To most people its just as obvious Intel CPU's are superior to AMD CPU's just as nVidia GPU's are superior to AMD GPU's.
However, the superiority of nVidia GPU's are AMD GPU's aren't nearly as substantial as Intel's CPU's over AMD. The real problem with AMD GPU's are with OEM's, software quality and product cycle alignment with nVidia.
Unlike AMD CPU's, AMD GPU's are exceptionally well engineered, just as well as nVidia's, albeit entirely different approaches. This makes AMD GPU's superior for datamining (Bitcoin, etc) where a competing nVidia GPU would need to be TESLA-based, and even then not as effective at mass execution. But the downside is AMD's less powerful but more plentiful cores are no competition to nVidia's more powerful but less plentiful core's in raw texel throughput, making them better for gaming.
Which is interesting when you consider AMD GPU's are in all the game consoles. The answer to that: they're cheaper. Just like their CPU's. Console OEM's would never hit their price target using Intel CPU's and nVidia GPU's. That approach is why the original XBOX wasn't as successful as the XBOX 360. It was just too expensive compared to the PS2 and Gamecube, and even at that price Microsoft lost tons of money on it, somewhere in the order of $100/ea at launch.
I would agree that nVidia's software and marketing are superior, but GCN is pretty close to the exact same µarch that nVidia has been using. Personally, I like aspects of both GPU competitors. There are things that you can find to dislike about both of them if you look hard enough. Anyone remember the 5000-series? Everyone has their 5000-series at some time... no one hits 100%.
I wonder hypothetically if the K10 would have had 2x the number of modules it does have, would it have been enough? I figured that's where they were going trying to save space and utilize resources more efficiently. It seemed like a key component like a process shrink or new silicon technology was missing from K10. It's a shame, but Intel went through the same growing pains with P4 and now we have enhanced Pentium-like micros returning in Core to the far foreseeable future. It really is a shame, we could've used some upstart to kickstart a performance race again just like the K6 vs. P2 (P3), and the K7 vs. P4... Those were the days...
That's only because they are writing the revenue from the game consoles for the GPU division, where in reality those SoCs both have a GPU and a CPU and AMD would not have won those contracts without its Jaguar CPU.
Which shows the absurdity of the current reporting, and why the reorganisation in reports coming from next quarter is necessary.
True, but we'll no longer see the split between machines where AMD is the central processor and add-on cards as it'll be CPU/APU/GPU lumped into one big "traditional" sack, I wish they'd just split out the new business from the GPU instead. In any case, I hope they manage to sell that inventory. It keeps going up...
I agree that their reporting needs a revamp. I would still expect to see their GPU division still being profitable. Their GPU division should actually get some credit from the CPU side, which would make the CPUs look worse. This would really sum up their whole current existence and marketing. People buy the CPUs because the GPUs keep them afloat performance wise.
The chipset division needs to be graded financially. They have been sitting on their asses too long. Name an innovation that came from there. They could do more there on the client and server side and make the whole platform more competitive. Intel is spanking them there for no good reason.
Hopefully when their next chipset comes out, with their next architecture in 2016, it will be a lot better. Also to note, AM3+ motherboards are getting pretty outdated.
The current 900 series chipset is essentially the exact same as the 800 series chipsets, which in turn were only a minor improvement over the 700 series chipsets. In-fact the 980G is a rebadged 880G chipset and some 960/970G Chipsets are re-badged 860/870G chipsets, IGP and all. Some OEM's however like Asus re-purposed the 5-6 year old 760G chipset to be AM3+ compatible like with the ASUS M5A78L-M/USB3 motherboard.
AMD has been stupidly lethargic with it's chipset innovation.
Here we are in the middle of 2014 and AMD's chipsets STILL don't support PCI-E 3.0 and USB 3.0 natively, not to mention the lack of Sata 3.2, heck even Sata 3.1, if you wan't those features they are included via secondary chips, which increases board complexity, thus the BoM and hence price and possibly impacting reliability.
Interests are basically the same thing as profits, i.e. money paid to investors instead of workers and suppliers. But usually you don't have to pay taxes before handing out interest payment. So most modern companies are actively trying to pay as much interest as they possibly can, and only keep a little money for the official profits. For AMD: They paid 177M$ of interest in 2013, but in 2009 (when their CPU business was still going stronger) they even paid 439M$ in interest. At the same time, AMD only paid 9M$ of income taxes in 2013, so thats just 5% of the money handed to their investors/creditors.
So really, the interesting result from a financial report are always the EBIT figures (Earnings before interest and taxes), not the largely random net income.
Wow your revenue went up $400 million but your profits only went up $80 million in the gpu division. Looks like they are not making a lot of margins on those game console chips, what is that, roughly 20%? I know they have to buy the chips from tmsc and gf, but you should barely have any R&D costs since you reused your jaguar and video card design, and those costs should have happened months ago, you are now just printing silicon.
The margins on console chips were debated a lot when nvidia took them as an excuse for the fact that they didn't have the chance to enter console market due to lack of an integrated CPU+GPU architecture (whose advantage is only to keep costs low). However, many interpreted those claims only as an excuse with no reference to reality. Many thought that margins were really high, but nvidia could not offer anything (as it couldn't).
Reality is that Sony and MS are not stupid and aimed since day 1 of their R&D for the next-gen consoles at lowering the costs as much as possible. And APUs are a good way to obtain that. And they knew. And so they wanted the APU to cost much less than an architecture with an Intel CPU+nvidia GPU. Both Intel and nvidia could lower their margins enough to create a cheap discrete architecture. But Sony and MS want even lower costs. AMD was there to satisfy both of them. Whoever think about this could never think that AMD would have high margins. Sony and MS were not surely keen to pay them (discrete architecture would be the best solution then at the same costs).
Someone (like me) said that. But hordes of fanboys just used nvidia statements to prove that AMD was the winner, the best, that was going to have great profits for the next years only based on console chip shippings. But, as we see, they were wrong. And we still do not know if the future APUs shrinks have to be paid by AMD or MS/Sony.
In the end, though nvidia statements on too much lower margins were a marketing necessity as they had nothing to offer to equal APUs costs, they were not telling bullshit.
My question is: seen the high profits and the fact that this "stream" of cash coming from console chips is now well consolidated (and will be similar for the next quarters) but the company could not make profits, when ever they are?
The executives have already explained this in the earlier Earnings Call, that the non-recurring engineering expenses of the semi-custom chips *were paid* by the customers. So basically the margin of the semi-custom chips is likely nothing but just profit.
Last year the Graphics revenue was 320mil and that includes the old consoles ,so it's safe to say that GPU this quarter was under 300 mil. But it is sad that the CPU side is fading away like that. They need to target "console like performance"with their 20nm APUs (or 14nm if they end up skipping 20nm) as soon as possible and then get those new cores out fast.
I don't think the lack of availability of the A8-7600 has helped as it appears to be the one APU that people are looking to purchase. A more custom approach to their APUs could also help; they did say they would create chips based on a modular design with different combinations of compute units to suit requirements but I imagine this isn't very high on their list of priorities right now.
Providing "console like performance" would be a difficult task; the CPU performance is there with Steamroller but even with "only" 512 cores, there's a lack of memory bandwidth. I'm not sure raising the number of cores to 768 would even get the APU to XBox One levels; the XBox One has a quad channel memory interface pushing 68GB/s which is practically double that of a dual-channel PC, so there's going to be diminishing returns for adding extra cores without the bandwidth to back them up, and that's without considering ESRAM or stacked memory.
What is the typical use case performance delta between Haswell and Ivy Bridge at the same frequency? Also, what is the frequency delta between the two, what is the performance delta there? I think Intel's last several years have shown what they think they need in terms of performance to sell CPU's. They've actually focused on other markets as a result too. Look at their restructuring...
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
31 Comments
Back to Article
Loki726 - Thursday, July 17, 2014 - link
Interesting to see the GPU division surpassing the CPU division in revenue....eanazag - Friday, July 18, 2014 - link
Not a total surprise as their products match up this way also.nunya112 - Saturday, July 19, 2014 - link
No surprise at all.The CPU side SUXS BALLS... HAIRY BALLS.... BIG HAIRY STINKY BALLS
And the GPU division ABSOLUTELY KILLING IT !
Samus - Sunday, July 20, 2014 - link
To most people its just as obvious Intel CPU's are superior to AMD CPU's just as nVidia GPU's are superior to AMD GPU's.However, the superiority of nVidia GPU's are AMD GPU's aren't nearly as substantial as Intel's CPU's over AMD. The real problem with AMD GPU's are with OEM's, software quality and product cycle alignment with nVidia.
Unlike AMD CPU's, AMD GPU's are exceptionally well engineered, just as well as nVidia's, albeit entirely different approaches. This makes AMD GPU's superior for datamining (Bitcoin, etc) where a competing nVidia GPU would need to be TESLA-based, and even then not as effective at mass execution. But the downside is AMD's less powerful but more plentiful cores are no competition to nVidia's more powerful but less plentiful core's in raw texel throughput, making them better for gaming.
Which is interesting when you consider AMD GPU's are in all the game consoles. The answer to that: they're cheaper. Just like their CPU's. Console OEM's would never hit their price target using Intel CPU's and nVidia GPU's. That approach is why the original XBOX wasn't as successful as the XBOX 360. It was just too expensive compared to the PS2 and Gamecube, and even at that price Microsoft lost tons of money on it, somewhere in the order of $100/ea at launch.
hardrock_ram - Monday, July 21, 2014 - link
What an insight ..samirotiv - Wednesday, July 23, 2014 - link
People who don't know where not to put apostrophes shouldn't be allowed to exist.Demiurge - Friday, July 25, 2014 - link
I would agree that nVidia's software and marketing are superior, but GCN is pretty close to the exact same µarch that nVidia has been using. Personally, I like aspects of both GPU competitors. There are things that you can find to dislike about both of them if you look hard enough. Anyone remember the 5000-series? Everyone has their 5000-series at some time... no one hits 100%.I wonder hypothetically if the K10 would have had 2x the number of modules it does have, would it have been enough? I figured that's where they were going trying to save space and utilize resources more efficiently. It seemed like a key component like a process shrink or new silicon technology was missing from K10. It's a shame, but Intel went through the same growing pains with P4 and now we have enhanced Pentium-like micros returning in Core to the far foreseeable future. It really is a shame, we could've used some upstart to kickstart a performance race again just like the K6 vs. P2 (P3), and the K7 vs. P4... Those were the days...
Gizmosis350k - Saturday, July 26, 2014 - link
Nvidia GPUs are not better than AMD - AMD is the bestaryonoco - Thursday, July 17, 2014 - link
That's only because they are writing the revenue from the game consoles for the GPU division, where in reality those SoCs both have a GPU and a CPU and AMD would not have won those contracts without its Jaguar CPU.Which shows the absurdity of the current reporting, and why the reorganisation in reports coming from next quarter is necessary.
Kjella - Friday, July 18, 2014 - link
True, but we'll no longer see the split between machines where AMD is the central processor and add-on cards as it'll be CPU/APU/GPU lumped into one big "traditional" sack, I wish they'd just split out the new business from the GPU instead. In any case, I hope they manage to sell that inventory. It keeps going up...eanazag - Friday, July 18, 2014 - link
I agree that their reporting needs a revamp. I would still expect to see their GPU division still being profitable. Their GPU division should actually get some credit from the CPU side, which would make the CPUs look worse. This would really sum up their whole current existence and marketing. People buy the CPUs because the GPUs keep them afloat performance wise.The chipset division needs to be graded financially. They have been sitting on their asses too long. Name an innovation that came from there. They could do more there on the client and server side and make the whole platform more competitive. Intel is spanking them there for no good reason.
icetorch - Friday, July 18, 2014 - link
Hopefully when their next chipset comes out, with their next architecture in 2016, it will be a lot better. Also to note, AM3+ motherboards are getting pretty outdated.StevoLincolnite - Sunday, July 27, 2014 - link
The current 900 series chipset is essentially the exact same as the 800 series chipsets, which in turn were only a minor improvement over the 700 series chipsets.In-fact the 980G is a rebadged 880G chipset and some 960/970G Chipsets are re-badged 860/870G chipsets, IGP and all.
Some OEM's however like Asus re-purposed the 5-6 year old 760G chipset to be AM3+ compatible like with the ASUS M5A78L-M/USB3 motherboard.
AMD has been stupidly lethargic with it's chipset innovation.
Here we are in the middle of 2014 and AMD's chipsets STILL don't support PCI-E 3.0 and USB 3.0 natively, not to mention the lack of Sata 3.2, heck even Sata 3.1, if you wan't those features they are included via secondary chips, which increases board complexity, thus the BoM and hence price and possibly impacting reliability.
Wreckage - Thursday, July 17, 2014 - link
"Total debt went up from $2.14 billion to $2.21 billion."I think this is the real killer for AMD. The interest payments alone on this debt have to be upwards of a $100 million or so a year.
SuperMecha - Friday, July 18, 2014 - link
It's closer $200 million a year.Death666Angel - Friday, July 18, 2014 - link
How do you pay 10% interest on loans these days. That's more than most struggling European countries pay today.ShieTar - Friday, July 18, 2014 - link
Interests are basically the same thing as profits, i.e. money paid to investors instead of workers and suppliers. But usually you don't have to pay taxes before handing out interest payment. So most modern companies are actively trying to pay as much interest as they possibly can, and only keep a little money for the official profits.For AMD: They paid 177M$ of interest in 2013, but in 2009 (when their CPU business was still going stronger) they even paid 439M$ in interest. At the same time, AMD only paid 9M$ of income taxes in 2013, so thats just 5% of the money handed to their investors/creditors.
So really, the interesting result from a financial report are always the EBIT figures (Earnings before interest and taxes), not the largely random net income.
eanazag - Friday, July 18, 2014 - link
Why wouldn't AMD be looked at like a struggling European country? Most of my kids have never seen a day where AMD was profitable.Roland00Address - Friday, July 18, 2014 - link
Wow your revenue went up $400 million but your profits only went up $80 million in the gpu division. Looks like they are not making a lot of margins on those game console chips, what is that, roughly 20%? I know they have to buy the chips from tmsc and gf, but you should barely have any R&D costs since you reused your jaguar and video card design, and those costs should have happened months ago, you are now just printing silicon.CiccioB - Friday, July 18, 2014 - link
The margins on console chips were debated a lot when nvidia took them as an excuse for the fact that they didn't have the chance to enter console market due to lack of an integrated CPU+GPU architecture (whose advantage is only to keep costs low).However, many interpreted those claims only as an excuse with no reference to reality. Many thought that margins were really high, but nvidia could not offer anything (as it couldn't).
Reality is that Sony and MS are not stupid and aimed since day 1 of their R&D for the next-gen consoles at lowering the costs as much as possible. And APUs are a good way to obtain that. And they knew. And so they wanted the APU to cost much less than an architecture with an Intel CPU+nvidia GPU. Both Intel and nvidia could lower their margins enough to create a cheap discrete architecture. But Sony and MS want even lower costs. AMD was there to satisfy both of them.
Whoever think about this could never think that AMD would have high margins. Sony and MS were not surely keen to pay them (discrete architecture would be the best solution then at the same costs).
Someone (like me) said that. But hordes of fanboys just used nvidia statements to prove that AMD was the winner, the best, that was going to have great profits for the next years only based on console chip shippings.
But, as we see, they were wrong. And we still do not know if the future APUs shrinks have to be paid by AMD or MS/Sony.
In the end, though nvidia statements on too much lower margins were a marketing necessity as they had nothing to offer to equal APUs costs, they were not telling bullshit.
My question is: seen the high profits and the fact that this "stream" of cash coming from console chips is now well consolidated (and will be similar for the next quarters) but the company could not make profits, when ever they are?
pTmdfx - Friday, July 18, 2014 - link
The executives have already explained this in the earlier Earnings Call, that the non-recurring engineering expenses of the semi-custom chips *were paid* by the customers. So basically the margin of the semi-custom chips is likely nothing but just profit.jjj - Friday, July 18, 2014 - link
Last year the Graphics revenue was 320mil and that includes the old consoles ,so it's safe to say that GPU this quarter was under 300 mil.But it is sad that the CPU side is fading away like that. They need to target "console like performance"with their 20nm APUs (or 14nm if they end up skipping 20nm) as soon as possible and then get those new cores out fast.
silverblue - Friday, July 18, 2014 - link
I don't think the lack of availability of the A8-7600 has helped as it appears to be the one APU that people are looking to purchase. A more custom approach to their APUs could also help; they did say they would create chips based on a modular design with different combinations of compute units to suit requirements but I imagine this isn't very high on their list of priorities right now.Providing "console like performance" would be a difficult task; the CPU performance is there with Steamroller but even with "only" 512 cores, there's a lack of memory bandwidth. I'm not sure raising the number of cores to 768 would even get the APU to XBox One levels; the XBox One has a quad channel memory interface pushing 68GB/s which is practically double that of a dual-channel PC, so there's going to be diminishing returns for adding extra cores without the bandwidth to back them up, and that's without considering ESRAM or stacked memory.
Hrel - Friday, July 18, 2014 - link
so, what? Maybe 6 years till they start selling off business assets and go away entirely.vortmax2 - Friday, July 18, 2014 - link
You don't want them to go away entirely...that's a bad thing for competition, innovation and prices.HighTech4US - Friday, July 18, 2014 - link
Well if they do go away then Charlie and his annoying fanboi flock would also disappear. So that would be a plus.CiccioB - Monday, July 21, 2014 - link
You are supposing that without AMD things remain the same for Intel.But that mey not be true.
Demiurge - Friday, July 25, 2014 - link
What is the typical use case performance delta between Haswell and Ivy Bridge at the same frequency? Also, what is the frequency delta between the two, what is the performance delta there? I think Intel's last several years have shown what they think they need in terms of performance to sell CPU's. They've actually focused on other markets as a result too. Look at their restructuring...Gizmosis350k - Saturday, July 26, 2014 - link
Your post stinks of Intel's cock.nunya112 - Saturday, July 19, 2014 - link
As a share holder I can't be happy at all with this report! My shares have lost value for 2 years straight!Gizmosis350k - Saturday, July 26, 2014 - link
Don't worry the 9th generation of consoles will save AMD