AMD Dual Graphics Analysis: Better Benchmarks; Same Experience?

[rmpumper]

Yes, a i3+7750 is faster than a A10+7750.

And cheaper (especially when you take the more expensive 2133 into account as well).

A10-6800K + 8GB 2133 RAM + 7750 > 150 + 90 + ~100 = $340
i3-3220 + 8GB 1600 RAM + 7750 > 130 + 70 + ~100 = $300

At this point, AMD's APUs are useless if you are going to use a discrete GPU with them. Hopefully, Kaveri will be a lot better.

 

[rmpumper]

Yes it is faster than the A10 + 7750 but not the A10 + 7750 in dual graphics mode.

Did you read the article? 7750 in dual mode is slower than single 7750.

 

[sarinaide]

Yes it is faster than the A10 + 7750 but not the A10 + 7750 in dual graphics mode.

Did you read the article? 7750 in dual mode is slower than single 7750.

1 - Game scales to Crossfire well so there is a difference between the 7750 v 7750/8670D
2 & 3 - Game doesn't utilise Dual Graphics scaling so performance is the same or slighly faster/slower.

Depends ultimately on the game, BF3 is another that scales well to dual graphics but is unplayable despite the good frame rates.

 

[Arief Widiatmoko]

Nice of testing system for analysing the smoothness of game, like the way tomshardware find this method, I think we really need the visual result of every game to know how smooth the game play and doesn't see only on the number only. it really simple to understand. Its great

 

[envy14tpe]

Nice of testing system for analysing the smoothness of game, like the way tomshardware find this method, I think we really need the visual result of every game to know how smooth the game play and doesn't see only on the number only. it really simple to understand. Its great

I too thought that was brilliant of tomshardware to show comparative videos. I had no idea the A10 was so choppy. It seemed smooth only when using the 7750...which defeats the point of the A10.

Once again, if you want a discrete GPU in your system, don't buy the A10 cuz the i3 is better and cheaper.

 

[ubercake]

So they don't even have crossfire fully fixed, but want to introduce this great new concept based around the already-flawed crossfire technology.

Doesn't make much sense, but there are a lot of people out there throwing their hard earned cash at something that doesn't work and vehemently defending their support of AMD with their flickery crossfire setup. Hey... If that's how you define value, more power to you. Don't get me wrong... The non-crossfire AMD setups are a deal, but AMD has historically not done a good job supporting their crossfire customers.

Even though the latest drivers by AMD are a step in the right direction, AMD has admitted they still have a ways to go with fixing crossfire's runt and dropped frame issue. Why a company would push more usage of a flawed technology is beyond me?

 

[yannigr]

I love Intel fanboys when they put a 7750 AND 2133 memory in the AMD setup to make it look more expensive than the Intel setup.

 

[ojas]

Um, Don...the article was informative, however...would it have made a difference had you tested at a lower resolution?

I'd not recommend an APU or the 6670 to anyone playing at above 720p, probably.

I mean, my GTX 560 struggles to exceed 40 fps at 1024x768 when the detail levels are turned up in some games.

Clarification: I mean this with respect to the stutter, if 1080p is too taxing a workload for these GPUs to produce a smooth sequence of frames, wouldn't it be better to give them a smaller resolution to render? Since that's more likely to be their intended use case.

 

[viewgamer]

For those that have already invested in a dual graphics solution, do not dispair, there is something you can do that will massively improve the experience.
Using the freeware RadeonPro, you can enable adaptive V-Sync, a frame limiter in essence.
You can set the frame limit to whatever you wish, setting it to your average frame-rate will completely remove micro-stutter.
http://www.tomshardware.com/reviews/radeon-hd-7990-devil13-7970-x2,3329-11.html

 

[ojas]

Being a APU owner and fully aware of past issues with Dual Graphics mode, notably the micro stutter, I was waiting for Catalyst 13.8 to re run the data and see if the fixes affected Dual Graphics in the manner it affected Crossfire. 13.8 has been out for 2-3 weeks and Toms still didn't incorporate this fix.

I do hope there will be a review on 13.8 in the near future and not some time down the line as this is subject matter that is particularly of interest to me and its hard to find reviews of this nature. That said I will probably have a 6670, 7730 and 7750 soon enough to run my own tests and to see whether the APU can crossfire with the GCN parts.

Like they've said in so many places, when they wrote the article 13.8 wasn't out, and AMD itself says that 13.8 doesn't affect this.

APUs are extremely effected by RAM. I'm not saying you guys are using bad RAM but I'm using RAM with better timings (average joe). I kinda expect you to be using the best., Tom. "We're using AMD's own RAM!" That isn't the best RAM.

And bandwidth will fix lack of frame pacing, how? Do you know, for example that even with all the bandwidth a 7990 or two 7970s have, they still suffer from similar issues? This is science and engineering, not magic.

A 6670 GDDR5/ 7750 + A10 6800k look like a promising heaven for those on a really tight budget, or those you just game on 1280x1024/1366x768

Are you serious? Faster RAM + A10 is better in gaming than A10 + 6670.
http://hexus.net/tech/reviews/cpu/46073-amd-a10-5800k-trinity-needs-faster-ram/?page=3
http://hexus.net/tech/reviews/cpu/46157-amd-a10-5800k-dual-graphics-evaluation/?page=2

Yes, a i3+7750 is faster than a A10+7750.

Yes it is faster than the A10 + 7750 but not the A10 + 7750 in dual graphics mode.

I await the further driver releases, aparently AMD are step releasing fixes and hopefully a dual graphics fix will come, in the interim i will try it out anyways.

Faster means jack if the frames are being dropped.

So they don't even have crossfire fully fixed, but want to introduce this great new concept based around the already-flawed crossfire technology.

Doesn't make much sense, but there are a lot of people out there throwing their hard earned cash at something that doesn't work and vehemently defending their support of AMD with their flickery crossfire setup. Hey... If that's how you define value, more power to you. Don't get me wrong... The non-crossfire AMD setups are a deal, but AMD has historically not done a good job supporting their crossfire customers.

Even though the latest drivers by AMD are a step in the right direction, AMD has admitted they still have a ways to go with fixing crossfire's runt and dropped frame issue. Why a company would push more usage of a flawed technology is beyond me?

This.

I love Intel fanboys when they put a 7750 AND 2133 memory in the AMD setup to make it look more expensive than the Intel setup.

LOL in my experience it's the AMD fanboys that always scream "WHY U NO USE FASTER RAM" every single time. Look at the second person i quoted, to see an example.

Wait, here's another one:

I wish the article included the A10-6800k + 2133 RAM + no discrete GPU. Seems like the A10 performs better with faster RAM rather than a better GPU.

How will that change anything? It's not about "MOAR FPS", in case you weren't reading.

 

[rmpumper]

I love Intel fanboys when they put a 7750 AND 2133 memory in the AMD setup to make it look more expensive than the Intel setup.

Whatever, troll. APU performance sucks with slower memory. That is a fact. Even if you put the same RAM with A10 and i3, Intel would still be 20 USD cheaper and faster than A10+7750. APU is only better if you will not be purchasing a GPU along with it.

 

[razvan_mz]

conclusion: get an Athlon X4 750k and a HD 7750 )

 

[InvalidError]

TL;DR: We worked with AMD prior to this piece's publication, and would have used 13.8, despite a lack of difference, if it had been available. If AMD told us 13.8 might change the outcome, we would have scrapped everything and started over.

People were probably expecting at least a line about doing a partial re-test to confirm what AMD said.

Anyone who's been in the tech business for a while should know that what the vendors say and what they actually deliver often disagree with each other in some ways since it is nearly impossible to ever be 100% certain software always behaves like expected... one little change in one place often has unexpected side-effects elsewhere. I would imagine AMD's frame-pacing fixes are made up of many tweaks all over the place.

 

[silverblue]

Considering the significant growth in APU sales over the past few months, I would expect AMD to put some serious work into their drivers to make these work in tandem with discrete cards, thus providing more value. I imagine Dual Graphics will never be more than a niche, but if you've bought a 6800 (for example) and just so happen to have a 6670/7750 or want a cheap boost to your framerates, you'd hope it actually worked properly. The measured frame rates are fine, but when performance is as jerky as when you're just using the IGP, you'd wonder why you even bothered.

 

[bwcbwc]

The drivers tested in this analysis are a tiny bit too old, just before the Crossfire Frame Pacing fix was released.
Could we see this performed again with Catalyst 13.8?

See page 2:

"You'll notice that we're using the Catalyst 13.6 Beta 2 driver instead of Catalyst 13.8 Beta, which adds a frame pacing feature for smoother, more consistent output. AMD tells us that the new driver does not affect Dual Graphics configurations. It only works with multiple discrete GPUs. Rest assured that the configuration we're presenting is as up to date as possible."

I guess the good news for AMD is that if they ever release a frame-paced driver for Dual Graphics with APUs, they should be able to achieve at least a 50% improvement over APU or low-end card alone. I don't expect to see the nearly perfect pacing that is attained with matching discrete GPUs, but if (for example) the 6670 processes 3 frames for each 2 frames handled by the APU, they should be able to get some pretty good results from a paced driver. The nasty part (or at least one nasty part) for them is to somehow provide tuning parameters for all the possible combinations of APU and card.

 

[DryCreamer]

I've been a long time stalker on Tom's and a member for about 2 years, and this is probably one of the most informative and surprising articles I've seen on here yet! Adding the video is an amazing feature!

Dry

 

[leagaleano]

You should have added some GPGPU benchmarks I'd like to check it for bitcoins 😛

 

[lpedraja2002]

Being a APU owner and fully aware of past issues with Dual Graphics mode, notably the micro stutter, I was waiting for Catalyst 13.8 to re run the data and see if the fixes affected Dual Graphics in the manner it affected Crossfire. 13.8 has been out for 2-3 weeks and Toms still didn't incorporate this fix.

I do hope there will be a review on 13.8 in the near future and not some time down the line as this is subject matter that is particularly of interest to me and its hard to find reviews of this nature. That said I will probably have a 6670, 7730 and 7750 soon enough to run my own tests and to see whether the APU can crossfire with the GCN parts.

Bet you didn't read past the first page.

Um, Don...the article was informative, however...would it have made a difference had you tested at a lower resolution?

I'd not recommend an APU or the 6670 to anyone playing at above 720p, probably.

I mean, my GTX 560 struggles to exceed 40 fps at 1024x768 when the detail levels are turned up in some games.

Clarification: I mean this with respect to the stutter, if 1080p is too taxing a workload for these GPUs to produce a smooth sequence of frames, wouldn't it be better to give them a smaller resolution to render? Since that's more likely to be their intended use case.

I agree, I also think the article was so PRO. The data gathered and those excellent videos make this site stand out even more now. But being these solutions are for the budget tight gamers I would also have liked to see test done in 720p resolutions and see if frame variance would be affected at all when the hybrid solution has an easier workload.

 

[JefferyD90]

Anyone notice that they declined to use the latest 13.8 driver? This is very typical of TomsHardware. If they would quit getting their pockets filled by Nvidia and Intel maybe then we would see some fair comparisons.

 

[ddpruitt]

It's interesting that many of the benchmarks were done using unplayable settings to begin with. I would like to see the benchmarks rerun with settings where even the lower end configurations run at playable frame rates. I'm also curious to see how RAM speed affects the overall result. We know APU's are sensitive to frame rates, I'm curious on this this affects dual graphics configurations.

 

[Sam Bittermann]

As an A10-6800K/6670 GDDR5/1866MHz DDR3 owner I'd like to thank Tom's for this article. I've experienced exactly what their results show and am severely disappointed with AMD at this time. Further more they won't give me a straight answer on whether a 7750 would work in dual graphics but it show it does AND gives the smoothest frame rates after a single card configuration. Seems I wasted my money on a 6670 and this whole system.

 

[Sam Bittermann]

As an A10-6800K/6670 GDDR5/1866MHz DDR3 owner I'd like to thank Tom's for this article. I've experienced exactly what their results show and am severely disappointed with AMD at this time. Further more they won't give me a straight answer on whether a 7750 would work in dual graphics but it show it does AND gives the smoothest frame rates after a single card configuration. Seems I wasted my money on a 6670 and this whole system.

 

[ddpruitt]

...

And bandwidth will fix lack of frame pacing, how? Do you know, for example that even with all the bandwidth a 7990 or two 7970s have, they still suffer from similar issues? This is science and engineering, not magic.

...

Faster means jack if the frames are being dropped.

...

LOL in my experience it's the AMD fanboys that always scream "WHY U NO USE FASTER RAM" every single time. Look at the second person i quoted, to see an example.

Wait, here's another one:
...

I think your missing the point. There are many benchmarks that indicate the APU's are very sensitive to memory bandwidth, double framerates in some cases. The reason a frame is dropped or becomes a runt is that the second GPU, in this case a more powerful discrete GPU, has a frame ready before the first one's done. If the APU is better able to keep up it may push out more complete frames. Since it wasn't tested we don't know if this is the case, however we can't dismiss it out of hand. Higher end graphics also suffer from runt frames but they are pushed much harder than this configuration. More importantly 7900s run equal GPUs and equal Memory. They aren't as unbalanced as APU + GPU (or DDR3 + GDDR5), we just want to try and even out the playing field a little.

Here's a case were we would like additional testing to disprove said hypothesis, like you said science not magic or guessing. Science and engineering involve testing, refining, and careful thought, clearly you don't understand the basic premise of either.

 

[Untruest]

"those additional investments are largely going to waste."... "Right now it's evangelizing a feature that helps improve benchmark results, but not actual gameplay. That's just not right"

The way I see it, a gamer was going to buy a CPU and going to buy a discrete graphics card. If you use that money to purchase a AMD APU (which cost less than Intel based CPUs) and purchase a AMD GPU, you have the option of using DUAL GRAPHICS. Meaning you can do no worse than your original plan anyway. How can this be a waste of investment?

Second Dual graphics is not perfect at the moment, yes AMD are the pioneers of this technology so there are lots of challenges but if I were to look on the bright side, I'd say at least it's only software so with an update hopefully this can be resolved. And to all those gamers on a budget playing at 20fps, the flawed extra boost in frame rate makes the game playable now at 30+fps at no noticeable glitch according to the article (i.e a happy customer)

 

[Eins San Luis]

Hey guys, I want to know if I can do the same with my ati 4850????