Iris pro benchmarks are in, and........they're VERY GOOD

[blackened23]

So "theoretical performance" aside, over three generations we've seen the performance double each time. It's rumored that GT4 will be up to 2 TFLOPs which would be another 2x gain. If Intel can deliver something that fits in a macbook pro form factor with that kind of power I would be very very happy. I don't understand what's not to like about that.

My line of thinking is the same. We're inching closer to the point to where 1080p quality gaming, even with the most demanding games, can be possible in a super small portable form factor. Currently, it isn't a realistic goal - small form factors such as ultrabooks mean trade-offs in terms of performance in exchange for long battery life. Meaning you can only play older games at 1080p and you'll need to dial down the quality settings. I see a day where this changes.

Folks can say what they want about mobile dGPU but the fact of the matter is, Iris Pro is overall matching the GT650M while having a lower overall TDP and better battery life. Additionally, BOM materials for Iris Pro will be cheaper since GT3e adds 57$ tray cost with EDRAM cache to a portable, while a GT750M costs about the same, but with some trade offs that will overall drive costs up: additional motherboard traces and complexity (ie more cost), GDDR5 VRAM (more cost), and a seperate cooling solution (higher cost). The only solutions in terms of dGPU that are truly cheaper are those that perform less. As well, notebook check seems to indicate that Iris Pro is in fact outperforming all AMD APUs in graphics. AMD just needs to release new parts - Iris Pro seems to be beating even the 7970m in a lot of tests. That is quite a feat.

With that being the case, this has NOTHING but good ramifications for us, the consumers - intel is bringing the heat here, and that means nvidia and AMD must step up. Competition is a great thing, folks. Can't wait to see what Broadwell GT4 does. This will force nvidia to lower prices and raise performance. It will force AMD to do the same.

What's not to like? I can't think of anything. This has only good implications for consumers.

 

[AtenRa]

AMD, otoh, is basically on the verge of losing all relevancy outside of desktop GPU. I'm not saying that to be mean, but just look at their current APUs. Even with the 6800K, you get pretty mediocre CPU performance, and the GPU performance isn't that hot either. Better than desktop Intel iGPU for sure, but the gap is getting thin, and will soon flip. Once you have lost :

Power efficiency
CPU performance per core
CPU performance overall
iGPU performance
iGPU perf/watt

Once all of that is lost, nothing to do but try to undercut on price, which is unsustainable in terms of bringing future competitive products to bear in order to attempt catching up.

Just have a look at the other topic to realize that Intel Mobile Graphics are still inferior to AMD at the same TDP levels even with a node Process advantage. A one year old 35W AMD Trinity is faster than 37W Haswell HD4600. 35W Ritchland is even faster due to better thermal management, higher clocks and 1866MHz memory support.

 

[blackened23]

Just have a look at the other topic to realize that Intel Mobile Graphics is still inferior to AMD at the same TDP levels even with a node Process advantage. A one year old 35W AMD Trinity is faster than 37W Haswell HD4600. 35W Ritchland is even faster due to better thermal management, higher clocks and 1866MHz memory support.

Well, in the LGA space AMD APUs may be better in terms of integrated graphics performance. That is definitely true - Mobile richland outperforms HD4600 in terms of graphics (but not CPU, obviously).

That said, it isn't exactly a secret that intel is chasing smaller form factors for their integrated graphics solutions, hence why only mobile SKUs have HD5000, 5100, and 5200 GT3. They are not available for LGA, they're only available for BGA form factors. With regards to the HD5000 SKUs, I don't think AMD has any answer that combines the same level of CPU and GPU performance within the same power consumption window. AMD mobile parts tend to have great idle power, but at full load things seem to go very awry. (ie high power use)

 

[AtenRa]

Well, in the LGA space AMD APUs may be better in terms of integrated graphics performance. That is definitely true - Mobile richland outperforms HD4600 in terms of graphics (but not CPU, obviously).

That said, it isn't exactly a secret that intel is chasing smaller form factors for their integrated graphics solutions, hence why only mobile SKUs have HD5000, 5100, and 5200 GT3. They are not available for LGA, they're only available for BGA form factors. With regards to the HD5000 SKUs, I don't think AMD has any answer that combines the same level of CPU and GPU performance within the same power consumption window. AMD mobile parts tend to have great idle power, but at full load things seem to go very awry. (ie high power use)

Thats to be expected, they are at 32nm currently vs 22nm for the Intel.

As for the HD5000/51000. I really will like to see the results, but i strongly believe those CPUs will be TDP limited and will not perform the way a lot of people believe.

 

[blackened23]

It's not intels' fault that AMD can't do 22nm. I don't see how that can possibly even enter the argument, intel paid their dues by paying billions upon billions of dollars in terms of R+D for their 22 and 14nm processes. If we rewind several years, you'll remember that AMD was ahead of intel in terms of x86 performance. What happened?

When AMD was riding high with the Athlon, their management completely wasted that opportunity by throwing their money away and making poor business choices. I really loved the AMD athlon, but their current situation is their fault and theirs alone - things COULD be much different had Hector Ruiz not run the company into the ground with extravagant spending on nonsense with little to no money spend on R+D - where it NEEDED to go.

AMD could easily be in a better position right now, but the fault of their current misfortunes are theirs and theirs alone. If their management had not been so excessively incompetent years ago, perhaps AMD *would* have 22 or 14nm. But they don't, and it's not Intel's fault. Anyway, back to your argument in question. I absolutely don't disagree that AMD APUs are better than HD4600 at the moment. I don't think they have proper answers for HD5000, 5100, and 5200, however - we'll see how Kaveri does next year, I suppose.

 

[AtenRa]

I just said it is logical that they have higher consumption at Load because of the 32nm process, there was no reason to go all that length about AMD, Hector etc.

 

[Haserath]

Does anyone expect Broadwell to have an updated 22nm chipset along with 256MB 22nm cache?

They should have quite a bit of capacity after they move the 'APU' to 14nm.
---
It also seems like Intel could improve efficiency if they didn't push such high clocks. HD 5000 fits in the MacBook Air and it does very well. That's fitting in an ultra book TDP with the chipset.

The CPU is also powerful for the GPU. Could use 2-3x the performance still.

 

[krumme]

Does anyone expect Broadwell to have an updated 22nm chipset along with 256MB 22nm cache?

They should have quite a bit of capacity after they move the 'APU' to 14nm.
---
It also seems like Intel could improve efficiency if they didn't push such high clocks. HD 5000 fits in the MacBook Air and it does very well. That's fitting in an ultra book TDP with the chipset.

The CPU is also powerful for the GPU. Could use 2-3x the performance still.

Acording to the orig. Anandtech article about the subject not much performance over 32Mb is shown. Intel made it 128mb to make it extra future proof. If thats the case then 128mb is what we get with broadwell also.

The edram is made on a custom 22nm proces optimized for low leakage instead of high freq. We will probably get the same 22nm process for broadwell also. Thats fine. It will keep cost down and gradually make it more mainstream.

We are very tdp constrained on the 5000 ultrabooks as shown by test at AT also. 14nm, efficiency gains, and just a few extra eu will bring us a great step forward on the ultra mobile scene.

 

[ShintaiDK]

Does anyone expect Broadwell to have an updated 22nm chipset along with 256MB 22nm cache?

They should have quite a bit of capacity after they move the 'APU' to 14nm.
---
It also seems like Intel could improve efficiency if they didn't push such high clocks. HD 5000 fits in the MacBook Air and it does very well. That's fitting in an ultra book TDP with the chipset.

The CPU is also powerful for the GPU. Could use 2-3x the performance still.

Broadwells chipset is the 9 series. 256MB cache? No.

There is no APU at Intel, only IGP.

Broadwell will expland the IGP to GT4 options, 2Tflop@1ghz.

 

[Fjodor2001]

Broadwell will expland the IGP to GT4 options, 2Tflop@1ghz.

Has the 2 Tflops and 1 GHz been mentioned officially somewhere, or is it some calculation/estimation based on what is known about Broadwell so far? Regardless, care to share the details?

 

[ShintaiDK]

Has the 2 Tflops and 1 GHz been mentioned officially somewhere, or is it some calculation/estimation based on what is known about Broadwell so far? Regardless, care to share the details?

 

[crashtech]

I see you switched the image host. Certain ones don't seem to be working here. Tinypic seems to be OK.

 

[ams23]

Isn't GT 650M about 1.5 years old now? Anyway, GT 750M should be well ahead of Iris Pro 5200 in most games. And if one measured actual GPU power consumption in these same games, I wouldn't be surprised to see GT 750M having higher perf. per watt than Iris Pro 5200 too. At the end of the day, most OEM's who want to differentiate their offerings from Intel integrated graphics will go with NVIDIA. And starting with the upcoming Kepler.M and heading towards Maxwell, the NVIDIA GPU perf. per watt will dramatically improve.

 

[ShintaiDK]

Isn't GT 650M about 1.5 years old now? Anyway, GT 750M should be well ahead of Iris Pro 5200 in most games. And if one measured actual GPU power consumption in these same games, I wouldn't be surprised to see GT 750M having higher perf. per watt than Iris Pro 5200 too. At the end of the day, most OEM's who want to differentiate their offerings from Intel integrated graphics will go with NVIDIA. And starting with the upcoming Kepler.M and heading towards Maxwell, the NVIDIA GPU perf. per watt will dramatically improve.

The 750M got alot lower perf/watt than Iris. The 750M got around the same TDP as the entire CPU.

And with IGP, you dont need to dedicate all that extra PCB and internal space. And your cooler can be smaller.

 

[krumme]

Isn't GT 650M about 1.5 years old now? Anyway, GT 750M should be well ahead of Iris Pro 5200 in most games. And if one measured actual GPU power consumption in these same games, I wouldn't be surprised to see GT 750M having higher perf. per watt than Iris Pro 5200 too. At the end of the day, most OEM's who want to differentiate their offerings from Intel integrated graphics will go with NVIDIA. And starting with the upcoming Kepler.M and heading towards Maxwell, the NVIDIA GPU perf. per watt will dramatically improve.

Yes 750 is the one on the market now, and the comparison should be to 750m. 750m is more or less the same as 650m. About plus 10% if i remember correctly, its a rebrand. Remember to get the gddr5 version

750m and the similar amd chips will do fine against 5200 as long as the prices is what they are for iris pro. The iris pro does not make a difference as it is. But its the start of the end of the midrange. Single chip is far more cost effective for the oem and ddr4 will take a toll on the gpu.

 

[blackened23]

Oh yeah that 57$ price for Iris Pro is killer. The quad mobile price stays static and Iris Pro is 57$ tray with EDRAM cache on top of the CPU price. That is as compared to mobile dGPUs which have the added cost of GDDR5 VRAM, motherboard tracers, an additional cooling solution, and higher overall system TDP. Let's say a GT650M costs 50 bucks. That doesn't include BOM cost of fan parts, GDDR5 VRAM, additional motherboard complexity, so on and so forth. So I were a vendor, why would I opt for a GT650M again, if it costs more and performs more or less equal to Iris Pro - while simultaneously the GT650M costs more and adds a higher TDP to the ultrabook.

The Iris Pro is not expensive as an added cost of a quad mobile i7 whose price obviously stays static. It is in fact cheaper than nearly every mobile discrete GPU except those which perform less. And there is no "similar" AMD chip whose performance matches the CPU performance of an intel quad mobile i7. Superior CPU performance aside, the Iris Pro also benchmarked at notebookcheck to be faster in most games than the 7970m chip. As far as I know that is the latest mobile dGPU from AMD and the Iris Pro is beating it in quite a few benchmarks and real world games.

 

[ams23]

The 750M got alot lower perf/watt than Iris.

I doubt that. To measure perf. per watt, one needs to measure gaming performance AND power consumption while gaming, for each respective gaming application. That type of data is not so easy to find, and cannot be derived from basic product specifications.

 

[SiliconWars]

Oh yeah that 57$ price for Iris Pro is killer. The quad mobile price stays static and Iris Pro is 57$ tray with EDRAM cache on top of the CPU price.

You also get a much slower CPU for the same price.

4700HQ - http://ark.intel.com/products/75116/
4750HQ - http://ark.intel.com/products/76087/

Considering what Intel charges for even 100 MHz, it's clear that the cost of the eDRAM (or Iris Pro is more accurate) is way higher than $57.

Superior CPU performance aside, the Iris Pro also benchmarked at notebookcheck to be faster in most games than the 7970m chip. As far as I know that is the latest mobile dGPU from AMD and the Iris Pro is beating it in quite a few benchmarks and real world games.

Uhh, not a chance in hell.

 

[Erenhardt]

Superior CPU performance aside, the Iris Pro also benchmarked at notebookcheck to be faster in most games than the 7970m chip. As far as I know that is the latest mobile dGPU from AMD and the Iris Pro is beating it in quite a few benchmarks and real world games.

Hahaha. That is funny as hell. I can understand someone might be enthusiastic about specific part, but to me it looks like you are the marketing guy @Intel that is resposible for sales of iris pro...

There was a linked review in which faster CPU+gtx750m consumes only 7W more in crysis3. That is with dedicated GDDR5 memory for GPU and whole banks of DDR3 dedicated only for system and applications.
Your 47 TDP chip is taking almost 2x the amount it can dissipate, you know what that means? After 5 minutes of acceptable gameplay CPU (and GPU, since it is integrated) will start to throttle - enjoy your slide show!
If you think about it: More TDP - better!
On top of that Iris pro performs around gtx650m. Slightly faster on low details, slower on high. There is a long way to 7970m from there.

 

[tential]

My line of thinking is the same. We're inching closer to the point to where 1080p quality gaming, even with the most demanding games, can be possible in a super small portable form factor. Currently, it isn't a realistic goal - small form factors such as ultrabooks mean trade-offs in terms of performance in exchange for long battery life. Meaning you can only play older games at 1080p and you'll need to dial down the quality settings. I see a day where this changes.

Folks can say what they want about mobile dGPU but the fact of the matter is, Iris Pro is overall matching the GT650M while having a lower overall TDP and better battery life. Additionally, BOM materials for Iris Pro will be cheaper since GT3e adds 57$ tray cost with EDRAM cache to a portable, while a GT750M costs about the same, but with some trade offs that will overall drive costs up: additional motherboard traces and complexity (ie more cost), GDDR5 VRAM (more cost), and a seperate cooling solution (higher cost). The only solutions in terms of dGPU that are truly cheaper are those that perform less. As well, notebook check seems to indicate that Iris Pro is in fact outperforming all AMD APUs in graphics. AMD just needs to release new parts - Iris Pro seems to be beating even the 7970m in a lot of tests. That is quite a feat.

With that being the case, this has NOTHING but good ramifications for us, the consumers - intel is bringing the heat here, and that means nvidia and AMD must step up. Competition is a great thing, folks. Can't wait to see what Broadwell GT4 does. This will force nvidia to lower prices and raise performance. It will force AMD to do the same.

What's not to like? I can't think of anything. This has only good implications for consumers.

Anandtech's review of mobility vs desktop just was released and it's funny how people on here are simply ignoring the fact that lets face it. Desktop is pretty much DONE. It's only a matter of time now. AMD isn't focusing on it. Intel isn't focusing on it. We are at a point where MOBILITY is important.

Intel is realizing how important battery life, and DECENT graphic performance is. They realize "Hey, we don't need top of the line performance, just good enough for people to game on w hile keeping good batterylife."

I'm going to see if I can hold out til the next iteration, but to those saying "Ultrabooks are a hunk of junk"(saw this a lot in this thread) just lol.....

CPU enthusiasts and gamers have typically been behind in seeing market trends and it's because most people who are CPU enthusiasts are stuck in their ways of benchmarking to get the highest score on an irrelevant program rather than seeing what the AVERAGE consumer wants.

The average consumer wants to be able to purchase a laptop, use it all day, MAYBE play some games, and do it at a relatively cheap price. Intel has been inching closer and closer to this and it's funny that INTEL made the ultrabook specification when in fact you'd have thought MS would have done it because in all reality MS needed it made more than Intel did.

In short though. Times are changing. Mobile is the future and Intel is understanding where its focus needs to be. I won't be surprised it intel dominates the whole CPU market within 5 years at this rate. Laptops and Desktops are basically theirs, they are taking a HUGE shot at even AMD and NVidia with this GPU move, and at the same time taking a shot at ARM.

 

[Fjodor2001]

@ShintaiDK:

So with Broadwell we'll basically get PS4 level of GPU performance, but with much better CPU performance, all in an Ultrabook CPU package? Nice!

But I guess the TDP will also be higher than current Iris Pro since there will be many more GPU cores... or will the 14 nm vs 22 nm be able to fully compensate for that?

 

[ShintaiDK]

@ShintaiDK:

So with Broadwell we'll basically get PS4 level of GPU performance, but with much better CPU performance, all in an Ultrabook CPU package? Nice!

But I guess the TDP will also be higher than current Iris Pro since there will be many more GPU cores... or will the 14 nm vs 22 nm be able to fully compensate for that?

TDP wont be higher. And yes, 14nm will allow it.

 

[Abwx]

The Iris Pro is not expensive as an added cost of a quad mobile i7 whose price obviously stays static.

A chip that consume as much as a A10 6700 65W DT CPU
is considered being a mobile dedicated chip??..

http://translate.google.de/translat...3/intel-iris-pro-5200-grafik-im-test/&act=url

 

[frozentundra123456]

@ShintaiDK:

So with Broadwell we'll basically get PS4 level of GPU performance, but with much better CPU performance, all in an Ultrabook CPU package? Nice!

But I guess the TDP will also be higher than current Iris Pro since there will be many more GPU cores... or will the 14 nm vs 22 nm be able to fully compensate for that?

That is a VERY optimistic prediction. In fact I would say it is totally impossible within the TDP constraints of a laptop or even more so an ultrabook. Current iris pro is maybe half the speed of a desktop 7750, and PS4 will be probably be twice that. So you would need 4x the performance in the same power envelope. IMO it could take several generations to see that, if ever.

It might be possible to build a huge monster desktop chip like that, but I just dont see how it could be made to fit in a mainstream ultrabook power/thermal envelope.

 

[Fjodor2001]

That is a VERY optimistic prediction. In fact I would say it is totally impossible within the TDP constraints of a laptop or even more so an ultrabook. Current iris pro is maybe half the speed of a desktop 7750, and PS4 will be probably be twice that. So you would need 4x the performance in the same power envelope. IMO it could take several generations to see that, if ever.

It might be possible to build a huge monster desktop chip like that, but I just dont see how it could be made to fit in a mainstream ultrabook power/thermal envelope.

I'm just summing up what ShintaiDK has claimed. The PS4 GPU is 1.8 TFlops, and he said the Broadwell Iris Pro IGPU will be 2 TFlops, all within the same or lower TDP than current Haswell based Iris Pro.