Palit 9600GT 1GB Sonic and HD3850 Super+1GB

Mike D. 18 Sep, 2008 at 4:50pm ET Article published in Tech

Palit Radeon HD3850 Super+1GB

Below are a list of specifications taken from the 3850 Super+ 1GB product page available at www.palit.biz.

Bus interface: PCI Express® 2.0
Memory Support: 1024MB DDR2
Memory Interface: 256 bit
RAMDACs: 400 MHz
Full Microsoft® DirectX® 10.1 Shader Model 4.1 support
320 stream processing units
128-bit floating point precision for all operations
OpenGL 2.0 support
Dynamic Geometry Acceleration
Superscalar unified shader architecture
ATI CrossFireX™ Multi-GPU Technology
ATI PowerPlay™
ATI Avivo™ HD video and display technology
Dual-link DVI output supports 2560×1600 resolution display
Native HDMI
Built for Microsoft® Windows Vista™

Aside from the usual HD3850 features, we see that Palit decided to use four times more onboard memory than the ATI reference HD3850. Most of the HD3850s shipped with only 256MB of GDDR3, and a few with 512MB. One interesting point that is easily overlooked, however, is that the Super+1GB utilizes DDR2, not GDDR3. This will likely reduce the card’s overall memory bandwidth quite a bit, so it will be interesting to see just what sort of difference this will make.

The HD3850 Super+1GB comes packaged in a colorful box. Red for ATI and Green for Nvidia I’ve noticed. Frobot dominates the front once again, as well as some key specifications.

The Super+1GB is really light when it comes to accessories. Nothing but a molex to PCI-E power adapter, a manual and a driver CD is included. I was surprised to see that a crossfire bridge was not included.

The Super+1GB is not your average HD3850; not only is the cooler non-standard, but neither is the PCB and component layout.

The Super+1GB comes with a single DVI, DSUB and HDMI connector. It would have been nice to see the DSUB connector replaced with a second DVI port, but the HDMI connection is definitely a nice feature.

Palit ditched the reference single slot cooler for an ‘orb’ style heatsink. The centrally mounted fan and open fins should allow the memory and PCB to get a fair bit of collateral cooling.

ATI’s HD3850 is well known for its power efficiency. Only a single 6-pin PCI-E power lead is required.

Unlike the reference HD3850 design, the PWM components are to the left of the GPU. I’m not sure why this design consideration was made, but the Super+1GB is about an inch shorter than the HD3850 reference; definitely a good thing.

Two crossfire connectors are present on the Super+1GB, allowing it to be used in a Crossfire-X configuration. Palit does not include a crossfire bridge. Buyers interested in crossfire will have to either buy one separately, or hope that their second card comes with one.

Above is a shot of the DDR2 used: Qimonda HYB18T512161BF-25. After taking a quick look online, it appears to be rated for 400/800MHz operation. Everything you could ever possibly want to know about these ICs can be found in Qimonda’s product document.

There is memory present on both sides of the card. A sticker is present over top of one of the HSF screws stating that the warranty is void if removed.

I was always amazed by the number of tiny components at the rear of the GPU. The reference HD3850 looks like this as well.

The reference card is about an inch longer than the Super+1GB. There are quite a few layout differences as well. The reference design also includes a small passive heatsink for the VRM components. The Super+1GB shouldn’t need a VRM sink as the components will receive airflow thanks to the orb-style heatsink.

GPU-Z correctly reports the DDR2 memory frequency of 400MHz and calculates a bandwidth figure of 25.3GB/s. The Reference 3850 with GDDR3 has a bandwidth rating of around 53GB/s, almost twice that of the Super+1GB.

« Previous Next page »

Pages: 1 2 3 4 5 6 7 8

Comments

19 Sep 2008 ~ 12:22am Winfrey Very good writeup as always mike! Before I read this article I didn't know very much at all about Palit in the video card industry. I got a pretty good impression from this article though:thumbup
19 Sep 2008 ~ 8:26am Zuntar Nice review Mike, easy to see charts too! Thanks!!!
21 Sep 2008 ~ 2:31am zero-counter Would it have made more sense to compare the 9600gt 1GB DDR3 version to the 9600gt 512MB DDR3 version, in proving differences in memory size?

The current comparison seems biased..considering memory type differences between the two 1 GB cards and the other card having a smaller memory size yet comparable memory type. These specifications can definitely impact the outcome of a benchmark comparison. On that note, your graphs showed the 3850/256MB/DDR3 as a very close contender for the 9600GT, given its memory deficiency.

It is obvious in your benchmarks, that the memory type made a difference when comparing the 3850 ref. 256MB/DDR3 model to the 3850 1GB/DDR2 one. Why not at least the 512MB version of the 3850 if you were not able to procure a 1GB one (only current one I have found is the CF version)? I am aware that the tests are limited to what is on hand and available to you.

Otherwise, the review was informative when comparing your charted info to other reviews out there to the 512MB version of the card in discussion (9600gt).

Please correct me if my logic is flawed. Thanks!
21 Sep 2008 ~ 11:22pm BuddyJ

zero.counter wrote:

Would it have made more sense to compare the 9600gt 1GB DDR3 version to the 9600gt 512MB DDR3 version, in proving differences in memory size?

Sure, it would have been great. Send us one.

The current comparison seems biased..considering memory type differences between the two 1 GB cards and the other card having a smaller memory size yet comparable memory type. bla bla bla

I don't think bias enters the picture. It's an apples to apples comparison for the ATI cards, with an orange thrown in because we were graciously provided with it, plus it gave readers a sense of what sort of performance they can get from a similarly priced card. Both Palit cards are right around $130.

I am aware that the tests are limited to what is on hand and available to you.

That should alone should have answered your original question.
22 Sep 2008 ~ 2:10am zero-counter

Buddy J wrote:

Sure, it would have been great. Send us one.

You may see the point I am attempting to push. If you want to be unbiased, yet informative, then you should procure samples of similar qualities for testing. If someone wants to pay shipping both ways, I could possibly loan a couple of things in the best interests of the site, let me know.

Buddy J wrote:

I don't think bias enters the picture. It's an apples to apples comparison for the ATI cards, with an orange thrown in because we were graciously provided with it, plus it gave readers a sense of what sort of performance they can get from a similarly priced card. Both Palit cards are right around $130.

Apples to apples as in Fujis to Granny Smiths? Or Fujis to Fujis? The 3850 GPU core is identical, yes. The orange, as you stated, was admittedly the control in the review albeit not related to the 9600gt. But the memory type can impact performance severly, especially when enabling AA and AF with higher resolutions. I understand the price issue, but there is more to consider here.

Buddy J wrote:

That should alone should have answered your original question.

This was obviously understood (as pointed out) and posted so that everyone can see that a lack of resources is not justification for a somewhat biased review. For future reference so as this issue does not repeat itself, reviews might be clear, concise, fair, and unjust (every other aspect of the review was absolutely great, just the odd comparison). I said that in hopes that future reviewers take this into consideration.
22 Sep 2008 ~ 8:06am Thrax I don't think bias means what you think it means.
22 Sep 2008 ~ 9:39am lemonlime You are 100% correct about including the other cards, zero.counter. Including results for a 512MB 9600GT and 512MB GDDR3 HD3850 would have been ideal. Believe me, I was of the same school of thought. I would have included them if I could have. The review definitely lacked the appropriate range of competitive products to provide an accurate picture of their abilities. None the less, I did the best I could with what I had to work with.

I should mention that the 256MB HD3850 was paid for out of my own pocket for this review. It wasn't overly expensive, so I didn't mind. Beyond that, I'm afraid I just can't buy three aging cards. I don't get to keep the two Palit cards, (they are on their way to other Icrontic reviewers for other projects) so this is definitely not something I can justify. Icrontic was not able to pitch in for additional hardware and we had nothing else coming from other manufacturers, so that is all I had.

It is very challenging to produce a solid review without all of the appropriate tools. I will see about adding a paragraph to the introduction of the article to forewarn readers about the lack of comparison cards and the apples to oranges comparison of the 9600GT and HD3850s. I do appreciate the feedback and the offer to lend hardware for review. We'll definitely take you up on that next time
22 Sep 2008 ~ 8:38pm zero-counter

Thrax wrote:

I don't think bias means what you think it means.

Really? Care to elaborate? There are two issues at hand, availability of parts and the decision to run the review...displaying a seemingly biased outcome, given the details. To push the review, knowing what i have stated, could be viewed as being bias. Relatively speaking of course.

Lemonlime...your article is great, and I understand the constraints. I was just bringing up a point that I am pretty sure the normal visitor of Icrontic would understand already, but someone googling the card and arriving at the article may not fully understand the comparison for whatever reason. Just a point. I appreciate your response and look forward to future reviews!

I love this site, and will always view it as a useful, entertaining, and intellectually enveloping place to visit.