Battle of the lightweights: AMD 7660G vs Intel HD4000 IGP

There is a lot of buzz about Intel's upcoming HD 5000 series, also coined GT2, GT3, integrated graphics in the Haswell CPU's. But before we get there I thought it might be good to set a baseline for the current generation of mobile IGP's from both AMD and Intel. AMD's flagship IGP in the Trinity chip is coined the 7660G, and the Intel Ivy Bridge chips sport the HD 4000 IGP.

While they are not 3D powerhouses by any means, they actually can do an adequate job with gaming in a pinch, as long as you stick with 1280x720 or 1366x768 resolution. Anything higher and they usually take a significant hit in performance. Both utilize system RAM for the video RAM. The 7660G is limited to 512MB max, whereas the HD 4000 is pretty much limited to the amount of system RAM that the system has. But honestly, the limited performance of the GPU's won't be able to take advantage of much more than 512GB RAM anyhow.

The systems that were tested are as specified:

HP Probook 6475b
AMD A10-4600m with 7660G IGP / Driver Catalyst 13.5 beta 3
2x4GB DDR3 1600MHz RAM (Samsung)
14" 1600x900 LCD
Intel X25-M G2 120GB SSD

Sager NP6110 (Clevo W110ER)
Intel i7-3610QM with HD 4000 IGP / Driver 15.31.3.64.3071
GT 650m (not used for these tests obviously)
2x4GB DDR3 1600MHz RAM (Samsung)
and for overclock: 2x4GB DDR3 2133MHz RAM (Kingston)
11.6" 1366x768 LCD
Samsung 840 500GB SSD

Before we pit these weaklings in battle, let's take a look at the details of each of the GPU's:

The AMD Radeon 7660G has a stock and fixed clock of 685MHz.

AMD Radeon 7660G



The Intel HD 4000 has a base clock of 650MHz but can boost to 1100MHz while active.

Intel HD 4000


From these images you can quickly tell the AMD IGP has significantly better specs, so it should perform better. 8 vs 2 ROPS, 48 vs 4 Texture Mapping Units, Pixel fill rate of 5.5 vs 1.3 GP/s, Texture fill rate 32.9 vs 2. GTexel/s. RAM bandwidth in these units are identical, since they have same RAM chips at 1600MHz.

Each of the IGP's were tested at stock settings as well as overclocked in the best way they could be overclocked. The AMD GPU can use AMD Overdrive to overclock the GPU. It was successfully overclocked to 900Mhz from stock 685MHz. The CPU is also limited to its peak speed of 3.2GHz, although it tends to not exceed 2.7GHz when fully loaded). The system RAM is not able to run faster than 1600MHz due to limitations of the CPU and system.

For the Intel system, special thanks to NBR member Prema for a modded BIOS that allowed overclock of the HD 4000. It was successfully overclocked from stock 1100MHz boost to 1300MHz boost and sytem RAM run at 2133MHz instead of 1600MHz. For the overclocked

See the GPU-Z speeds of the stock and overclocked IGP's:

AMD stock speeds on left, overclock on Right (GPU Temps are NOT correct):



Intel stock speeds on left, overclock on right:



BENCHMARKS

Legend for the benchmark results

- "7660G STOCK" = AMD 7660G STOCK CLOCKS (GPU @ 685MHz, CPU @ 2.7GHz)
- "7660G OC" = Overclocked AMD 7660G GPU @ 900MHz, CPU @ 2.7GHz
- "HD 4000 STOCK" = Intel HD 4000 STOCK CLOCKS (GPU @ 1100MHz, CPU @ 2.4-3.1GHz)
- "HD 4000 OC" = Overclocked Intel HD 4000 GPU @ 1300MHz, CPU @ 3.1GHz Throttlestop

Benchmarks run at 720p unless otherwise noted.

Benchmark	HD 4000	HD 4000 OC	7660G	7660G OC
ARTIFICIAL BENCHMARKS
3DMark Vantage	2567	5354	4651	4968
3DMark Vantage Graphics	2003	4303	4253	4553
3Dmark 11	470	883	1295	1328
3DMark 11 Graphics	399	752	1214	1235
3DMark 13 Ice Storm	35547	56987	42498	45220
3DMark 13 Ice Storm Graphics	37378	65803	57892	62347
3DMark 13 Cloud Gate	3689	6593	4170	4371
3DMark 13 Cloud Gate Graphics	3549	6814	6567	6568
3DMark 13 Fire Strike	371	748	796	814
3DMark 13 Fire Strike Graphics	411	810	863	879
Allbenchmark Catzilla RC3 CAT	1251	1516	1531	1598
Unigine Heaven Score	292	293	538	556
Unigine Heaven FPS	11.6	11.6	21.3	22.1
GAME BENCHMARKS
Bioshock Infinite Avg FPS (High)	19.3	23.2	26.0	28.1
DiRT 3 Avg FPS (High)	38.5	46.5	40.6	45.8
Far Cry 3 Avg FPS (Low)	29	36	33	40
HAWX 2 (High)	69	81	58	65
Just Cause 2 (High)
Concrete Jungle Avg FPS	14.5	17.1	24.4	26.0
Desert Sunrise Avg FPS	23.7	28.4	48.1	53.0
Dark Tower Avg FPS	14.6	17.4	31.3	32.6
Metro Last Light Avg FPS (Lowest)	15.0	15.3	23.0	25.0
Resident Evil 6 Score (Low)	1712	2056	2633	3111
Skyrim Avg FPS (High NoAA 16xAF)	25	30	33	38
Sleeping Dogs Avg FPS (Med)	35.1	42.3	38.3	45.3
Tomb Raider Avg FPS (Normal)	16.5	20.1	36.9	41.0
Trackmania Canyon Avg FPS (Track A05 - Nicer)	35	43	56	57

Graphs for benchmarks and comments are below in spoiler tag:

Spoiler

ARTIFICIAL BENCHMARKS

3DMark Vantage



3DMark 11



3DMark 13
Ice Storm


Cloud Gate


Fire Strike



Allbechmark Catzilla RC3 - CAT setting



Unigine Heaven



----------------------------------------

GAME BENCHMARKS


Bioshock Infinite

Playable on lower detail resolutions than high that was tested for all four IGP conditions.

DiRT 3

Plays beautifully on all four IGP conditions.

Far Cry 3

Fairly smooth with an OC 7660G, stock was a little choppy. HD 4000 was very choppy in either case.

HAWX 2

No issues playing with any IGP condition. Although somehow Intel trumps AMD in this game.

Just Cause 2

Wouldn't recommend with an HD4000, but 7660G is very playable, especially if you drop detail a little bit.

Metro Last Light

Surprisingly this ran ok on the 7660G OC. Borderline playable.

Resident Evil 6

Low FPS in the benchmark, I can't imagine it would be too enjoyable with the final game.

Skyrim

Playable for all IGP conditions, HD 4000 stock would dip to lower FPS while gaming though.

Sleeping Dogs

Very playable for all IGP conditions. Impressed.

Tomb Raider

This is not a game for Intel's IGP. AMD Radeon HD 7660G handles it very well. Double the performance of the Intel.

TrackMania Canyon

Very fun game with well optimized engine plays on pretty much any hardware.

----------------------------------------


Power Consumption

This is system power consumption. So it would all depend on the system, so maybe not a completely fair assessment since a ULV dual core Intel power usage would be considerably less, but then I believe the FPS would drop considerably due to CPU performance. In any case, the power consumption is about 10-15% less on the AMD system.

Temperatures

Like power consumption, temperatures are dependent on the system, and admittedly the Sager NP6110 tends to run a little hotter than if it were in a larger notebook. But just looking at the AMD 7660G, temperatures are remarkably low for the level of performance it offers.

FINAL THOUGHTS

This little experiment showed us several things. For one, the 7660G is clearly a better performer for games despite being outclassed in CPU performance by the Intel quad core. This can also be achieved with low temperatures and power consumption. The AMD IGP seems to tip the scale between making more recent AAA titles playable than an annoying stutter-fest or constant dips below playable FPS like is frequently found with the HD 4000. Average FPS shown do not necessarily tell the full story either. In general the HD 4000 had more noticeable frame lag or choppiness in general compared with the Radeon HD 7660G. Of course with Haswell this may likely change soon, although early indications shows GT2 likely won't outperform AMD's current Trinity generation of IGP's, the 7660G being the current flagship. We will see once the new Intel CPU's get released and I get my hands on one.

 

Reserved for future post.

 

wow great review HT :thumbsup:

 

questions:
1. did you use the new driver from intel?
2. can you do a shogun 2 run?
3. do you give ice cream to those who read it all?
4. Can you post a graph showing the variations of the FPS in those benchies?

and good job

 

Good point, I didn't indicate drivers, will update with driver versions used, but yes it was the latest from AMD and Intel.

Don't own Shogun 2, so no, unless there's an independent benchmark I'm not aware of. (I guess I cold buy it, have a Shogun 2 coupon on Steam to get it for $7.50)

No, but I'll give you a cookie instead.

You mean real-time FPS, or min-max-avg? I didn't record those. It was more or less to check (a) difference in general performance between mobile Intel and AMD IGP's and (b) what games, even modern ones, are playable with them. I suppose I could run the OC versions again and record real time FPS.

Thanks. You too, good points.

 

Are those homemade cookies? if so yes! I never had those, only store bought

Ah well, too bad about shogun 2 I do own it, I wonder if we can make something out of it, what TW do you own? ETW? NTW? though earlier titles didnt have benchies built in (nor S2tw it was added on a patch)

Regarding the fps I meant real time, its more valuable to see what happened and where, but I got your point, its indeed a good thing to see what runs and how on those lightweights. I wonder what happens on the battle of HD 5200 and kaveri, thats one Im anxious to see.

 

Man, great report, thanks for the info! I look forward to seeing how Haswell comes into play! Curious, I wasn't aware overclocking the HD4000 was possible, is it safe enough to run overclocked regulary?

I know this took a good bit of time, thanks for putting it together. +1

 

HD 4000 requires a BIOS that will allow you to use faster clocked RAM as well as clocking the GPU higher. If your board supports Intel XTU then likely you can do the same. But Intel is less likely and more difficult to overclock than AMD where you just need software to do so.

I'll run Hitman Absolution and Shogun 2 later today.

edit: downloaded Shogun 2 demo today and wouldn't work unless I forced windowed mode in the cfg file. Sorry not gonna spend a dime on something that's broken after being released a couple years.

edit edit: Looks like it may be an issue with drivers somehow.

 

Truly remarkable thread! Kudos to HTWingNut for the amount of work done!

 

Interesting and very nice work. And by very nice work also looks like very lot of work, good job.

Clear the AMD kicks butt. BIOS vs overdrive? I know you know but talk about unfair. Heroic level of effort on the Intel vs off the shelf soft mod OC on AMD. Sure already a test bed issue and of course I do understand why but wow what an imbalance and deviation from even the appearance of level playing field. Don't get me wrong I think what you did is very cool but...... One OC likely reproducible on many/most AMD vs OC that is 100% system specific.

I am very impressed with your HD4000 OC and that deserves evaluation in and of itself I would say in OC context. Less so as the cross platform comparison.

Is there a reason BIOS mod cannot be done on the AMD or just no one has done? If can't be done then from an OC standpoint important.

Once again the work in and of itself is very cool. The level of work is outstanding and worthy of recognition. Problem is that really in any true sense it is not as laid out what it claims to be. It is about as valid as comparing a HD400 to a 650m. I can draw out the relevant data and analyze I just find the presentation lacking in any "scientific" comparative construct while being presented as such. You want to compare the two units fine but it is not a fair GPU comparison. Your lack of ability to OC AMD to the level I suspect might your shortcoming and/or unit.

I did not pick the title. It is really the battle of the HP Probook 6475bvs the Sager NP6110 (Clevo W110ER).

I do question why Intel set or settled the HD4000 at the levels they did with such seemingly hidden performance headroom. I do know that the variation in chip performance does exist. But the extreme variation in performance above from a chip which we do not know if high or low characteristics almost certainly would have allowed some level of increased clocks across the entire line.

AMD's performance crushes Intel. At such a level that even if you believe Intel's claims on Haswell AMD will still hold the performance crown. Intel claims 50% increase but the AMD tested here has in my estimation a 80%/100% performance advantage over the Intel.

 

I think it would be embarrassing for AMD if Intel EVER released a faster igpu. It will never happen while the 2 companies interact the way they do now. Haswell claims 50% increase in performance. Fair enough but AMD's next release will also be better.

 

Well it kind of is HP vs Sager notebooks, but in reality the CPU performance is the CPU performance, and likely won't vary much from machine to machine configured similarly. Items unique to the system would be the temperature and power measurements.

I'm sure a BIOS could be modded for the AMD as well, but finding the right person to do so is problematic. Prema is well known here and at other tech sites for modifying Clevo BIOS especially, so am fortunate in that respect. I just find it odd how so much pomp and circumstance around the Intel GT2/GT3 when the current, nearly year old tech from AMD will likely be as powerful if not slightly better, and their upcoming GPU tech will put GT3 to shame.

I'll be sure to add a chart to show relative performance setting HD 4000 stock as the baseline so you can see actual improvement.

 

Added % improvement over stock HD 4000 to main post.