HD 4650 Low profile 64bit AX4650 512MD2-LHV2 review
Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee
-
- -- Vendor --
- Posts: 372
- Joined: Sun Jul 12, 2009 10:30 am
- Location: Canada
- Contact:
HD 4650 Low profile 64bit AX4650 512MD2-LHV2 review
This is a strange video card I just purchased.
I needed a cheap video card and this was on sale at newegg.ca if I get my 15$ mail in rebate it will end up at 32$ + 9.82$ shipping.
For the money I will be able to get a fanless 4350 but I was curios how a 4650 will perform with only half of the memory bus used.
4650 uses RV730 PRO so 320 stream processors compared to 80 on the RV710 used in 4350 and 4550
The strange fact about this card is that it uses only 64bit for memory bus and this is half of the 128bit the RV730 can address. I only found this made by Powercolor and one made by Asus but I fond no benchmark or reviews on this 64bit 4650.
It seems that memory bandwidth will limit the ability of this GPU.
I made some benchmarks using 3Dmark 2006 and also made some power consumption measurements.
System was G31 MB with 4GB DDR2/800 and Q8400s at 2.66Ghz
See this Table
Just curios if you have a 4350 with DDR2 600/400 can you do a 3Dmark 2006 test to compare the score
I needed a cheap video card and this was on sale at newegg.ca if I get my 15$ mail in rebate it will end up at 32$ + 9.82$ shipping.
For the money I will be able to get a fanless 4350 but I was curios how a 4650 will perform with only half of the memory bus used.
4650 uses RV730 PRO so 320 stream processors compared to 80 on the RV710 used in 4350 and 4550
The strange fact about this card is that it uses only 64bit for memory bus and this is half of the 128bit the RV730 can address. I only found this made by Powercolor and one made by Asus but I fond no benchmark or reviews on this 64bit 4650.
It seems that memory bandwidth will limit the ability of this GPU.
I made some benchmarks using 3Dmark 2006 and also made some power consumption measurements.
System was G31 MB with 4GB DDR2/800 and Q8400s at 2.66Ghz
See this Table
Just curios if you have a 4350 with DDR2 600/400 can you do a 3Dmark 2006 test to compare the score
Last edited by electrodacus on Tue Apr 13, 2010 10:07 pm, edited 1 time in total.
-
- Posts: 419
- Joined: Mon Aug 31, 2009 12:30 am
- Location: Italy
I was about to purchase one of these before I realized they came with that damn crippled bus width. Actually other vendors who use the 64 bit bus for a 4650 model are Sapphire (the one which has a cooler with no plastic shroud and black fan) and Asus (their 512 mb model has half of the bus data width compared to the 1 gb one).
Thanks for the comparative results. Will try to do some testing with my 4550 clocked at the speeds you ask, sometime in the future. Were you able to get the memories anything higher than 500 MHz?
Thanks for the comparative results. Will try to do some testing with my 4550 clocked at the speeds you ask, sometime in the future. Were you able to get the memories anything higher than 500 MHz?
-
- -- Vendor --
- Posts: 372
- Joined: Sun Jul 12, 2009 10:30 am
- Location: Canada
- Contact:
It will be interesting to see the results. I used Ati Catalyst to overclock the memory and 500MHz is the maximum is allowed. They also have a Auto overclock or something like this and after some tests the selection was 640MHz GPU and 490MHz memory so 10Mhz lower than max allowed but I used at 500Mhz with no problem in benchmarks.Parappaman wrote: Thanks for the comparative results. Will try to do some testing with my 4550 clocked at the speeds you ask, sometime in the future. Were you able to get the memories anything higher than 500 MHz?
-
- Posts: 419
- Joined: Mon Aug 31, 2009 12:30 am
- Location: Italy
-
- -- Vendor --
- Posts: 372
- Joined: Sun Jul 12, 2009 10:30 am
- Location: Canada
- Contact:
Will this work with cards from other manufacturer? Fan is not a problem it will be removed and replaced with a larger an much more quiet one.Parappaman wrote:You definitely have to take a look at MSI Afterburner, much more flexible clocking and an excellent fanspeed configuration to boot.
Also I just overclocked for benchmark but i will used default or underclocked.
-
- Posts: 419
- Joined: Mon Aug 31, 2009 12:30 am
- Location: Italy
-
- -- Vendor --
- Posts: 372
- Joined: Sun Jul 12, 2009 10:30 am
- Location: Canada
- Contact:
Thanks, I just downloaded the MSI Afterburner and it works but within the same limits 650MHz for GPU and 500Mhz for memory I can select max 520MHz for memory but if I apply it will jump to the last correct value.Parappaman wrote:Yes, it works with any card as it is based on RivaTuner's engine. I'm quite busy these days so won't be able to do the testing, but I'll be back with the results soon.
Though I have an E8400 @ 3.6 GHz plus 900+ MHz memories, so the results may be quite different due to cpu.
I also did some tests using all 4 cores on my CPU and with only one core you can set this in Windows or Linux and it seems the 3DMark 06 is influenced by the CPU with all 4 cores I get 3169 points and with only one core I get 2858 points settings are 600Mhz GPU and 500MHz memory the CPU is default at 2.66Ghz.
ZOOM
ZOOM
-
- Posts: 419
- Joined: Mon Aug 31, 2009 12:30 am
- Location: Italy
So, I finally ran this 3D Mark thing
At 600/450 clocks (can't get to 400 on memories as the 4550 is equipped with GDDR3 which can't clock that low) my score is 2543, though I have an E8400 clocked at 3,6 GHz so this should have improved the score a bit. Clocking the memories to the maximum stable value, which is 873 MHz, I got a much neatier 3954 points, which is already higher than your 4650, showing that this tight memory bus is really the limiting factor. In fact, by going haywire on the core at 745 MHz the improvement is much less pronounced, getting to 4334 points.
So, the bottom line is: I did a good thing by going for a 4550 instead of one of those crippled 4650, as it turned out to be a higher performing card when overclocked.
At 600/450 clocks (can't get to 400 on memories as the 4550 is equipped with GDDR3 which can't clock that low) my score is 2543, though I have an E8400 clocked at 3,6 GHz so this should have improved the score a bit. Clocking the memories to the maximum stable value, which is 873 MHz, I got a much neatier 3954 points, which is already higher than your 4650, showing that this tight memory bus is really the limiting factor. In fact, by going haywire on the core at 745 MHz the improvement is much less pronounced, getting to 4334 points.
So, the bottom line is: I did a good thing by going for a 4550 instead of one of those crippled 4650, as it turned out to be a higher performing card when overclocked.
-
- -- Vendor --
- Posts: 372
- Joined: Sun Jul 12, 2009 10:30 am
- Location: Canada
- Contact:
Thanks for testing.
and Yes 4550 is better because it uses DDR3 so memory bandwidth is the most important factor at least for the ATI cards.
Is a bit stupid that they pair powerful GPU with much slower memory so that the GPU is less than 20% or 30% utilization.
My card will get 2800 at 600/450 so just a bit better at the same clock as the 4550 but maybe is also depended on the OS and CPU.
This 4650 that I purchased was at the same price as a 4350 so you get what you pay for. Anyway I do not game so is not important for me I just needed the cheapest card.
and Yes 4550 is better because it uses DDR3 so memory bandwidth is the most important factor at least for the ATI cards.
Is a bit stupid that they pair powerful GPU with much slower memory so that the GPU is less than 20% or 30% utilization.
My card will get 2800 at 600/450 so just a bit better at the same clock as the 4550 but maybe is also depended on the OS and CPU.
This 4650 that I purchased was at the same price as a 4350 so you get what you pay for. Anyway I do not game so is not important for me I just needed the cheapest card.