Mobile Trinity Graphics: Not really better than Intel HD Graphics 4000?

[Obsoleet]

The GPUs are within 20% of each other, but the Ivy Bridge CPU is twice as good as the A10 IMO. Look at the performance-per-watt. For a laptop it's a no brainer.

Lots of IMO but all the facts- 20% performance advantage, 38X less frame lag in some cases (in worst case for the A10, it's much, much closer). Sure, the IB CPU is favorable, but that still doesn't stop the A10 from trouncing it hard. We're comparing graphics performance, not CPU performance that not all of us care about. Not to mention all the raw numbers and anecdotal evidence from users of both IGP solutions that say that Trinity is a much better multitasker.

Needless to say, on performance per watt, you need the have the 'performance' in performance per watt.

1050/819 = 28% increase in GFLOPS for Kaveri over Trinity - both using 100W! If the bandwidth can also be increased by over 20% then we'll be seeing at least 25% increase in overall performance. Even with a truly unified memory arch, there still needs to be considerably more bandwidth due to the CPU also being clocked 15-20% higher (and probably 25% more bandwidth-hungry) as well.

ONLY 25%?!??! But, but but, the HD4000 is 'only' 20% behind now.
There's been months, if not year after year debates with discrete cards over less of a performance gap.

 

[bgt]

Get 2 laptops, Intel i3 & AMD Trinity based, with only IGP's, next to eachother and look/feel the difference. Trinity is always smooth. The HWsites look at benches and say.................oh, the Trinity's are slow but thats not reallife experience. They should make a bench like: have a room of 20 or more ordinary people using the pc/laptops and put their experiences in a graphical form and look then. I have both brands for years and I cannot say benches represent real life situations/use. 10-30% differences are useless in everyday life. You don't notice it. Or you are forever benchmarking systems.

 

[Obsoleet]

Get 2 laptops, Intel&AMD based, with only IGP's next to eachother and look/feel the difference. Trinity is always smooth. The HWsites look at benches and say.................oh, the Trinity's are slow but thats not reallife experience. They should make a bench like: have a room of 20 or more ordinary people using the pc/laptops and put their experiences in a graphical form and look then. I have both brands for years and I cannot say benches represent real life situations/use. 10-30% differences are useless in everyday life. You don't notice it. Or you are forever benchmarking systems.

I'm pretty much convinced from the frame latency results, the 20% raw advantage, and better multitasking benchmark results you presented. I also use VMs all the time, and recently found out that the entire AMD lineup has all the virtualization features, while Intel artificially limits them between i3/i5/i7. No thanks.

I'm going for A8 or A10 for a laptop when I next purchase, and A10 for desktop. Hope my existing stuff doesn't break before Kaveri is released though.

 

[bgt]

Just reading some user experiences for a HP Pavilion g6-2051sd(A10-4600), very positive

 

[Blitzvogel]

GCN probably has a superior cache system to the past VLIW4/5 too, hence why even with similar bandwidth to pre-GCN GPUs, it can still absolutely trounce them no question. It's because of this we can still expect Kaveri to still actually improve heavily on Richland and Trinity despite the moderately higher RAM support but still limited to dual-channel.

I would love to see AMD beat Intel to DDR4. Too bad it's highly unlikely to happen, but it would make so much sense for AMD's Fusion products.

 

[Obsoleet]

Once Haswell comes out it's game over for AMD. AMD will then have vastly inferior CPUs and moderately inferior GPUs inside their APUs. What a mess. Intel came out on top and they didn't even have to buy ATi. One could argue that nVidia's CPU division is now worth more than that of AMD.

This did NOT come true.
http://hothardware.com/Reviews/Intel-Core-i74770K-Review-Haswell-Has-Landed/?page=16
According to these tests, Haswell is 11% faster at best vs what is likely to be the comparable priced part (HD4600), when the current A10 at it's best is 27% faster. Small sample size, but it is suggestive. HD5200 doesn't really blow the socks off the A10 according to Anand's review, and that's the top dog.

Then there's the matter of HD3000/HD4000 having terrible frame latency results, which you'll have to worry about with Haswell due to its drivers.
http://techreport.com/review/22932/amd-a10-4600m-trinity-apu-reviewed/12

Oh well, better luck next gen. Looking forward to Kaveri more than ever now.

 

[Red Hawk]

Then there's the matter of HD3000/HD4000 having terrible frame latency results, which you'll have to worry about with Haswell due to its drivers.
http://techreport.com/review/22932/amd-a10-4600m-trinity-apu-reviewed/12

Out of the three games tested, AMD had lower frame latency in 2, and Intel had lower frame latency in one. If anything that indicates that frame latency is a problem for both platforms, and the sample size is much too small to establish one having an overall advantage against the other.

 

[Obsoleet]

Out of the three games tested, AMD had lower frame latency in 2, and Intel had lower frame latency in one. If anything that indicates that frame latency is a problem for both platforms, and the sample size is much too small to establish one having an overall advantage against the other.

That would be a misrepresentation of the data.
The 2 that AMD wins are blowouts by a factor of 4x to almost 40x worse.
The AMD loss is by a factor of 2x.

Time spent beyond 50ms
Intel won 711 vs 1521ms
AMD won 257 vs 2448ms
AMD won 27 vs 1034ms

It might not be an exhaustive test, but where I come from those are bad odds.
I'd avoid Intel graphics, even this so-called and previously hyped "Trinity killer", and go for AMD.