Published in Reviews

Zotac 8500GT, the fastest G86 core

by on30 May 2007

Index

Image


Review:
Sold overclocked to 700 MHz

 

 

We continued to test cheap DirextX 10 cards and today we bring you the Zotac Geforce 8500GT. Zotac is a fairly new name in the industry - it is in fact Nvidia's PC Partner brand. Zotac launched a number of their products based on Nvidia's DirectX 10 chips and they are currently trying hard to get their piece of the market. Many of their products are over-clocked and are running on greater speeds than Nvidia’s reference clocks. The Zotac GeForce 8500GT is no exception.

The Zotac 8500GT (SKU#: ZT-85TE250-FPP) is a card with the fastest core among the GeForce 8500GT family, at least among the ones we've tested so far. The core runs at 700MHz which is 250MHz faster than Nvida's default speed for the 8500GT which is 450MHz. Quite an impressive start for the Zotac 8500GT card.

 Image

 


However, the memory is still at the default 800MHz. Zotac didn't opt for GDDR3 memory, although we've seen some Nvidia partners, such as Gainward do this. With such a fast core, it is quite sad to see the memory not running at greater speed. The only way to change this would be over-clocking, and we will be talking about it later on, but we'll tell you this - we were impressed with the Hynix HY 5PS561621A-FP25 memory chips' capabilities. These modules are rated at 2.5ns which makes for 400MHz, and we easily ran them at 530MHz and 1060MHz in GDDR2 mode. Whether this is enough to beat Gainward's GDDR3 memory, we'll find out later. The memory interface is still 128 bit and stays the same no matter what type of memory used. That was Nvidia's decision and it applies to all 8600 and 8500 cards, meaning G84 and G86 cores. On the GeForce 8500GT with 256 MB of memory, the bandwidth measures up to 12.8 GB/s, the core is 80nm and it has 16 Stream Processors. Image


Zotac also worked on the design - the I/O side has DVI and VGA connectors painted yellow on black metal, and it fits in with the rest of the card. The cooler is small but sufficient for a G86, which is quite small itself. The fan would sometimes run a bit louder than usual, but it's bearable. The RoHS sticker on the back of the card says that the materials used in the manufacturing process contained no harmful elements.

Image

The picture on the box shows the same macho Zotac soldier we saw on the card's cooler. It also has some additional info like "DVI to HDMI adapter". That's correct, besides the S-Video and component cable as well as a CD containing the drivers, you'll find the afore-mentioned adapter. It seems that in this range of cards, Zotac is the only one giving away such a juicy gift.

Image







Testbed

Image


Motherboard:
Nforce 680i EVGA board (supplied by EVGA)
 
CPU:
Intel Core 2 Duo 6800 Extreme Edition (supplied by Intel)

CPU Cooler:
Akasa EVO AK 922 Blue cooler for Athlon 64/X2/FX and Intel CPU's (supplied by Akasa)

Memory:
OCZ Reaper PC2-8500 1,066MHz 5-5-5-15  (supplied by OCZ)
CL5-5-5-15-CR2T at 2.3V

Graphics Card:
Zotac Geforce 8500GT (supplied by Zotac)

Gainward BLISS 8500GT Golden Sample PCX (supplied by Gainward)

Asus EN8600GT (supplied by Asus)

PSU:
OCZ GameXStream 700W (supplied by OCZ)

Hard Drive:
Seagate Barracuda 7200.9 500GB SATA (supplied by Seagate)






Testing and overclocking

Today we're looking at a GeForce 8500GT comparison test. The Gainward 8500GT utilizes better GDDR3 memory running at 1,400 MHz, while Zotac sticks to the Nvidia's reference design using GDDR2 memory running at 800MHz. There is quite a difference in speed between these two, and we sincerely doubt that Zotac, with its 100 MHz faster core, will compensate for its 600 MHz slower GDDR2 memory. Whether GDDR3 and its speed play an important role - we'll see soon enough.

The Zotac 8500GT card with 700MHz core speed is much faster than the default 450 MHz of most other 8500GTs.

To refresh the table we incorporated the results of the recently tested Asus EN8600GT card, which runs at Nvidia's default speed of 540/1,400 MHz. We will not discuss this card's results as it's not in the same performance or price range as the GeForce 8500GT cards. If you want to know more about Asus EN8600GT card, you can read the recent review here.

As far as over-clocking goes, Zotac pushed the card almost to its maximum. We still managed to squeeze out an additional 45 MHz from the core, so we reached a final core speed of 745 MHz. The card handled these speeds well, but for maximum stability in Fear and Lost Planet we lowered the frequency to 735 MHz. We mentioned the memory before and we said it runs stable at 1,060 MHz. We over-clocked it to run 130 MHz faster than the default 400 MHz, which is a good result for 2.5 ns memory.

3DMarks

Composite Figures 3Dmark 03

3DMark 03

Game2

Game3

Game4

8500GT referenc 450/800 MHz

8328

61.4

51.0

48.4

Zotac 8500GT 700/800 MHz

10251

72.0

61.9

66.4

Gainward BLISS 8500GT PCX 600/1400 MHz

11196

89.3

69.7

64.2

Asus EN8600GT 540/1400MHz

15247

105.5

88.7

107.0

 

 

 

 

 

Composite Figures 3Dmark 03

Single Texturinng

Multi Textur.

Vertex Shader

Pixel Shader 2.0

8500GT referenc 450/800 MHz

1324.4

2885.5

29.0

112.8

Zotac 8500GT 700/800 MHz

1328.1

3988.9

38.3

127.9

Gainward BLISS 8500GT PCX 600/1400 MHz

2164.8

3721.2

34.0

164.4

Asus EN8600GT 540/1400 MHz

2023.7

7028.5

55.1

168.2

 

 

 

 

 


The Zotac 8500GT scored 10,251 which is 945 marks less than the Gainward 8500GT. Almost a thousand marks less for a card with the same chip (running at 100MHz faster) is not a good result, and it is most certainly caused by the weak DDR2 memory.

The core speed helps in Game4 (Nature Test) where Zotac emerges as the winner. Different results in the Composite part of 3DMark are again caused by Zotac’s faster core or Gainward’s greater memory bandwidth. Vertex Shader test resulted in lower performance from Gainward, but that’s because memory doesn’t play a significant role in this test. In Single Texturing and Pixel Shader 2.0 tests memory plays a key role – and GDDR3 memory with greater bandwidth clearly secures a win. Multitexturing test profits from core-speed that draws 64 texture layers on objects as quick as possible.

We can’t compare the Asus 8600GT with the two 8500GT cards because it has 32 Stream Processors, which is twice as many as the 8500GT cards.

By comparing the Zotac 8500GT with Nvidia’s reference card, we see that the over-clocked core brings Zotac to the winner’s spot with a 2,000 mark advantage.

Composite Figures 3Dmark 05

3DMark 05

Game1

Game2

Game3

8500GT referenc 450/800 MHz

4725

22.5

14.4

20.9

Zotac 8500GT 700/800 MHz

6165

28.8

19.1

27.2

Gainward BLISS 8500GT PCX 600/1400 MHz

6327

30.0

19.2

28.1

Asus EN8600GT 540/1400 MHz

8924

38.5

28.7

41.2

 

 

 

 

 

Composite Figures 3Dmark 05

Single Texturinng

Multi Textur.

Pixel Shader

VS/VS
complex

8500GT referenc 450/800 MHz

1464.0

3007.3

62.3

54.0/53.0

Zotac 8500GT 700/800 MHz

1442.3

5347.8

83.7

54.0/59.9

Gainward BLISS 8500GT PCX 600/1400 MHz

2230.7

4196.0

85.4

88.4/74.7

Asus EN8600GT 540/1400MHz

2166.7

7399.2

140.1

63.0/95.1


In 3DMark 05 Gainward’s 8500GT card, which uses GDDR3 memory, beats Zotac 8500GT, but only by 162 marks. Single and Multi Texturing tests work on the same principles as the 3DMark 03 tests so the results are similar. The Pixel Shader test should depend on the memory bandwidth, but here we see Zotac’s 12.8GB/s is enough to compete with the Gainward 8500GT. That is not the case in the Vertex Shader Simple/Complex test though, as Gainward thanks to its memory, emerges as the winner.

The Zotac Geforce 8500GT is still significantly ahead of the reference card.

Composite Figures 3Dmark 06

3DMark 06

 

 

 

8500GT referenc 700/800 MHz

2291

 

 

 

Zotac 8500GT 700/800 MHz

2936

 

 

 

Zotac 8500GT 745/1060 MHz

3447

 

 

 

Gainward BLISS 8500GT PCX 600/1400 MHz

3101

 

 

 

Asus EN8600GT 540/1400MHz

4513

 

 

 

 

 

 

 

 

SM2.0 Test

Score

GT1

GT2

 

8500GT referenc 700/800 MHz

878

6.913

7.726

 

Zotac 8500GT 700/800 MHz

1158

9.279

Last modified on 31 May 2007
Rate this item
(0 votes)