Intel HD 4600 vs. AMD A10-5800/6800K vs. GPU
With Haswell's IGP Intel had set itself ambitious goals. One for example is that they want to heat up the competition with AMD when it comes to integrated grahics performance. Wether Intel is going to succeed or not and how Intel's HD 4600 as well as AMD's HD 7660D and HD 8670D are going to stack up again discrete GPU's that's what we're going to show with this article.
Introduction
[pagebreak]
Technical Details
First
of all there are quite a few different integrated GPU's with Haswell, which is
why we've put together a small list:
- HD Graphics GT1 (Desktop)
- HD Graphics 4200 GT2 (Desktop)
- HD Graphics 4400 GT2 (Desktop)
- HD Graphics 4600 GT2 (Desktop)
- HD Graphics 5000 GT3 (Notebook)
- Iris Graphics 5100 GT3 (Notebook)
- Iris Pro Graphics 5200 GT3e (Notebook)
One
rather interesting thing is that the most powerful integrated GPU's didn't make
it's way into the desktop CPU's. Looking for the best performing desktop IGP
you'll find the offers to top out with the HD Graphics 4600 G2 offer. A closer
look at the notebook processors shows, that there are even HD Graphics 5000
Series IGP's and with the Iris and Iris Pro Graphics, or in other words GT3,
there are integrated graphics cores from Intel, which are supposed to bring
chipzilla's graphics power to another level.
So
far so good, there are quite a few new and different graphics processors within
Intel's CPU range, but what has actually been improved? Unfortunately there
isn't too much information on this topic, but at least there are some
interesting thnigs we can tell you. Checking out the GT2 IGP, there are 20 of
Intel's so called execution units and when looking at the GT3 IGP's there are no
less than 40 EU's. Maybe a short recap regarding the HD 4000 you got with Ivy
Bridge processors wouldn't be a bad idea at this point. A bit more an one year
ago Intel introduced an IGP, which came with 16 EU's. To put things into
persepective, the new HD 4600 IGP offers 25 percent more execution units, and it
might be interesting to see how this translates performance wise.
With
the desktop CPU's we see that there might be a 25 percent performance boost over
Ivy Bridge, but what's actually really interesting, and still not yet available
are the notebook graphics chips. With the HD 5000 Intel keeps upscaling their
existing architecure and in this case we find 40 execution units, which is 2.5
times more than what you got from Ivy Bridge. Thing get even more interesting
when checking out the HD 5200 GT3e. Next to the 40 execution units, there is a
128 Megabyte on-chip cache. The benefit of deploying an on-chip cache can bee
seen, when looking at the memory bandwidth. Whereas the DDR3-1600 Interface with
Ivy Bridge is capable of squeezing up to 25 Gigabyte of data through the bus,
the integrated cache offers twice as much, in other words 50 Gigabyte per
second.
According
to Intel 128 Megabyte of this eDRAM cache should be plenty. Nevertheless it
makes sense to offer a decently sized cache, since higher resolutions as well as
anti aliasing can quite quickly demand for a lot of cache. The downside of
adding quite a few more transistors to a processor is that it is automatically
going to burn more power. If the cache is under full load, the CPU will draw 4.5
Watt more power and when the cache is in idle, the additional power consumption
tops out at 0.5 Watt. As a desktop user these values don't make you worry, but
in a notbook, where manufacturers desperately try to keep power consumption as
low as possible, this means, that battery life will go down a bit.
At
last there are few other details Intel was working on. All new integrated
graphics units do now support DirectX 11.1, OpenGL 4.0 as well as OpenCL 1.2.
Furthermore Haswell CPU's will make it possible to send different pictures to up
to three independant displays. Another addition is DisplayPort 1.2 support and
improved 4K x 2K resolutions. Last but not least there have been improvements
when it comes to decoding JPEG as well as encoding MPEG.
[pagebreak]
Test Setup
| Motherboard |
- ASUS Maximus V Gene
- ASUS Maximus VI Gene
- ASUS F2A85-V Pro
|
| CPU |
- Intel Core i7-2600K @ Stock (Sandy Bridge)
- Intel Core i7-3770K @ Stock (Ivy Bridge)
- Intel Core i7-4770K @ Stock (Haswell)
- AMD A10-5800K @ Stock (Tirnity)
- AMD A10-6800K @ Stock (Richland)
|
| Graphics card
/ IGP |
- Intel HD 3000
- Intel HD 4000
- Intel HD 4600
- Intel HD 7660D
- Intel HD 8770D
- AMD Radeon HD 5550
- AMD Radeon HD 6770
- GTX550 Ti
|
| Driver |
- Intel 9.17.10.3062 (Sandy Bridge)
- Intel 9.18.10.3165 (Ivy Brdige & Haswell)
- Force Ware 320.49 NVIDIA
- Catalyst 13.6 AMD
|
| OS |
Windows 7 x64 |
| Benchmarks & Games |
- 3DMark
- Ice Storm
- Cloud Gate
- Fire Strike
- Unigine Heaven 4.0
- Basic Preset (Fullscreen)
- Extreme Preset (Fullscreen)
- Crysis 3
- 1280 x 1024 Low Detail
- 1920 x 1080 High Detail
- Dirt Showdown
- 1280 x 1024 Low Detail
- 1920 x 1080 High Detail
- Metro Last Light
- 1280 x 1024 Low Detail
- 1920 x 1080 High Detail
- Sleeping Dogs
- 1280 x 1024 Low Detail
- 1920 x 1080 High Detail
|
| SSD |
OCZ Octane 512 GB |
| Power
Supply |
Seasonic Platinum SS-1000XP / 1000 Watt |
[pagebreak]
3DMark & Unigine Heaven 4.0
3DMark
| 3DMark Ice Storm (Total) |
Points |
Index |
| HD 3000 (Sandy Bridge) |
33640 |
69.00 % |
| HD 4000 (Ivy Bridge) |
48752 |
100.00 % |
| HD 4600 (Haswell) |
52003 |
106.67 % |
| |
|
|
| HD 7660D (Trinity) |
52928 |
108.57 % |
| HD 8770D (Richland) |
54566 |
111.93 % |
| |
|
|
| GeForce GTX 550 Ti |
99738 |
204.58 % |
| Radeon HD 5550 |
48894 |
100.29 % |
| Radeon HD 6770 |
87720 |
179.93 % |
| 3DMark Ice Storm (Graphics) |
Points |
Index |
| HD 3000 (Sandy Bridge) |
31629 |
64.77 % |
| HD 4000 (Ivy Bridge) |
48836 |
100.00 % |
| HD 4600 (Haswell) |
53172 |
108.88 % |
| |
|
|
| HD 7660D (Trinity) |
60660 |
124.21 % |
| HD 8770D (Richland) |
61512 |
125.96 % |
| |
|
|
| GeForce GTX 550 Ti |
135138 |
276.72 % |
| Radeon HD 5550 |
49233 |
100.81 % |
| Radeon HD 6770 |
108619 |
222.42 % |
| 3DMark Cloud Gate (Total) |
Points |
Index |
| HD 3000 (Sandy Bridge) |
3674 |
59.75 % |
| HD 4000 (Ivy Bridge) |
6149 |
100.00 % |
| HD 4600 (Haswell) |
7893 |
128.36 % |
| |
|
|
| HD 7660D (Trinity) |
5687 |
92.49 % |
| HD 8770D (Richland) |
5918 |
96.24 % |
| |
|
|
| GeForce GTX 550 Ti |
12994 |
211.32 % |
| Radeon HD 5550 |
5929 |
96.42 % |
| Radeon HD 6770 |
12003 |
195.20 % |
| 3DMark Cloud Gate (Graphics) |
Points |
Index |
| HD 3000 (Sandy Bridge) |
3244 |
55.43 % |
| HD 4000 (Ivy Bridge) |
5852 |
100.00 % |
| HD 4600 (Haswell) |
7989 |
136.52 % |
| |
|
|
| HD 7660D (Trinity) |
7482 |
127.85 % |
| HD 8770D (Richland) |
7578 |
129.49 % |
| |
|
|
| GeForce GTX 550 Ti |
15992 |
273.27 % |
| Radeon HD 5550 |
5547 |
94.79 % |
| Radeon HD 6770 |
14147 |
241.75 % |
| 3DMark Fire Strike (Total) |
Points |
Index |
| HD 3000 (Sandy Bridge) |
no DX11 |
0.00 % |
| HD 4000 (Ivy Bridge) |
629 |
100.00 % |
| HD 4600 (Haswell) |
862 |
137.04 % |
| |
|
|
| HD 7660D (Trinity) |
963 |
153.10 % |
| HD 8770D (Richland) |
989 |
157.23 % |
| |
|
|
| GeForce GTX 550 Ti |
1980 |
314.79 % |
| Radeon HD 5550 |
641 |
101.91 % |
| Radeon HD 6770 |
1805 |
286.96 % |
| 3DMark Fire Strike (Graphics) |
Points |
Index |
| HD 3000 (Sandy Bridge) |
no DX11 |
0.00 % |
| HD 4000 (Ivy Bridge) |
681 |
100.00 % |
| HD 4600 (Haswell) |
934 |
137.15 % |
| |
|
|
| HD 7660D (Trinity) |
1025 |
150.51 % |
| HD 8770D (Richland) |
1053 |
154.63 % |
| |
|
|
| GeForce GTX 550 Ti |
2036 |
298.97 % |
| Radeon HD 5550 |
647 |
139.06 % |
| Radeon HD 6770 |
1840 |
270.19 % |
Unigine Heaven 4.0
| Unigine Heaven 4.0 Basic Preset |
Points |
Index |
| HD 3000 (Sandy Bridge) |
235 |
56.63 % |
| HD 4000 (Ivy Bridge) |
415 |
100.00 % |
| HD 4600 (Haswell) |
576 |
138.80 % |
| |
|
|
| HD 7660D (Trinity) |
572 |
137.83 % |
| HD 8770D (Richland) |
582 |
140.24 % |
| |
|
|
| GeForce GTX 550 Ti |
1282 |
308.92 % |
| Radeon HD 5550 |
418 |
100.72 % |
| Radeon HD 6770 |
1108 |
266.99 % |
| Unigine Heaven 4.0 Extreme Preset |
Points |
Index |
| HD 3000 (Sandy Bridge) |
no DX11 |
0.00 % |
| HD 4000 (Ivy Bridge) |
84 |
100.00 % |
| HD 4600 (Haswell) |
128 |
152.38 % |
| |
|
|
| HD 7660D (Trinity) |
142 |
169.05 % |
| HD 8770D (Richland) |
147 |
175.00 % |
| |
|
|
| GeForce GTX 550 Ti |
366 |
435.71 % |
| Radeon HD 5550 |
97 |
115.48 % |
| Radeon HD 6770 |
280 |
333.33 % |
[pagebreak]
Battlefield 3 & Bioshock: Infinite
Battlefield 3 - 1280 x 1024, Low Details
| Battlefield 3 - 1280 x 1024, Low Details |
fps |
Index |
| HD 3000 (Sandy Bridge) |
15.2 |
69.09 % |
| HD 4000 (Ivy Bridge) |
22.0 |
100.00 % |
| HD 4600 (Haswell) |
28.0 |
127.27 % |
| |
|
|
| HD 7660D (Trinity) |
26.8 |
121.82 % |
| HD 8770D (Richland) |
29.4 |
133.64 % |
| |
|
|
| GeForce GTX 550 Ti |
67.0 |
304.55 % |
| Radeon HD 5550 |
22.2 |
100.91 % |
| Radeon HD 6770 |
48.3 |
219.55 % |
Battlefield 3 - 1920 x 1080, High Details
| Battlefield 3 - 1920 x 1080, High Details |
fps |
Index |
| HD 3000 (Sandy Bridge) |
5.0 |
80.65 % |
| HD 4000 (Ivy Bridge) |
6.2 |
100.00 % |
| HD 4600 (Haswell) |
7.8 |
125.81 % |
| |
|
|
| HD 7660D (Trinity) |
8.8 |
141.94 % |
| HD 8770D (Richland) |
9.8 |
158.06 % |
| |
|
|
| GeForce GTX 550 Ti |
24.2 |
390.32 % |
| Radeon HD 5550 |
6.2 |
100.00 % |
| Radeon HD 6770 |
18.6 |
300.00 % |
Bioshock: Infinite - 1280 x 1024, Low Details
| Bioshock: Infinite - 1280 x 1024,
Low Details |
fps |
Index |
| HD 3000 (Sandy Bridge) |
14.45 |
58.79 % |
| HD 4000 (Ivy Bridge) |
24.58 |
100.00 % |
| HD 4600 (Haswell) |
28.15 |
114.52 % |
| |
|
|
| HD 7660D (Trinity) |
39.69 |
161.47 % |
| HD 8770D (Richland) |
40.39 |
164.32 % |
| |
|
|
| GeForce GTX 550 Ti |
94.85 |
385.88 % |
| Radeon HD 5550 |
28.85 |
117.37 % |
| Radeon HD 6770 |
74.67 |
303.78 % |
Bioshock: Infinite - 1920 x 1080, High Details
| Bioshock: Infinite - 1920 x 1080,
High Details |
fps |
Index |
| HD 3000 (Sandy Bridge) |
7.51 |
58.26 % |
| HD 4000 (Ivy Bridge) |
12.89 |
100.00 % |
| HD 4600 (Haswell) |
14.73 |
114.27 % |
| |
|
|
| HD 7660D (Trinity) |
21.41 |
166.10 % |
| HD 8770D (Richland) |
21.72 |
168.50 % |
| |
|
|
| GeForce GTX 550 Ti |
51.18 |
397.05 % |
| Radeon HD 5550 |
15.41 |
119.55 % |
| Radeon HD 6770 |
41.93 |
325.29 % |
[pagebreak]
Crysis 3 & Dirt Showdown
Crysis 3 - 1280 x 1024, Low Details
| Crysis 3 - 1280 x 1024, Low Details |
fps |
Index |
| HD 3000 (Sandy Bridge) |
no DX11 |
0.00 % |
| HD 4000 (Ivy Bridge) |
15.8 |
100.00 % |
| HD 4600 (Haswell) |
20.8 |
131.65 % |
| |
|
|
| HD 7660D (Trinity) |
25.6 |
162.03 % |
| HD 8770D (Richland) |
25.4 |
160.76 % |
| |
|
|
| GeForce GTX 550 Ti |
61.4 |
388.61 % |
| Radeon HD 5550 |
19.0 |
120.25 % |
| Radeon HD 6770 |
49.6 |
313.92 % |
Crysis 3 - 1920 x 1080, High Details
| Crysis 3 - 1920 x 1080, High Details |
fps |
Index |
| HD 3000 (Sandy Bridge) |
no DX11 |
0.00 % |
| HD 4000 (Ivy Bridge) |
1.6 |
100.00 % |
| HD 4600 (Haswell) |
2.4 |
150.00 % |
| |
|
|
| HD 7660D (Trinity) |
3.0 |
187.50 % |
| HD 8770D (Richland) |
3.0 |
187.50 % |
| |
|
|
| GeForce GTX 550 Ti |
7.4 |
462.50 % |
| Radeon HD 5550 |
1.8 |
112.50 % |
| Radeon HD 6770 |
6.2 |
387.50 % |
Dirt Showdown - 1280 x 1024, Low Details
| Dirt Showdown - 1280 x 1024, Low Details |
fps |
Index |
| HD 3000 (Sandy Bridge) |
33.18 |
68.60 % |
| HD 4000 (Ivy Bridge) |
48.37 |
100.00 % |
| HD 4600 (Haswell) |
70.70 |
146.16 % |
| |
|
|
| HD 7660D (Trinity) |
78.05 |
161.36 % |
| HD 8770D (Richland) |
80.03 |
165.45 % |
| |
|
|
| GeForce GTX 550 Ti |
170.67 |
352.84 % |
| Radeon HD 5550 |
59.83 |
123.69 % |
| Radeon HD 6770 |
146.80 |
303.49 % |
Dirt Showdown - 1920 x 1080, High Details
| Dirt Showdown - 1920 x 1080, High
Details |
fps |
Index |
| HD 3000 (Sandy Bridge) |
13.66 |
98.63 % |
| HD 4000 (Ivy Bridge) |
13.85 |
100.00 % |
| HD 4600 (Haswell) |
13.79 |
99.57 % |
| |
|
|
| HD 7660D (Trinity) |
13.98 |
100.94 % |
| HD 8770D (Richland) |
14.10 |
101.81 % |
| |
|
|
| GeForce GTX 550 Ti |
20.61 |
148.81 % |
| Radeon HD 5550 |
13.82 |
99.78 % |
| Radeon HD 6770 |
20.94 |
151.19 % |
[pagebreak]
Metro Last Light & Sleeping Dogs
Metro Last Light - 1280 x 1024, Low Details
| Metro Last Light - 1280 x 1024, Low Details |
fps |
Index |
| HD 3000 (Sandy Bridge) |
11.71 |
61.47 % |
| HD 4000 (Ivy Bridge) |
19.05 |
100.00 % |
| HD 4600 (Haswell) |
23.92 |
125.56 % |
| |
|
|
| HD 7660D (Trinity) |
24.73 |
129.82 % |
| HD 8770D (Richland) |
25.32 |
132.91 % |
| |
|
|
| GeForce GTX 550 Ti |
66.03 |
346.61 % |
| Radeon HD 5550 |
20.68 |
108.56 % |
| Radeon HD 6770 |
51.34 |
269.50 % |
Metro Last Light - 1920 x 1080, High Details
| Metro Last Light - 1920 x 1080, High Details |
fps |
Index |
| HD 3000 (Sandy Bridge) |
1.87 |
61.61 % |
| HD 4000 (Ivy Bridge) |
3.05 |
100.00 % |
| HD 4600 (Haswell) |
3.98 |
130.49 % |
| |
|
|
| HD 7660D (Trinity) |
4.62 |
151.48 % |
| HD 8770D (Richland) |
4.71 |
154.43 % |
| |
|
|
| GeForce GTX 550 Ti |
10.01 |
328.20 % |
| Radeon HD 5550 |
2.65 |
86.89 % |
| Radeon HD 6770 |
7.00 |
229.51 % |
Sleeping Dogs - 1280 x 1024, Low Details
| Sleeping Dogs - 1280 x 1024, Low Details |
fps |
Index |
| HD 3000 (Sandy Bridge) |
16.8 |
52.34 % |
| HD 4000 (Ivy Bridge) |
32.1 |
100.00 % |
| HD 4600 (Haswell) |
37.1 |
115.58 % |
| |
|
|
| HD 7660D (Trinity) |
48.1 |
149.84 % |
| HD 8770D (Richland) |
47.5 |
147.98 % |
| |
|
|
| GeForce GTX 550 Ti |
110.5 |
344.24 % |
| Radeon HD 5550 |
40.3 |
125.55 % |
| Radeon HD 6770 |
103.3 |
321.81 % |
Sleeping Dogs - 1920 x 1080, High Details
| Sleeping Dogs - 1920 x 1080 High
Details |
fps |
Index |
| HD 3000 (Sandy Bridge) |
FAIL |
0.00 % |
| HD 4000 (Ivy Bridge) |
5.9 |
100.00 % |
| HD 4600 (Haswell) |
5.9 |
100.00 % |
| |
|
|
| HD 7660D (Trinity) |
6.0 |
101.69 % |
| HD 8770D (Richland) |
6.1 |
103.39 % |
| |
|
|
| GeForce GTX 550 Ti |
11.7 |
198.31 % |
| Radeon HD 5550 |
5.0 |
84.75 % |
| Radeon HD 6770 |
13.0 |
220.34 % |
[pagebreak]
Performance Index
| |
Index |
| HD 3000 (Sandy Bridge) |
65.36 % |
| HD 4000 (Ivy Bridge) |
100.00 % |
| HD 4600 (Haswell) |
127.68 % |
| |
|
| HD 7660D (Trinity) |
139.98 % |
| HD 8770D (Richland) |
143.47 % |
| |
|
| GeForce GTX 550 Ti |
318.61 % |
| Radeon HD 5550 |
107.46 % |
| Radeon HD 6770 |
267.13 % |
[pagebreak]
Conclusion
At this
point we will first of all talk about Haswell before we then move on to discuss
global improvements regarding integrated graphics processors.
During our testing of Haswell we soley focussed on the HD 4600 graphics unit.
Since Intel increased the number of execution units for the HD 4600 by 25
percent when directly comparing it to Ivy Bridge's HD 4000, we expected that the
overall performance improvement should also be somewhere in the region of 25
percent. According to our performance index we've seen performance go up by 27
percent. After all it is quite surprising to see that there is an overall almost
perfect linear scaling with the number of execution units. Percentage wise the
largest jumps were when it comes we look at Unigine Heaven 4.0 Extreme Preset.
In this case the HD 4600 is a whopping 52 percent faster than the HD 4000. When
it comes to games the largest improvement was seen when running Crysis 3, where
we measured 50 percent difference.
Overall we can say that Intel's HD 4600 IGP is suitable for casual gaming at
1280 x 1024 resolution with low graphics details. In certain cases it would even
be possible to increase the detail level but in other cases is would also make
sense to set an even lower resolution. What you really shouldn't expect are
miracles: the HD 4600 is an improvemtn over the HD 4000 but it's certainly not a
revolution. It will be very interesting to see what the HD 5200 Pro is capable
of. Since we've seen an almost linear scaling from HD 4000 to HD 4600, when
comparing performance numbers with the number of execution units, we would
expect HD 5200 Pro to be more than twice as fast than HD 4600. In this case you
could then even consider to choose a higher resolution than 1280 x 1024 in games
our increase the detail level significantly.
Looking at IGP's these day on a global scale by putting things into
perspective, while adding AMD's A10-5800K as well as A10-6800K to the comparison
tables, we see that these days, IGP's can compete with discrete entry level
graphics cards from three and a half years ago. Since we were interested to see
how much faster a lower mid range card is (GTX 550 Ti) we ran the same tests on
this card and ou performance index tells us that there is almost a factor of
three inbetween this card and today's IGP's.
At this point we want to put Intel's improvements into perspective. Maybe you
remember that Sandy Bridge first came to market roughly two years ago. Within
those two years Intel has been able to double graphics performance of their
IGP's, which means that even in this context Intel is considering Moore's Law as
the way to go to be successful in this industry. And then the company still has
an ace up their sleeve with HD 5200 Pro, which will certainly be substantially
faster than the HD 4600 we've tested here. Overall this also means that Intel is
starting to apply pressure on AMD, when it comes to IGP's. AMD will soon have to
release there GCN based APU's to remain the leading company, when it comes to
IGP's.
Article in English 
Artikel in Deutsch
Article en français
Articolo in italiano
