading) (1+ threads)
Skyrim (1+ threads)

They don't "run at once", they swap as needed, per the OS scheduler. Thats the entire point of preemptive multithreading. Using your example, at any one instant, only four of these applications could actually be running. The other one would be waiting until either its priority was greater then another application, or another application was blocked. But at any instant, you are limited to running FOUR threads at a time.

This also means that threads don't "share clocktime"; they are scheduled by the OS, given a priority, and the CPU will run the threads accordingly. Now, if a thread blocks (for any number of reasons), modern CPU's are smart enough to run some other thread instead, but there isn't any magical "Oh, this thread is getting 70% runtime, and this one gets 30%" that is determined beforehand.