can Intel's Iris PRO inside Broadwell be a low-end discrete video card killer?

[Opcode]

Different testing protocols, Anandtech used the built-in benchmark. I haven't checked if it's the case with GTA V, but built-in benchmarks often give completely different results than the game itself.

In game benchmarks use scripted gameplay. So either way there is something majorly fishy about Toms Hardware results. I have never condoned anyone using Toms Hardware as they simply aren't credible and the entire OCN community will stand behind that. Anandtech has been a far more credible source and has results that at least look legit. You're not going to squeeze 100% more performance out of eight extra execution units (1 slice) nor any architecture revision. The testing methodologies could be different although Toms results are severely skewed somehow (likely a scene with very little detail raising the average). Anandtech came to the conclusion of ~20% which matches Intel's own projections. I feel sorry for the sucker going to buy a Broadwell chip thinking they're going to get 100+ FPS effortlessly in GTA V.

[Sakkura]

In game benchmarks use scripted gameplay. So either way there is something majorly fishy about Toms Hardware results. I have never condoned anyone using Toms Hardware as they simply aren't credible and the entire OCN community will stand behind that. Anandtech has been a far more credible source and has results that at least look legit. You're not going to squeeze 100% more performance out of eight extra execution units (1 slice) nor any architecture revision. The testing methodologies could be different although Toms results are severely skewed somehow (likely a scene with very little detail raising the average). Anandtech came to the conclusion of ~20% which matches Intel's own projections. I feel sorry for the sucker going to buy a Broadwell chip thinking they're going to get 100+ FPS effortlessly in GTA V.

 

100% more performance than what? If you're comparing with HD 4600, remember that it's 48 EUs vs. 20, and with a serious extra boost from the eDRAM. Getting more than twice the performance is exactly what you'd expect there.

[Opcode]

100% more performance than what? If you're comparing with HD 4600, remember that it's 48 EUs vs. 20, and with a serious extra boost from the eDRAM. Getting more than twice the performance is exactly what you'd expect there.

You're not going to get a 100% performance uplift going from 40 Gen7.5 to 48 Gen8 execution units. I'm surprised I'm the only one who caught on to Toms horrendous numbers. I'm not saying they didn't achieve these numbers in game although everyone knows if you're standing there staring straight up to the sky outside of the city limits that doesn't account for real world performance. Their numbers are misleading to consumers who think based on five conductive tests done that the i5-5675C averaged at 122 FPS. As to where Anandtech shows what would be real world performance. Keep in mind Toms and Anandtech benchmarks were done at the same resolution and settings. There's no way Iris Pro 6200 doubles A10-7850k performance (Anandtech backs that up as well).

[Sakkura]

You're not going to get a 100% performance uplift going from 40 Gen7.5 to 48 Gen8 execution units.

 

But that's not what they're showing. They're showing a >100% increase going from 20 EUs to 48 EUs + eDRAM. Which is entirely reasonable.

 

Their comparisons with the AMD APUs may not be entirely fair, depending on your point of view. It looks like Tom's Hardware picked generally less graphically demanding games/settings than Anandtech, which means framerates will scale higher, which means the CPU cores are more likely to bottleneck performance. Which will definitely be more of a problem for an AMD APU than for Intel CPUs.

[ivan134]

I don't think some people on this forum know what the phrase "budget gamer" means, if they're seriously saying these Broadwell chips are better alternatives for them.

[Megahurt]

 

I know it's possible, but who knows if it will be widely supported.

[Opcode]

But that's not what they're showing. They're showing a >100% increase going from 20 EUs to 48 EUs + eDRAM. Which is entirely reasonable.

 

Their comparisons with the AMD APUs may not be entirely fair, depending on your point of view. It looks like Tom's Hardware picked generally less graphically demanding games/settings than Anandtech, which means framerates will scale higher, which means the CPU cores are more likely to bottleneck performance. Which will definitely be more of a problem for an AMD APU than for Intel CPUs.

That's exactly what they're showing. Toms Hardware has a horrendous reputation in the big dog tech communities and this is why. Regardless of testing methodology there is no way Iris Pro 6200 is going to ever twofold Godavari performance even in scenes that are more IPC dependent. There's something extremely wrong with their numbers when you know a credible source puts only 7% between the two meanwhile they put 87% between them. Bad numbers are bad numbers any which way you look at them leaving Toms Hardware to live up to their reputation, yet again.

[mathijs727]

Toms Hardware has earned their reputation from examples like this. There's no way IP 6200 got a 100% performance uplift with just eight extra execution units. Toms Hardware is an unreliable source and we (on other bigger tech forums) always tell people to not bother sourcing from them for this very reason. Although the Anandtech numbers look within reason.

Their numbers are clearly fabricated if you look through and compare them enough.

GTA V w/ i5-5675C

• Toms Hardware = 122 FPS
• Anandtech = 56 FPS

The difference between i5-5675C vs A8-7650k is only 11% (55.25 vs 49.76 FPS) in Anadtech's review but in Tom's review it's 103% (122 vs 60 FPS).Reading is hard apperently.

Anandtechs chart does NOT contain FPS numbers but the amount of frames that took less then 16ms, they're not comparable.

[Victorious Secret]

I don't get why people keep saying this. Look at all the gaming benchmarks in the reviews. Even if Intel's drivers are bad (have not seen anything to support this) the 5675C is still beating the crap out of the A10-7870K. Are you saying Intel's lead would go from big to massive if they upgraded their drivers?

I don't get that argument either. No one had experience until they do something. Intel is doing something. Intel has agreements with both sides for their GPU tech.

Intel didn't just pull this out of their asses with no idea on how to sustain it. Iris Pro 5200 was the beta test. 6200 is gen 1. It works. It only gets better.

Intel knows that APUs will be a future, theirs just happen to be...better? It's all fun to say Zen will crush Broadwell. It better, since we will be on 7xxx Iris Pro by this fall with Skylake. We have only current products to compare and currently, Intel is winning. Can't argue that.

[Opcode]

Reading is hard apperently.

Anandtechs chart does NOT contain FPS numbers but the amount of frames that took less then 16ms, they're not comparable.

Look again, that's the average FPS chart taken from their review.

[Sakkura]

That's exactly what they're showing. Toms Hardware has a horrendous reputation in the big dog tech communities and this is why. Regardless of testing methodology there is no way Iris Pro 6200 is going to ever twofold Godavari performance even in scenes that are more IPC dependent. There's something extremely wrong with their numbers when you know a credible source puts only 7% between the two meanwhile they put 87% between them. Bad numbers are bad numbers any which way you look at them leaving Toms Hardware to live up to their reputation, yet again.

No, they can't show that. They did not include any CPU with Iris Pro 5200 to compare with the new Iris Pro 6200, they only compared against the inferior HD Graphics 4600 that the Iris Pro 6200 will obviously crush with ease.

 

They also didn't compare with Godavari, just Kaveri.

[Opcode]

No, they can't show that. They did not include any CPU with Iris Pro 5200 to compare with the new Iris Pro 6200, they only compared against the inferior HD Graphics 4600 that the Iris Pro 6200 will obviously crush with ease.

 

They also didn't compare with Godavari, just Kaveri.

They can with Iris Pro 5200 benchmarking just below Kaveri in a majority of established benchmarks (lots of them). All of their numbers (not just GTA V) are putting the Iris Pro 6200 >= 40-87% over Kaveri. Leaving their testing methodology extremely flawed. Especially when their numbers don't even remotely match previous benchmarks they've done on the same hardware. It's safe to say their benchmarks are far from inclusive.

[patrickjp93]

Can IGPUs be overclocked?

If the CPU is unlocked, so is the iGPU.

[patrickjp93]

ah ok, I didnt know this.

Look at the 4980HQ to see what Intel managed to get into some high end ultra books and laptops such as the MacBook Pro Retina 15"

[patrickjp93]

The menu in Shadow Warrior can't even run on intel integrated. It artefacts to hell. There is a reason why AMD and NVidia bring out monthly drivers and Intel does it like once a year or so.

I don't understand why anyone is hyped that a 350$ Intel CPU is beating a 150$ AMD APU. Buy a dedicated graphics card for the 200$ difference and see what's best?

That problem disappeared months ago. Also, Intel's updating drivers about every other month now.

[patrickjp93]

Toms Hardware has earned their reputation from examples like this. There's no way IP 6200 got a 100% performance uplift with just eight extra execution units. Toms Hardware is an unreliable source and we (on other bigger tech forums) always tell people to not bother sourcing from them for this very reason. Although the Anandtech numbers look within reason.

Their numbers are clearly fabricated if you look through and compare them enough.

GTA V w/ i5-5675C

• Toms Hardware = 122 FPS
• Anandtech = 56 FPS

The difference between i5-5675C vs A8-7650k is only 11% (55.25 vs 49.76 FPS) in Anadtech's review but in Tom's review it's 103% (122 vs 60 FPS).Intel also revised the GPU architecture from 7.5 to 8, changing a lot of internal busses and opening internal bandwidth that was a bit crunched on Iris Pro 5200. Also be aware there's a significant improvement when over locking Intel's Iris Pro graphics.

[dragosudeki]

Though now, wouldn't you lose the features that both AMD and Nvidia have (GSync/FreeSync, Hairwork/TressFX, etc.).  I find it hard to see them getting more and more into graphics. Probably will remain entry level until GSync/FreeSync (and others) becomes much more common.

[patrickjp93]

Though now, wouldn't you lose the features that both AMD and Nvidia have (GSync/FreeSync, Hairwork/TressFX, etc.). I find it hard to see them getting more and more into graphics. Probably will remain entry level until GSync/FreeSync (and others) becomes much more common.

I'm sure Intel will invest in something like FreeSync if its community demands it. Intel's more focused on preparing their architecture for the supercomputer war in 2016/2017, where the Cannonlake graphics can come to Skylake E5/E7 Xeons for customers who want it.

[esteszeke]

Probably not quite. The Minimum FPS is still pretty low at 720p and the jump to 1080p is quite large. Still it's very impressive for a iGPU to do this. 

Unfortunately a 1080p monitor has more than two times as many pixels to render than in a 720p display. Could see some big hits.

[harrynowl]

I wonder if this means Intel might invest some time into improving their DirectX driver as they don't exactly optimise for gaming like discrete manufacturers.

[patrickjp93]

I wonder if this means Intel might invest some time into improving their DirectX driver as they don't exactly optimise for gaming like discrete manufacturers.

Have you used their more recent drivers? While Intel is definitely focused on compute for the future, I'm sure it'd be happy to take some wind out of Nvidia's sails in the meantime as long as AMD isn't trampled underfoot.

[harrynowl]

Have you used their more recent drivers? While Intel is definitely focused on compute for the future, I'm sure it'd be happy to take some wind out of Nvidia's sails in the meantime as long as AMD isn't trampled underfoot.

I run the latest driver on my laptop however it is an Ivy Bridle ULV i3 so there isn't much to work with there

 

However to my understanding Intel have never really optimised their driver for games beyond it not running like garbage on titles it can run.

[Sakkura]

They can with Iris Pro 5200 benchmarking just below Kaveri in a majority of established benchmarks (lots of them). All of their numbers (not just GTA V) are putting the Iris Pro 6200 >= 40-87% over Kaveri. Leaving their testing methodology extremely flawed. Especially when their numbers don't even remotely match previous benchmarks they've done on the same hardware. It's safe to say their benchmarks are far from inclusive.

 

There were games where Iris Pro 5200 outperformed the A10-7850K, and adding 20% more EUs and fiddling with the architecture would just add to that.

 

 

Those earlier reviews also showed a lot of variability, which seems to be what we see repeated today. Since both Tom's and Anandtech have sampled a limited number of games each, the fact that they come up with very different pictures of performance doesn't have to indicate foul play.

[Opcode]

Intel also revised the GPU architecture from 7.5 to 8, changing a lot of internal busses and opening internal bandwidth that was a bit crunched on Iris Pro 5200. Also be aware there's a significant improvement when over locking Intel's Iris Pro graphics.

Which shows in Anandtech benchmarks although Toms Hardware results are just absurd, like always.

 

There were games where Iris Pro 5200 outperformed the A10-7850K, and adding 20% more EUs and fiddling with the architecture would just add to that.

 

 

Those earlier reviews also showed a lot of variability, which seems to be what we see repeated today. Since both Tom's and Anandtech have sampled a limited number of games each, the fact that they come up with very different pictures of performance doesn't have to indicate foul play.

There were a few games that did but not on a scale that high in every game.

[patrickjp93]

Which shows in Anandtech benchmarks although Toms Hardware results are just absurd, like always.

 

There were a few games that did but not on a scale that high in every game.

They're not absurd, and the quoted numbers of Tom's are several areas of gameplay whereas Anand only uses the bundled benchmark which is always worse than in-game results. Tom's doesn't even include the previous Iris Pro results, some of which are 20-30% above the 7850K anyway as Sakkura quoted, and actually yes, after Intel updated its drivers, it is on that scale.

 

EDIT: Did you happen to get my message/will you get back to me on it?