With the current rate of Intel CPU performance increases, could AMD be catching up?

[Phynaz]

Chips weren't that much cheaper to design back then.

They are immensely more expensive to design today.

Think about how much larger the design teams are. The cost of a mask set today is 10x what it was a decade ago.

When you say that Intel had 40% margins back then, you are forgetting about volume. All that R&D is spread over a much larger number of chips today.

Just look at the charts of R&D spending that IDC has posted. I don't know an exact number, but if you take the 11% 2010 - 2011 increase and extrapolate over a decade that comes out to a nearly 3x cost increase.

 

[AtenRa]

Does this mean that AMD now has a chance to catch up performance wise?

They(AMD) already have catched up in performance (Throughtput). Just see AMD Trinity (Dual Module 4 Threads) against Core i3 2100/3220 (Dual Core 4 Threads) in Multithreaded apps. Even in newer games the Trinity is faster (BF3, Crysis 3 etc).

For almost the same CPU Core size(32nm) they have more Throughput performance than Intel, if AMD had 20nm today they would be even closer in performance/watt.

In GPU performance they are still way infront of Intel and i dont see even Haswell being able to give the lead to Intel(in desktop) against Trinity/Richland.

The only thing is that AMD is not focusing on the HighEnd desktop, so dont expect miracles there. They push the APUs more and thats what they wanted from the start, remember "FUSION is the future" and OpenCL through HSA.

Intel following in AMDs footsteps the last years, Haswell GT3 iGPU is almost 50% of the CPU die size, like Trinity. They know that the future is not single thread performance and that IPC cannot scale like it used to be.

 

[ShintaiDK]

They(AMD) already have catched up in performance (Throughtput). Just see AMD Trinity (Dual Module 4 Threads) against Core i3 2100/3220 (Dual Core 4 Threads) in Multithreaded apps. Even in newer games the Trinity is faster (BF3, Crysis 3 etc).

For almost the same CPU Core size(32nm) they have more Throughput performance than Intel, if AMD had 20nm today they would be even closer in performance/watt.

In GPU performance they are still way infront of Intel and i dont see even Haswell being able to give the lead to Intel(in desktop) against Trinity/Richland.

The only thing is that AMD is not focusing on the HighEnd desktop, so dont expect miracles there. They push the APUs more and thats what they wanted from the start, remember "FUSION is the future" and OpenCL through HSA.

Intel following in AMDs footsteps the last years, Haswell GT3 iGPU is almost 50% of the CPU die size, like Trinity. They know that the future is not single thread performance and that IPC cannot scale like it used to be.

Wow thats a lot of marketing. But just no, AMD have not and will not catch up. You also forget the performance/watt part.

Not to mention Fusion is dead and HSA already looks like a flop. OpenCL is about to end like OpenGL. DirectCompute already won that game.

 

[AtenRa]

Wow thats a lot of marketing. But just no, AMD have not and will not catch up.

From
AMD A10-5800K & A8-5600K Review: Trinity on the Desktop, Part 2

Cinebench (An Intel Optimized app)
Trinity is faster than Core i3 3220
http://images.anandtech.com/graphs/graph6347/50398.png

7zip
http://images.anandtech.com/graphs/graph6347/50401.png

3d smax
http://images.anandtech.com/graphs/graph6347/50392.png

Povray
http://images.anandtech.com/graphs/graph6347/50393.png

Par2
http://images.anandtech.com/graphs/graph6347/50394.png

Excel
http://images.anandtech.com/graphs/graph6347/50395.png

x2664 HD v3.03
http://images.anandtech.com/graphs/graph6347/50400.png

It seams to me they have catched up in perfromance (they are even in the lead most of the time).

You also forget the performance/watt part.

I havent forget about it but the OP was talking about performance alone.

Not to mention Fusion is dead and HSA already looks like a flop. OpenCL is about to end like OpenGL. DirectCompute already won that game.

LoL, simple NO. You are trying so hard to promote AVX2 as the next best thing ever hapend in the PC ,but OpenCL has already a lot of apps you can use TODAY and more are coming every year.

http://openclnews.com/apps

You should try some and see how your Intel APU perform 😉

Also just to add that Tensilica has Joined HSA Foundation and that HSA is close to setting hardware specs.
So not only OpenCL is not dead but HSA is starting to produce some reasults when AVX2 is not even availiable yet.

 

[Fjodor2001]

Do I really have to spell it out for you?

Since you never really said much in your original post, please do.

No. But it's improbable.

Because...?

And you think AMD of all companies is going to be the one to run them over? How laughable.

They've done it before. A few years before that they were also counted out and standing in the shadow of Intel.

Pretty sure AMD's running a similar average right now [8% CPU performance increase from one CPU generation to the next]. And you're absolutely short-sighted if you're looking at performance as your only measuring stick.

See: http://www.anandtech.com/show/6347/amd-a10-5800k-a8-5600k-review-trinity-on-the-desktop-part-2/2

Single threaded performance went up 21% going from Llano (A8-3850) to Trinity (A10-5800K).

Also see: http://www.anandtech.com/show/6347/amd-a10-5800k-a8-5600k-review-trinity-on-the-desktop-part-2/3

SYSMark 2012 Overall performance up 29%.

For example, Intel could very, very easily release a higher clocked version of Ivy Bridge or Haswell at similar TDP's to AMD's FX. Ta-freaking-da, you now have that dumb performance increase you're looking for. And if you think that Intel doesn't have the fab maturity to do this, you're clueless.

The fact is that TDP goes up very quickly when increasing the CPU frequency on IB. AMD A8-5800K is at 100 W, Intel 3770K is at 77W (and Haswell 4770K will be at 84 W). That gives Intel a 16-23 W advantage. Not much. And that's even with Intel being on 22 nm vs AMD at 32 nm!

Also, AMD has an iGPU advantage compared to Intel. If Intel wants to catch up, it means the TDP of Intel CPUs will increase due to that (adding more GPU cores / EUs).

You and everyone else needs to stop blindly looking at performance as the only thing that matters. You're an enthusiast. Learn how to overclock.

I've never said that performance is the only thing that matters. But it's the topic of this thread.

Regarding the OC comment I'm not really sure what you're point is. Since it's possible to OC, performance increases brought by CPU improvements do not matter? 😕

Intel's already got a fire under their rear: ARM. Make all the vapid comments about Intel "slacking off" that you'd like. They're not.

Yes, they are getting competition in that sector too. That does not mean they cannot get competition in other sectors at the same time.

Perhaps that's the problem for Intel. They are now getting competition on several fronts, and currently they seem to focus primarily on the ARM problem, so they are leaving the other flanks unprotected.

 

[mrmt]

URL]The fact is that TDP goes up very quickly when increasing the CPU frequency on IB. AMD A8-5800K is at 100 W, Intel 3770K is at 77W (and Haswell 4770K will be at 84 W). That gives Intel a 16-23 W advantage. Not much. And that's even with Intel being on 22 nm vs AMD at 32 nm!

Oh dear, this makes Atera's number twisting becomes child's play. So now you take Intel quad-core, something that is definetly in another league performance wise than any of AMD's APU, and by decree Intel advantage in power efficiency is now gone?

I can barely wait to see you make the die-size advantage disappear.

 

[SPBHM]

They(AMD) already have catched up in performance (Throughtput). Just see AMD Trinity (Dual Module 4 Threads) against Core i3 2100/3220 (Dual Core 4 Threads) in Multithreaded apps. Even in newer games the Trinity is faster (BF3, Crysis 3 etc)

i3 2100 still looks faster in BF3 even when compared to the clearly faster (compared to trinity for gaming) 4300!
also, when the 2100 was released, 2 years ago, the PII X4 955 was being sold at a similar price I think, and it was faster for MT (like video encoding), so I don't really think it has changed much, the i3 still offers clearly better perf/watt when it comes to CPU,

 

[Fjodor2001]

Oh dear, this makes Atera's number twisting becomes child's play. So now you take Intel quad-core, something that is definetly in another league performance wise than any of AMD's APU, and by decree Intel advantage in power efficiency is now gone?

Fair enough, the AMD 4 core / 2 module solution is not directly comparable to Intel 4C. Instead it's somewhere inbetween an Intel 2C/4T (with HT) and pure 4C.

And in multithreaded performance this shows. However in Cinebench 11.5 single threaded performance the 3770K is "only" about 50% faster than the A10-5800K.

But anyway, the main point is that if AMD keep improving CPU performance by 20-30% per CPU generation (as shown above) when Intel only improves by ~8%, then AMD will eventually catch up.

 

[ShintaiDK]

But anyway, the main point is that if AMD keep improving CPU performance by 20-30% per CPU generation (as shown above) when Intel only improves by ~8%, then AMD will eventually catch up.

Those numbers is abit off.

But also flawed as conclusion. 10% to index 100 is what? 5% to index 200 is what?

 

[mrmt]

But anyway, the main point is that if AMD keep improving CPU performance by 20-30% per CPU generation (as shown above) when Intel only improves by ~8%, then AMD will eventually catch up.

What you see in Piledriver has more to do with GLF learning curve in improving 32nm process and the marketing shenanigan of breaching their own TDP specifications than with AMD tweaking the architecture, and by the way Bulldozer was wasn't a great starting point. Once you factor those things out, you are left with the same 5-10% increases that Intel gives us in each generation.

 

[NTMBK]

What you see in Piledriver has more to do with GLF learning curve in improving 32nm process and the marketing shenanigan of breaching their own TDP specifications than with AMD tweaking the architecture, and by the way Bulldozer was wasn't a great starting point. Once you factor those things out, you are left with the same 5-10% increases that Intel gives us in each generation.

Piledriver was more of an improvement from BD than SB was from IB- and IB got to use an entirely new process.

 

[Pilum]

They(AMD) already have catched up in performance (Throughtput). Just see AMD Trinity (Dual Module 4 Threads) against Core i3 2100/3220 (Dual Core 4 Threads) in Multithreaded apps.

Which very few people use to any signficant extent.

Even in newer games the Trinity is faster (BF3, Crysis 3 etc).

In a very few modern games. There are many modern games which still require good single-threaded performance, and of course there are hundreds of older games still being played which prefer single-threaded performance to many threads.

In GPU performance they are still way infront of Intel and i dont see even Haswell being able to give the lead to Intel(in desktop) against Trinity/Richland.

Which seems to be irrelevant, as Intel with its weak iGPUs makes lots of money, while AMD with its strong iGPUs doesn't. Which tells us not many people care about the iGPUs. Which in turn tells us that Intel is making the right decisions, in offering products people actually want to spend money on.

 

[Pilum]

Cinebench (An Intel Optimized app)
Trinity is faster than Core i3 3220

Relevance? Do you really think that people who spend between $1000 and $3700 on a single program then run said program on a weak-ass AMD APU?

You're a funny guy.

 

[Homeles]

Since you never really said much in your original post, please do.

Intel was blindly pursuing clock speed at any cost. They no longer do that -- ergo, they have learned from their mistake.

They've done it before. A few years before that they were also counted out and standing in the shadow of Intel.

They certainly won't be doing it anytime soon. Steamroller, at best, can bring them close to Intel on a performance level, while still being woefully behind with power consumption. Excavator is allegedly working on the power consumption part, at the expense of clock speed.

See: http://www.anandtech.com/show/6347/amd-a10-5800k-a8-5600k-review-trinity-on-the-desktop-part-2/2

Single threaded performance went up 21% going from Llano (A8-3850) to Trinity (A10-5800K).

Also see: http://www.anandtech.com/show/6347/amd-a10-5800k-a8-5600k-review-trinity-on-the-desktop-part-2/3

SYSMark 2012 Overall performance up 29%.

Performance per watt went nowhere.

And when you look at Phenom II -> Zambezi -> Vishera, that average does come out to around the 8% number, considering that Bulldozer went backwards.

The fact is that TDP goes up very quickly when increasing the CPU frequency on IB. AMD A8-5800K is at 100 W, Intel 3770K is at 77W (and Haswell 4770K will be at 84 W). That gives Intel a 16-23 W advantage. Not much. And that's even with Intel being on 22 nm vs AMD at 32 nm!

The point is that Intel's shrinking their power usage while simultaneously increasing performance. It's silly to be complaining about Intel not going anywhere with performance, because they're making huge leaps in performance per watt. Intel's top-level power rating has changed -- why can't the performance junkies that are whining in this thread and everywhere else on the internet see this?

Ivy Bridge is an oddball in a rather long history of large gains in clock speed and performance. Claiming it's the end of the world based on one data point is ridiculous.

Also, AMD has an iGPU advantage compared to Intel. If Intel wants to catch up, it means the TDP of Intel CPUs will increase due to that (adding more GPU cores / EUs).

That has no bearing on the progress being made with their CPUs.

I've never said that performance is the only thing that matters. But it's the topic of this thread.

Regarding the OC comment I'm not really sure what you're point is. Since it's possible to OC, performance increases brought by CPU improvements do not matter? 😕

It's about not being bound by Intel's TDP ratings. If Intel kept performance stagnant, but managed to drop power consumption by 75%, by the logic that's being thrown around here, they wouldn't be moving anywhere.

Yes, they are getting competition in that sector too. That does not mean they cannot get competition in other sectors at the same time.

Perhaps that's the problem for Intel. They are now getting competition on several fronts, and currently they seem to focus primarily on the ARM problem, so they are leaving the other flanks unprotected.

Worst case scenario, Intel has to adjust their price. They aren't going to hold onto those wide margins forever.

 

[mrmt]

Piledriver was more of an improvement from BD than SB was from IB- and IB got to use an entirely new process.

Really? I didn't see 20-30% in power consumption improvements in AMD line up.

 

[VirtualLarry]

Wow thats a lot of marketing. But just no, AMD have not and will not catch up. You also forget the performance/watt part.

Fanboy reading comprehension fail. He said "performance", not performance/watt. Obviously, because Intel has leading process technology, they are going to be ahead in performance/watt. But the very fact that AMD is indeed competitive in performance, speaks volumes about the quality of their engineering.

Not to mention Fusion is dead and HSA already looks like a flop. OpenCL is about to end like OpenGL. DirectCompute already won that game.

Fanboy proponents of doom and gloom. How is Fusion dead? It's selling, even today!

If I were to use your logic, I would declare AVX2 to be a flop already. I mean, look at how few apps on the market use AVX2 already.

Extending your logic likewise, I would say that Intel is doomed, because there are so few AVX2 apps on the market.

 

[386user]

IGP is where the battle is

if haswell gt3 is better than richland..then no

value is a whole different story though

 

[colonelciller]

"The point is that Intel's shrinking their power usage while simultaneously increasing performance."

why should desktop users care about small decreases in power usage?
is that $4 in extra electricity per year going to break the bank or something?

 

[colonelciller]

Really? I didn't see 20-30% in power consumption improvements in AMD line up.

Straw man argument if I'm not mistaken.
AMD case -- more improvement per generation compared to intel (speaking of computing performance)

Intel counterargument -- we have lower power consumption.

Rational mind:
*CPU cooler - Check
*Desktop plugged into wall - check
*$4 higher electricity bill - irrelevant

 

[Fjodor2001]

Those numbers is abit off.

In what way?

But also flawed as conclusion. 10% to index 100 is what? 5% to index 200 is what?

Doesn't matter.

Index_AMD : Starting performance index for AMD
Index_Intel : Starting performance index for Intel
Index_AMD and Index_Intel are normalized to some common index.

P_AMD : Performance increase per year for AMD CPUs (P_AMD>1)
P_Intel : Performance increase per year for Intel CPUs (P_Intel>1)

Then this will always be true if P_AMD > P_Intel, regardless of Index_AMD and Index_Intel:

Index_AMD*P_AMD^n > Index_AMD*P_Intel^n

when n -> infinity

😎

 

[ShintaiDK]

In what way?

Did you calculate for the baseclock for FX8350 went up 400Mhz compared to the FX8150?

And IPC didnt increase 20-30% either. Not to mention the chip run out of spec with power consumption as well.

 

[mrmt]

Rational mind:
*CPU cooler - Check
*Desktop plugged into wall - check
*$4 higher electricity bill - irrelevant

The lower power consumption means that there is enough headroom for extra frequency if Intel needs, meaning that there are improvements, it's just that those improvements aren't where you want them to be, but they are there.

As a matter of fact, Intel's 32nm and 22nm chips can reach clocks as high as any AMD chips but with much less power consumption. It's just marketing that prevents a 4 or even a 4.2 IVB-4C SKU to be launched.

 

[frozentundra123456]

Fair enough, the AMD 4 core / 2 module solution is not directly comparable to Intel 4C. Instead it's somewhere inbetween an Intel 2C/4T (with HT) and pure 4C.

And in multithreaded performance this shows. However in Cinebench 11.5 single threaded performance the 3770K is "only" about 50% faster than the A10-5800K.

But anyway, the main point is that if AMD keep improving CPU performance by 20-30% per CPU generation (as shown above) when Intel only improves by ~8%, then AMD will eventually catch up.

Yes, but you can extrapolate a lot of things that never happen.

 

[frozentundra123456]

Fanboy reading comprehension fail. He said "performance", not performance/watt. Obviously, because Intel has leading process technology, they are going to be ahead in performance/watt. But the very fact that AMD is indeed competitive in performance, speaks volumes about the quality of their engineering.


Fanboy proponents of doom and gloom. How is Fusion dead? It's selling, even today!

If I were to use your logic, I would declare AVX2 to be a flop already. I mean, look at how few apps on the market use AVX2 already.

Extending your logic likewise, I would say that Intel is doomed, because there are so few AVX2 apps on the market.

Not a valid comparison. Fusion has been on the market for several years and AMD has lost market share during that time. Haswell is not even available yet.

 

[Piroko]

In a very few modern games. There are many modern games which still require good single-threaded performance, and of course there are hundreds of older games still being played which prefer single-threaded performance to many threads.

To be fair, there aren't that many older games which need single thread performance above what a decently clocked Trinity can provide. Starcraft 2 and Arma 2 (+3?) wants more, maybe Diablo 3 and Skyrim? A lot of games will definitely pick up more performance, but will already be beyond stable 60 fps.