[gamegpu.ru] APU gaming including Skylake GT2 and Broadwell Iris Pro 6200

[Erenhardt]

There's the goalpost move.

So now it's "The best APU as long as you don't use it as a CPU". So ridiculous.

No as ridiculous as the best apu as long as you don't overflow the eDRAM while using IGP.

Business is slow, and Baby needs a new pair of shoes :biggrin:

And not as ridiculous as the reason why you are not yet banner for posts like these.

 

[Phynaz]

TOPIC title : [gamegpu.ru] APU gaming including Skylake GT2 and Broadwell Iris Pro 6200

I will not say anything else......be free to believe what you want.

Again:

the A8-7600 is the undisputed perf/$ champion of APUs

No, it's not.

It should be:

the A8-7600 is the undisputed perf/$ champion of APUs in the scenario I'm posting about, and only that scenario.

Anything else is dishonest.

 

[Dresdenboy]

AtenRa has a point that Intel iGPUs seem to fall off in performance when you up the quality settings, but aren't you going to be playing at fairly low settings anyway on an iGPU?

If you want excellent 3D performance in AAA, integrated graphics simply won't do.

Agreed. This is an easily predictable result of available power and RAM bandwidth (mitigated by the eDRAM). And the low settings simply reduce required iGPU shader performance (primitive setup and simple shading is not the bottleneck) to move that to the CPU cores, emphasizing processors with faster CPUs.

But I also recognized, that many posters here are blind on the price eye. Russian buyers with $200 salary surely are not.

When I was in Moscow for six weeks in 1999, Athlon already was the performance king. And nobody noticed that.

 

[Arachnotronic]

But I also recognized, that many posters here are blind on the price eye. Russian buyers with $200 salary surely are not.

Good point!

 

[Sweepr]

But I also recognized, that many posters here are blind on the price eye. Russian buyers with $200 salary surely are not.

Russians also know that Core i5-5675C/i7-5775C provides state of the art CPU performance for dGPU gaming, sometimes beating Skylake. Having the top notch iGPU is just a bonus.

 

[Dresdenboy]

Anything else is dishonest.

Are you sure? Full analysis and reasoning please. He was talking about perf/$. When I checked some the other sites given in this thread, I simply didn't see a significant amount of gaming tests even at Intel favouring low quality, where the 4770K or the Iris 6200 Pro CPUs were 2x to 3x faster.

 

[Dresdenboy]

Russians also know that Core i5-5675C/i7-5775C provides state of the art CPU performance for dGPU gaming, sometimes beating Skylake. Having the top notch iGPU is just a bonus.

Of course they know. The same way I know, that our car maker groups' Bugatti Veyron is one of the fastest and strongest cars out there, but I won't fall for this psychological marketing AIDA or comparison stuff ..

Wanna bet, that a i7-5775C rig would never be seriously beaten in gaming performance by another system, which costs the same.

 

[Sweepr]

Of course they know. The same way I know, that our car maker groups' Bugatti Veyron is one of the fastest and strongest cars out there, but I won't fall for this psychological marketing AIDA or comparison stuff ..

Wanna bet, that a i7-5775C rig would never be seriously beaten in gaming performance by another system, which costs the same.

The day AMD provides this level of CPU+iGPU performance at bargain APU prices, then you can call Broadwell-K expensive.
Vishera's green bars look great at the bottom of the performance charts BTW, I wonder where Kaveri/Godavari would fit.

 

[MrTeal]

Of course they know. The same way I know, that our car maker groups' Bugatti Veyron is one of the fastest and strongest cars out there, but I won't fall for this psychological marketing AIDA or comparison stuff ..

Wanna bet, that a i7-5775C rig would never be seriously beaten in gaming performance by another system, which costs the same.

What do you mean by this, exactly? It's not hard at all to imagine an i7-5775C using the iGPU getting beaten by a similar priced system. You can buy an i7-4790k and an R7 370 for the same price as an i7-5775C, and that would run rings around the Iris Pro solution even if you overclock them to the same speed.

 

[frozentundra123456]

Are you sure? Full analysis and reasoning please. He was talking about perf/$. When I checked some the other sites given in this thread, I simply didn't see a significant amount of gaming tests even at Intel favouring low quality, where the 4770K or the Iris 6200 Pro CPUs were 2x to 3x faster.

Lots of quibbling over semantics, but he was referring to the statement by Aten that the A8-7600 is the undisputed performance king at its price point.

That statement to be correct should have specified igpu performance.

Edit: and like I say every time in these threads, it is kind of ironic that APU fans talk about performance per dollar when 50 to 100% better performance is available with a cheap CPU like the x4 860k and a discrete card for at most 10% more cost based on the cost of an entire system.

 

[TheELF]

Are you sure? Full analysis and reasoning please. He was talking about perf/$.

Well look at what gamegpu tests, they have a video on each test that shows the testing that has been done,
war thunder=scripted benchmark,sure does not seem like someone actually playing*
Dirt Rally=scripted benchmark*
World of tanks=the player is always alone with max one other tank visible,not really the way that someone would actually play the game is it now,it's a single thread game and the more that is going on the more the APU will slow down.


*when you run scripted the software knows beforehand what will happen and can break up rendering into many threads thus making multicored cpus look much better then if you would actually play the game ,just look at the benchmarks atten does on his own,they are all based on this concept,never actual gameplay but only build in benchmarks that are just 3d render benchmarks for the gpu.

So yes, best perf/$ if you are only going to be "playing" in-build benchmarks...or you'r only looking at floors, one tree, one tank or otherwise take care not to tax the CPU.

 

[TheELF]

What do you mean by this, exactly? It's not hard at all to imagine an i7-5775C using the iGPU getting beaten by a similar priced system. You can buy an i7-4790k and an R7 370 for the same price as an i7-5775C, and that would run rings around the Iris Pro solution even if you overclock them to the same speed.

Yes but it wouldn't be a APU then...

(super reasoning the OP uses)

 

[MrTeal]

Yes but it wouldn't be a APU then...

(super reasoning the OP uses)

Yeah, stating the i7-5775C is the fastest APU available is a pretty reasonable statement, but the other part about pitting it against another similar priced rig seems off since the i7 is more than 3x more expensive than AMD's top APU. He was quoting a post about dGPU gaming.

 

[bystander36]

Edit: and like I say every time in these threads, it is kind of ironic that APU fans talk about performance per dollar when 50 to 100% better performance is available with a cheap CPU like the x4 860k and a discrete card for at most 10% more cost based on the cost of an entire system.

All the APU talks would make the most sense on mobile devices where power is an issue. Of course these are all desktop parts we are looking at, which for the most part don't seem to have much place.

 

[Phynaz]

And not as ridiculous as the reason why you are not yet banner for posts like these.

Did you bother Googling what I posted? It's a famous line from an American movie.

Perhaps you should consider sticking to Polish speaking websites to eliminate language as a barrier to communication, just as I stick to English speaking websites.

Assuming that by "banner", you meant "banned" of course. If you actually meant "banner", I don't think my posts are particularly sticky worthy, as I'm just some random guy on the internet.

 

[MiddleOfTheRoad]

Yeah, stating the i7-5775C is the fastest APU available is a pretty reasonable statement, but the other part about pitting it against another similar priced rig seems off since the i7 is more than 3x more expensive than AMD's top APU. He was quoting a post about dGPU gaming.

Totally Agree.

Plus, from the benches I've seen the 5775C only edges out the A10 7850K by about 10% - $15% in most 3D games, the i7 is definitely not worth triple the retail cost of the A10 if you are just gaming on the integrated graphics.

I am basing this on the numbers in this review:
http://www.overclockers.com/intel-i7-5775c-broadwell-cpu-review/

 

[MiddleOfTheRoad]

All the APU talks would make the most sense on mobile devices where power is an issue. Of course these are all desktop parts we are looking at, which for the most part don't seem to have much place.

Well, it really depends on the build. For an ITX build, I'd rather run an APU especially in the tight/small cases. For a mid-tower or better, give me the dedicated video card.

 

[Dresdenboy]

The day AMD provides this level of CPU+iGPU performance at bargain APU prices, then you can call Broadwell-K expensive.
Vishera's green bars look great at the bottom of the performance charts BTW, I wonder where Kaveri/Godavari would fit.

Vishera is kind of outdated. Since even 4C/4T CPUs do much better in most tests, it looks like a typical DX11-single-core-performance bottleneck to me. In this regard the newer AMD CPUs might improve somewhat, especially XV (in BR) with larger L1 D$ and other improvements.

What do you mean by this, exactly? It's not hard at all to imagine an i7-5775C using the iGPU getting beaten by a similar priced system. You can buy an i7-4790k and an R7 370 for the same price as an i7-5775C, and that would run rings around the Iris Pro solution even if you overclock them to the same speed.

Of course. That is related to the discussion about hardware choices of budget limited Russians. They might leave the path of iGPU glory and choose a dGPU instead. Power is cheap there I think.

Lots of quibbling over semantics, but he was referring to the statement by Aten that the A8-7600 is the undisputed performance king at its price point.

That statement to be correct should have specified igpu performance.

Edit: and like I say every time in these threads, it is kind of ironic that APU fans talk about performance per dollar when 50 to 100% better performance is available with a cheap CPU like the x4 860k and a discrete card for at most 10% more cost based on the cost of an entire system.

Of course. Enthusiasts and DIY builders might choose an iGPU based system due to power efficiency. But still the bigger market is that of off the shelf systems, where labor cost, space, etc. add up when using cheap dGPUs.

I think, most CPU/APU products out there have their individual rights of existence as they fulfill very different needs. It's just useless to discuss any product based on needs it isn't meant for.

 

[TheELF]

Of course. Enthusiasts and DIY builders might choose an iGPU based system due to power efficiency.

iGPUs are not power efficient they are low performance and thus low consumption,and throttling the CPU when the iGPU is loaded or throttling the iGPU when the CPU is loaded is not power efficiency.

 

[Dresdenboy]

Well look at what gamegpu tests, they have a video on each test that shows the testing that has been done,
war thunder=scripted benchmark,sure does not seem like someone actually playing*
Dirt Rally=scripted benchmark*
World of tanks=the player is always alone with max one other tank visible,not really the way that someone would actually play the game is it now,it's a single thread game and the more that is going on the more the APU will slow down.


*when you run scripted the software knows beforehand what will happen and can break up rendering into many threads thus making multicored cpus look much better then if you would actually play the game ,just look at the benchmarks atten does on his own,they are all based on this concept,never actual gameplay but only build in benchmarks that are just 3d render benchmarks for the gpu.

So yes, best perf/$ if you are only going to be "playing" in-build benchmarks...or you'r only looking at floors, one tree, one tank or otherwise take care not to tax the CPU.

Of course scripted benchmarks have their drawbacks. But this also depends on the way it is implemented. But where have you seen cases, where software tricks itself by using advance knowledge? And doesn't still the DX11 multithreading bottleneck limit such tries?

BTW I didn't find a benchmark mode in WoT. It looks like testers still have to do that by running a replay. But even then the engine just gets commands like in a real game.
And there is a reason that Russians like to benchmark WoT and WT.

But even recording the FPS at actual MP gameplay poses multiple problems:

• It causes more work to be prepared and analyzed (just multiply the ## of CPUs and games tested).
• With deep buffering the FPS would look fine, while the player experiences some serious lag.
• MP scenarios vary.

After decades of running benchmarks (Norton SI and Landmark anyone?) I think we need some more sophisticated testing methods to give a true analysis of perceived performance. FPS statistics with 99% percentiles etc. are just the beginning.

 

[Dresdenboy]

iGPUs are not power efficient they are low performance and thus low consumption,and throttling the CPU when the iGPU is loaded or throttling the iGPU when the CPU is loaded is not power efficiency.

Throttling is power efficient per se, as the highest clock frequencies are the least efficient ones and even might cause idle cores to leak more (voltage islands).

So do you have a chart showing FPS over power?

 

[TheELF]

Of course scripted benchmarks have their drawbacks. But this also depends on the way it is implemented. But where have you seen cases, where software tricks itself by using advance knowledge? And doesn't still the DX11 multithreading bottleneck limit such tries?

Render Frame Ahead
max pre-rendered frames
(Ever heard of that in all your years of benchmarking? )
If you know every frame that is going to appear, and because it is scripted you can, you can render ahead and get way better FPS.
And there are plenty of ways to get multiple threads to render frames faster under Dx11 just look at what Crysis3/cryengine does, if you don't need the cpu for game code it maxes out threads to produce more frames.
https://www.youtube.com/watch?v=lEafN99Q4K4

BTW I didn't find a benchmark mode in WoT. It looks like testers still have to do that by running a replay. But even then the engine just gets commands like in a real game.

Look at the video on gamegpu.

World of tanks=the player is always alone with max one other tank visible,not really the way that someone would actually play the game is it now,it's a single thread game and the more that is going on the more the APU will slow down.

Throttling is power efficient per se, as the highest clock frequencies are the least efficient ones and even might cause idle cores to leak more (voltage islands).

Ok but it does make for confusing/misdirecting benchmarks, you can't play a game on a APU without pushing both CPU+iGPU,looking at benches that push only the iGPU gives you false results.

 

[MiddleOfTheRoad]

iGPUs are not power efficient they are low performance and thus low consumption,and throttling the CPU when the iGPU is loaded or throttling the iGPU when the CPU is loaded is not power efficiency.

"Low" performance is subjective. What you call "low performance" is probably fine with a large number of PC gamers. Look at the steam survey results: Intel HD Graphics 4000 was #2 last month for most common graphics.

http://store.steampowered.com/hwsurvey/videocard/

It is also very energy efficient not to run a video card -- many gaming video cards triple the power draw of a desktop computer.

 

[MiddleOfTheRoad]

https://www.youtube.com/watch?v=lEafN99Q4K4
Look at the video on gamegpu.
World of tanks=the player is always alone with max one other tank visible,not really the way that someone would actually play the game is it now,it's a single thread game and the more that is going on the more the APU will slow down.

Why? The Kaveri APU's are comparable to Haswell Celerons / Pentiums in single threaded performance. So anyone advocating a G3258 instead of a Kaveri APU for single threaded performance is just spinning their wheels -- because they perform roughly the same.

 

[frozentundra123456]

All the APU talks would make the most sense on mobile devices where power is an issue. Of course these are all desktop parts we are looking at, which for the most part don't seem to have much place.

Agreed, but APUs in mobile, due of course to TDP constraints, are even more anemic than on the desktop.

I do wish however, that Intel would make GT3e, or whatever they are calling the Skylake equivalent, standard on the mobile i7 quads. Seems like it would make a decent mobile low end gaming platform. Of course they would have to make the price competitive vs say 950/960m discrete systems with mobile quads.