Why can't Intel compete with AMD in GPU Sector?

[Bateluer]

have anyone seen how a dGPU runs on 4K with high Anti Aliasing? it can't even pull off 1 FPS proper. Even the GTX Titan which is so advanced can't even pull off that unless we are talking 1400W power supply and 3 Way SLI

*snip*

I don't think any serious gamer is going to be buying iGPU to run in 4K with AA

Someone called you on both these tidbits as total BS right? I didn't read the entire thread.

 

[tential]

Someone called you on both these tidbits as total BS right? I didn't read the entire thread.

The comment was so ridiculous I don't think anyone felt the need to actually correct him.

But he's relatively new to the hardware section and to hardware history in general (from the topics he's posted anyway).

Anyway

Best optimized game of the year getting 30 FPS on GTX 980 SLI 4K 2x AA (/sarcasm on best optimized game of the year).

So 4K with AA is extremely possible not sure why he wouldn't have just looked up ONE benchmark of any current game to see that it is.

Edit: A couple more....

Dat 1 FPS..... I seriously wonder where some people get their information....

 

[AtenRa]

Broadwell-K's IGP should be faster than anything AMD has at the time of it's release. Unless it gets delayed again. It will of course be in a different price range.

Broadwell-K will only have GT2 with 24 EUs. It will not even going to reach Kaveris level of performance in desktop.

 

[ame132]

Broadwell-K will only have GT2 with 24 EUs. It will not even going to reach Kaveris level of performance in desktop.

i dont think so
http://www.anandtech.com/show/7875/new-unlocked-iris-pro-cpu-broadwell

 

[tential]

i dont think so
http://www.anandtech.com/show/7875/new-unlocked-iris-pro-cpu-broadwell

I doubt it will be the normal i7-k SKU but a specialty SKU costing an arm, leg, and possibly a kidney.

If Iris Pro is coming to all K Processors at the normal Iris Pro mark up... may god have mercy on our souls....

 

[ame132]

I doubt it will be the normal i7-k SKU but a specialty SKU costing an arm, leg, and possibly a kidney.

If Iris Pro is coming to all K Processors at the normal Iris Pro mark up... may god have mercy on our souls....

again, i dont think so
cpu with iris costs 358 http://ark.intel.com/products/76642/Intel-Core-i7-4770R-Processor-6M-Cache-up-to-3_90-GHz
cpu w/o iris costs 350 http://ark.intel.com/products/75123/Intel-Core-i7-4770K-Processor-8M-Cache-up-to-3_90-GHz

 

[III-V]

again, i dont think so
cpu with iris costs 358 http://ark.intel.com/products/76642/Intel-Core-i7-4770R-Processor-6M-Cache-up-to-3_90-GHz
cpu w/o iris costs 350 http://ark.intel.com/products/75123/Intel-Core-i7-4770K-Processor-8M-Cache-up-to-3_90-GHz

Yup.

Broadwell-K will only have GT2 with 24 EUs. It will not even going to reach Kaveris level of performance in desktop.

I think you haven't had your coffee today

Broadwell-K is GT4.

 

[witeken]

Yup.

I think you haven't had your coffee today

Broadwell-K is GT4.

Huh, since when does Broadwell GT4 exist and since when will it be used instead of GT3? Broadwell-K should have 1TFLOPS of HSA Gen8 performance.

 

[Enigmoid]

Huh, since when does Broadwell GT4 exist and since when will it be used instead of GT3? Broadwell-K should have 1TFLOPS of HSA Gen8 performance.

But intel's perf/FLOP is quite low.

 

[witeken]

But intel's perf/FLOP is quite low.

You know more than we do, when will you post the benchmarks of your Broadwell-K Gen8 IGP?

 

[sm625]

Future AMD APUs probably will have the texture compression technology we saw in Tonga too.

Careful. This is exactly the kind of thing that intel could easily jump on and do better than AMD. Honestly I'm really surprised they havent jumped to the forefront in fixed function texture compression. It makes even more sense than quicksync. And, as with quicksync, intel doesnt really care if some people think the resulting quality isnt that great. As long as its fast, right?

 

[III-V]

Huh, since when does Broadwell GT4 exist and since when will it be used instead of GT3? Broadwell-K should have 1TFLOPS of HSA Gen8 performance.

GT3e. Sounds like I need my coffee too.

 

[ultimatebob]

I think the real question is "Why would Intel invest billions of dollars in R&D to become competitive in a declining market like desktop gaming PC's?"

Laptops and tablets are where the money is now, and most of them have integrated graphics. Intel is making big improvements in this area. Hell... the Iris Pro GPU component is actually pretty damn good.

 

[AtenRa]

hmmm, Broadwell-K 4+3e ??? close to 150-160mm2 + eDRAM, that is a huge die for 14nm process.

Carrizo with HDL on 28nm could be close to 180mm2, that will be a huge battle between the two. And both should be release close to each other. Cant wait to test them both

 

[ShintaiDK]

hmmm, Broadwell-K 4+3e ??? close to 150-160mm2 + eDRAM, that is a huge die for 14nm process.

Carrizo with HDL on 28nm could be close to 180mm2, that will be a huge battle between the two. And both should be release close to each other. Cant wait to test them both

But its not. Carrizo is the same size as Kaveri.

http://www.eetimes.com/document.asp?_mc=RSS_EET_EDT&doc_id=1324643&page_number=2

AMD will disclose Carrizo, an integrated processor with its latest x86 core. The 28 nm chip measures 244.62 mm2 and packs more than 3.1 billion transistors.

 

[AtenRa]

But its not. Carrizo is the same size as Kaveri.

http://www.eetimes.com/document.asp?_mc=RSS_EET_EDT&doc_id=1324643&page_number=2

You are right, i was thinking of Kaveri with HDL

 

[witeken]

But its not. Carrizo is the same size as Kaveri.

http://www.eetimes.com/document.asp?_mc=RSS_EET_EDT&doc_id=1324643&page_number=2

I notice TSMC's 1.9x density scaling number is spot on.

 

[Enigmoid]

You know more than we do, when will you post the benchmarks of your Broadwell-K Gen8 IGP?

Not but Haswell GT3e was pretty poor. I'm not expecting suddenly intel to gain parity with nvidia with broadwell.

HD5200e = 830 GFLOPS @ 1300mhz
650m = 691 GFLOPS @ 900 mhz

650m is generally 30% faster.

 

[ame132]

Not but Haswell GT3e was pretty poor. I'm not expecting suddenly intel to gain parity with nvidia with broadwell.

HD5200e = 830 GFLOPS @ 1300mhz
650m = 691 GFLOPS @ 900 mhz

650m is generally 30% faster.

probably mostly due to texturing
http://www.anandtech.com/show/6993/intel-iris-pro-5200-graphics-review-core-i74950hq-tested/15

 

[SAAA]

Not but Haswell GT3e was pretty poor. I'm not expecting suddenly intel to gain parity with nvidia with broadwell.

HD5200e = 830 GFLOPS @ 1300mhz
650m = 691 GFLOPS @ 900 mhz

650m is generally 30% faster.

The most dreadful thing about Intel IGP is performance/mm^2 right now. Looking at the competitors it seems they are 5-10 years behind in this and they even have a decent process to hide the problem!
Imagine the same architecture on 28 or 32nm process...

Overall performance isn't bad for an IGP, especaially the GT3e variant, but hopefully Broadwell brings some damned improvements in texturing and real world performance at the same clocks.
Is there any review of core-M who focuses on this? Did they get better or not with gen8?

 

[ShintaiDK]

Sampler performance gets a huge increase in Gen8. So it will be interesting to see how it turns out after that bottleneck is reduced.

 

[witeken]

Performance per mm² isn't stellar because Intel didn't focus on density until 14nm.

 

[AtenRa]

Performance per mm² isn't stellar because Intel didn't focus on density until 14nm.

We are talking about the design of the iGPU not the density of the process.

 

[Enigmoid]

The most dreadful thing about Intel IGP is performance/mm^2 right now. Looking at the competitors it seems they are 5-10 years behind in this and they even have a decent process to hide the problem!
Imagine the same architecture on 28 or 32nm process...

Overall performance isn't bad for an IGP, especaially the GT3e variant, but hopefully Broadwell brings some damned improvements in texturing and real world performance at the same clocks.
Is there any review of core-M who focuses on this? Did they get better or not with gen8?

Haswell GT2 - 177 mm^2 * 0.31 = 55mm^2
Kaveri - 245 mm^2 * 0.47 = 115 mm^2

Kaveri is not quite twice as fast (iris pro is generally a tiny bit faster and has less than 100% scaling over GT2). Density must be factored in as well as drivers (intel's hardware is capable of more but is held back by their drivers). The video block, quicksync and other non-performance areas of the GPU form a larger percentage of intel's igp as well (the actual EU space is roughly half of the igp space) compared to AMD's much larger in proportion performance area of the igp.

 

[AtenRa]

Haswell GT2 - 177 mm^2 * 0.31 = 55mm^2
Kaveri - 245 mm^2 * 0.47 = 115 mm^2

Kaveri is not quite twice as fast (iris pro is generally a tiny bit faster and has less than 100% scaling over GT2). Density must be factored in as well as drivers (intel's hardware is capable of more but is held back by their drivers). The video block, quicksync and other non-performance areas of the GPU form a larger percentage of intel's igp as well (the actual EU space is roughly half of the igp space) compared to AMD's much larger in proportion performance area of the igp.

Iris Pro(HD5200) iGPU is close to 100-110mm2, add 85mm2 for the eDRAM and we are close to 200mm2 vs 110-120mm2 for Kaveri.

Intel Iris (HD5000) without eDRAM is considerable slower than Kaveri at the same die size with a full node advantage. Kaveris iGPU at 22nm would be smaller than Intels HD5000 and perform faster.