2900XT + Review MB Used

[Pepsiennis]

As I read through reviews of the HD 2900XT around the web, I began noticing a pattern. Before I comment on it, I'd just like to remark that this post is for the ATI camp, so it'd be much apperciated if the knuckle dragging, frat-boy mentality, spelling-challenged lot from across the aisle just not post. If you find anything I say noteworthy, please discuss it amongst yourselves over in the Nvidia threads. Thank you for your consideration.
- Next, I was never expecting this card to take on the 8800GTX. Anyone who has followed its development should also not have expected it, so from the outset it seemed aimed squarely at the 8800GTS.
- I don't give a rat's ass about power consumption or heat. With the guys running two 8800GTX's in SLI using enough to power a NY city block jumping and hooting like monkeys atop the issue of the HD 2900XT's power use, the hypocrisy is too much to allow for a discussion.

Now, having said all that, here's the observation:

Most reviewers used the eVGA 680i SLI. The 680i chipset includes a technology called Linkboost, that automatically detects an Nvidia GPU and assigns system resources to boost performance. Now in the Bios there is an option to disable it, but anyone who has ever used an Nvidia chipset knows sometimes it will haul off and start implimenting a feature you had disabled (1T, 2T anyone?) while telling you it is still as you'd set it.

The few reviews that used a neutral (in relation to an Nvidia gpu on an Nvidia chipset, at least) Intel chipset almost every time had the HD 2900XT outperforming the GTS 640mb by sizeable margins, and on a few rare occasions even pulling close to the GTX.

I asked our noble leader Kyle Bennett if using a different, more neutral chipset (like the Intel P975 on the Badaxe used for the power consumption tests) would affect the results Brent Justice's review arrived at. His (Mr. Bennett's) reply was succinct: "No."

There are still troubling issues with this card, but they are not nearly as large as some would have me believe on a process level, and I have every confidence in ATI's driver team. I've already ordered my HD 2900XT. My aim with this post is to hearten those who allowed themselves to get caught up in some rather puerile gloating and taunting going on since yesterday, and to encourage you to pay closer attention to the hardware used in the reviews around the web and see if you notice a pattern of MB chipset and gameplay framerates. Best regards.

 

[Pepsiennis]

So people are still in Stage 1, eh?

So, people still can't read, eh?

 

[Mr. Stryker]

Let us know how it goes when you get your 2900XT.

 

[Arcygenical]

Eh, it only raises the PCIe frequency... Doesn't exactly do much.

The benefits are small at best.

Still interesting though.

 

[mitcity22]

everyone wants to have the latest and greatest gpu......it is only natural that the ppl. w/ 8800 series gpus want to make sure that everyone knows they still have the best gpu out there.

 

[Pepsiennis]

Eh, it only raises the PCIe frequency... Doesn't exactly do much.

The benefits are small at best.

Still interesting though.

I ran Ntune on my Striker Extreme with my X1900XT, and set it to only optimize the CPU. It started with the PCIe frequency anyway. Wouldn't stop when I hit the Cancel button either, took it up to 145%. How many atmosphere's before a beer can begins to crunch? I imagine it's the same with a system bus and throughput, but I may well be mistaken.

everyone wants to have the latest and greatest gpu......it is only natural that the ppl. w/ 8800 series gpus want to make sure that everyone knows they still have the best gpu out there.

Their reaction was childish in its relief. They were genuinly scared, and still unsure and clinging to the [H] review as the ONLY one that matters.

 

[Mr. Stryker]

Why isn't that Striker Extreme PC in your sig? :-p

 

[ivan_gromov]

In all honestly, I have been running on ATI's cards for 4 years now, and being a college student, I do not have a ton of money. So, sticking with only ATI will no longer be an option. It is about getting the best product for the best price at this point, soon will have a time when I will have to replace my videocard and the fact is that it will most likely be Nvidia. ATI has failed to deliver on several occasions now, and i am not about to throw 300 some bones toward a product that is half assed, I am sorry but I expected something that would perform with 8800gts, and it seem to be doing just that, but the driver issues are simply unacceptable.

 

[Pepsiennis]

Why isn't that Striker Extreme PC in your sig? :-p

Because it up and died after four days of running at stock. So, I thought I'd leave Asus alone for a few years and try the Abit I'd been eyeing. A good choice it was, too.

 

[Pepsiennis]

... I expected something that would perform with 8800gts, and it seem to be doing just that, but the driver issues are simply unacceptable.

ATI has a pattern over the last year or more of making big strides with each driver release, and geting them out each month.

 

[JDAdams]

LinkBoost is not present on 680i boards - see http://www.nvidia.com/page/nforce_600i_tech_specs.html.

 

[Pepsiennis]

LinkBoost is not present on 680i boards - see http://www.nvidia.com/page/nforce_600i_tech_specs.html.

What it says is, and I quote:

"**Note: NVIDIA LinkBoost™ technology has been removed as a feature from NVIDIA nForce® 680i SLI"

Linkboost was only recently (in the last month or two) discontinued for the 600 series, and the majority of boards out using 680i will still have the feature capacity engineered into them. I doubt many of the post-linkboost MCP's got through channels, into the boars, to the partner, packaged, to the vendor, then into the reviewer or customer's hands in just two months.

What's more, this .pdf presently on the Nvidia website explains the features of the 600 for Intel series also explains what Linkboost does (page 2):
http://www.nvidia.com/docs/IO/35382/nforce_600_intel_linecard_0307.pdf

"NVIDIA LinkBoost™ Technology Select NVIDIA nForce MCPs automatically increases bandwidth when select NVIDIA® GeForce® graphics cards are detected."

 

[Mr. Stryker]

I noticed "NVIDIA LinkBoost not Detected" at the bottom of my BIOS screen. I was stumped.

 

[Eastcoasthandle]

LinkBoost is not present on 680i boards - see http://www.nvidia.com/page/nforce_600i_tech_specs.html.

LOL, nice try but it says that linkboost was removed, not "never used". Therefore, your use of that term here is really a play on words. Because there is no way to disprove that any review using the 680i didn't use linkboost through updated bios, MB revision, etc.

One question I have:
How & When was it removed? Was it removed through MB revision or bios update? Please provide source.

 

[JDAdams]

Mystery solved with a little searching - the review clearly states they used "the latest BIOS at time of evaluation" . For the eVGA board used LinkBoost support was removed as of BIOS revision P26, released on 14th March - see http://hardforum.com/showthread.php?t=1169329. Therefore LinkBoost was not used in the review.

 

[Eastcoasthandle]

Try asking Nvidia, it's their webpage. I don't know, I've never wanted to buy a Nvidia board before yesterday. All I know is that when looking up specs on the boards that's Nvidia's take on it - as you can see it just says it's been removed, which I read as "it isn't present vs. earlier chipsets which supported it". You'd think the info on their own site would be reliable.

In this case you should understand it's a feature they stop using (recently). That's all that really means. Therefore, those who are current users (who had the MB for a while now) of the MB may still linkboost available. That's why I asked:
-When was it removed (it had to be within the last month or 2)?
-What method was used to remove linkboost? Was it through bios update or did the MB need a revision.

These are solid questions here that cannot be answered in that link to nvidia. Therefore, only validates the OP opinion on why the HD 2900XT performs poorly on the 680i vs the 975 chipset.

 

[Commander Suzdal]

So people are still in Stage 1, eh?

So, people still can't read, eh?

And moving quickly into Stage 2! Glad he's making progress.

 

[Pepsiennis]

LOL, nice try but it says that linkboost was removed, not "never used". Therefore, your use of that term here is really a play on words. Because there is no way to disprove that any review using the 680i didn't use linkboost through updated bios, MB revision, etc.

One question I have:
How & When was it removed? Was it removed through MB revision or bios update? Please provide source.

Removed via BIOS, which suggests the specs are still in the hardware. This archived page from PC Perspective has information on the removal.
http://www.pcper.com/content_home.php?s=3

 

[Pepsiennis]

And moving quickly into Stage 2! Glad he's making progress.

Ah yes, I am indeed progressing. You, on the other hand, have not yet learned to read either. What is becoming of the education system in this country? Tsk, tsk. You don't see me skulking over in the Nvidia forums, correcting spelling, grammar, and suggesting logical organization of a presented argument, do you? Run along now, I'm sure you're late for the green beer kegger.

 

[JDAdams]

See my edited post - it's been unavailable as of the BIOS released on 14/03, which means it wasn't a factor in the review. The 2900XT is just bad, I've accepted that and I've never bought a Nvidia card in my life. I've even avoided their chipsets (except for an nForce 2 Ultra-based Athlon XP board, there weren't really any viable alternatives for those procs). It's time to look at the figures, hang up my red gloves and go buy a couple of GTSs for me.

 

[Eastcoasthandle]

See my edited post - it's been unavailable as of the BIOS released on 14/03...

And the source of this information is from? When I checked the Asus Striker Extreme it wasn't remove until bios update on 4-20-07.

 

[JDAdams]

eVGA - their BIOS release notes are quoted in the thread I linked to

 

[Pepsiennis]

Automatic Linkboost was disabled via bios, but the ability is still there. They didn't respin the whole MCP just to drop Linkboost. The 680i is not without it's undocumented flaws. Killing memory, unexplained reboots. It wouldn't be out of character for it to see a brand new architecture and do odd things. Sodium and water have nothing against each other, but in the right amounts you still get a big 'boom'. I'll say it again, there are questions in my mind concerning using the HD 2900XT's arch competitor's chipset to gauge this card, especially when a more neutral alternative was well at hand for every reviewer on every site. I'm not at all saying the card does not have its demerits, it's the fact that when that neutral alternative was implimented the HD 2900XT performed better, which raises questions as to the accurate level of the card's demerits.

 

[JDAdams]

LinkBoost IS the automatic feature. Manual FSB or PCIe bus overclocking are not branded as LinkBoost and are not specific to Nvidia's GPUs or their boards, I can do both on my A32R-MVP with my X1900s. More to the point, both can be done equally well for either ATi or Nvidia GPUs on any board that supports the features and Brent doesn't mention changing any BIOS settings between tests; there is no "magic" going on here.

 

[Eastcoasthandle]

JDAdams there is no information that I see in your link that provides a date in which linkboost was removed. Anyone reading a review from this prospective would have to know the bios version of that MB to know that it was removed from the bios . Therefore, this doesn't disprove the OP opinion. Also you are going a bit OT here. The issue is that some reviews show higher scores while others show lower scores. There is no "magic" here, there is a reason for this happening.

 

[JDAdams]

I have provided clear dates. The review clearly states that Brent used the latest BIOS available - that would be P28, released on 05/07 and after the March P26 release which turned off LinkBoost. Even if you say [H] were a little slack updating (since the window from the 5th to the date of testing is quite small), the previous release, P27, was on 07/04 and still had no LinkBoost. Are you suggesting that Brent would mean a release prior to P26 (14/03) by "the latest"?

 

[Eastcoasthandle]

I have provided clear dates. The review clearly states that Brent used the latest BIOS available - that would be P28, released on 05/07 and after the March P26 release which turned off LinkBoost. Even if you say [H] were a little slack updating (since the window from the 5th to the date of testing is quite small), the previous release, P27, was on 07/04 and still had no LinkBoost. Are you suggesting that Brent would mean a release prior to P26 (14/03) by "the latest"?

LOL,
There is no documentation to the bios used in most reviews I have come across.

 

[Pepsiennis]

The review clearly states that Brent used the latest BIOS available - that would be P28, released on 05/07 and after the March P26 release which turned off LinkBoost.

I trust you're right, JD. But it's still engineered into the chipset, the removal was software only. And Linkboost is only an example, anyway. My point simply is that a chipset designed with in-built optimizations for one brand of GPU isn't the most neutral platform on which to gather one's data concerning a different brand's GPU. And many reviews reflect this opinion. When another platform that is not optimized for a specific brand of GPU was used, the data gatherd was notably (not drastically, only notably) different. It warrants consideration by many minds, not just one.

 

[Commander Suzdal]

Ah yes, I am indeed progressing. You, on the other hand, have not yet learned to read either. What is becoming of the education system in this country? Tsk, tsk. You don't see me skulking over in the Nvidia forums, correcting spelling, grammar, and suggesting logical organization of a presented argument, do you? Run along now, I'm sure you're late for the green beer kegger.

Pardon me if you would, oh sage progenitor of wisdom, for indulging myself in a harmless scintilla of whimsy. I can perhaps engage you more productively in analysis of the thesis you have put forward, if such might be permitted. Ah, but what is this? It seems that in a public discussion forum, the originator of a thread does not possess autocratic control of either the content or the participants. Therefore, I presume I may proceed regardless of your disposition toward the prospect.

Kyle and Brent have considered the possibility in the past that the motherboard chipsets of competing graphics card vendors might offer hidden advantages to their own products. Their findings upon testing this hypothesis have offered no support to it whatsoever. The reason that [H] (and presumably, other quality review sites) use the 680i chipset is that it has proven to be the highest-performing product available for Core 2 Duo processors.

If some of the sites reviewing the 2900XT are using 975X chipsets, it could well be an indication that they do not maintain the highest standards with regard to the equipment they invest in for testing purposes, a shortcoming that, if true, could reasonably be expected to be reflected in the results of their product testing methodologies as well. After all, as I understand it, even P965 motherboards offer superior performance to those based on the 975 chipset.

It could also be a sign that the sites concerned avoid the 680i chipset because they dislike nVidia and its products. Perhaps they shun SLI and embrace Crossfire. If so, it is not unreasonable to suppose that their biases might be reflected in their review methodologies.

Contrast these concerns with the history and reputation of the staff members of [H]ardOCP, who spare no expense in using and testing the best available equipment, and have demonstrated impartial zeal in the pursuit of the truth regarding every product of interest to their target audience. In light of such considerations, I would submit that if there was cause for concern regarding the accurate investigation of a product, they would act upon it without delay.

Essentially, this is the situation: Your thesis, while not without interest, is hardly the only or even the most likely explanation for the observed phenomena. It is incumbent upon you, as the proponent of the thesis, to carry the burden of proof. Either find a reputable and impartial site that compares performance of the 2900XT on different motheboard chipsets within the context of one review (so as to eliminate as many uncontrolled variables as possible) or, once your card arrives, test it on different motherboard chipsets yourself. Given your lengthy and tortuous rationalizations in favor of the 2900XT so far, I must confess I would struggle to lend credence to the objectivity of your results, but I would attend to them with interest nonetheless.

Cordially, your servant Suzdal.

 

[JDAdams]

LOL,
There is no documentation to the bios used in most reviews I have come across.

I'm talking about the [H] review, which is the most scathing one. If we know the worst review used a BIOS without LinkBoost, this shows that even if LinkBoost was enabled on other reviews it only helped the 2900XT.

The problem with the more general issue around optimised chipsets is that it's not good business sense for a manufacturer to make a board that only works well with certain cards, even if they're their own. Otherwise if ATi had come out with a killer card 680i/650i sales would go way down.

Nevertheless, I would like to see a 2900XT benched on RD600, ATi's own chipset, but I don't expect stellar results. If for some reason such results were great, I still wouldn't buy it since RD600 has no dual 16x Crossfire support and I want two cards in my next rig.

 

[Eastcoasthandle]

Pardon me if you would, oh sage progenitor of wisdom, for indulging myself in a harmless scintilla of whimsy. I can perhaps engage you more productively in analysis of the thesis you have put forward, if such might be permitted. Ah, but what is this? It seems that in a public discussion forum, the originator of a thread does not possess autocratic control of either the content or the participants. Therefore, I presume I may proceed regardless of your disposition toward the prospect.

Kyle and Brent have considered the possibility in the past that the motherboard chipsets of competing graphics card vendors might offer hidden advantages to their own products. Their findings upon testing this hypothesis have offered no support to it whatsoever. The reason that [H] (and presumably, other quality review sites) use the 680i chipset is that it has proven to be the highest-performing product available for Core 2 Duo processors.

If some of the sites reviewing the 2900XT are using 975X chipsets, it could well be an indication that they do not maintain the highest standards with regard to the equipment they invest in for testing purposes, a shortcoming that, if true, could reasonably be expected to be reflected in the results of their product testing methodologies as well. After all, as I understand it, even P965 motherboards offer superior performance to those based on the 975 chipset.

It could also be a sign that the sites concerned avoid the 680i chipset because they dislike nVidia and its products. Perhaps they shun SLI and embrace Crossfire. If so, it is not unreasonable to suppose that their biases might be reflected in their review methodologies.

Contrast these concerns with the history and reputation of the staff members of [H]ardOCP, who spare no expense in using and testing the best available equipment, and have demonstrated impartial zeal in the pursuit of the truth regarding every product of interest to their target audience. In light of such considerations, I would submit that if there was cause for concern regarding the accurate investigation of a product, they would act upon it without delay.

Essentially, this is the situation: Your thesis, while not without interest, is hardly the only or even the most likely explanation for the observed phenomena. It is incumbent upon you, as the proponent of the thesis, to carry the burden of proof. Either find a reputable and impartial site that compares performance of the 2900XT on different motheboard chipsets within the context of one review (so as to eliminate as many uncontrolled variables as possible) or, once your card arrives, test it on different motherboard chipsets yourself. Given your lengthy and tortuous rationalizations in favor of the 2900XT so far, I must confess I would struggle to lend credence to the objectivity of your results, but I would attend to them with interest nonetheless.

Cordially, your servant Suzdal.

Clock for clock performance
P5W 975 vs Striker Extreme Test 1
P5W DH 975 vs Striker Extreme 680i Test 2

P5W vs Commando vs eVGA 680i Part 1
P5W vs Commando vs eVGA 680i Part 2
P5W vs Commando vs eVGA 680i Part 3
P5W vs Commando vs eVGA 680i Part 4
Looks like the P5W DH, using the 975 chipset is doing a bit better then the 680i So it makes sense why people are looking for the R600 review on this MB (for example).

 

[Eastcoasthandle]

I'm talking about the [H] review, which is the most scathing one. If we know the worst review used a BIOS without LinkBoost, this shows that even if LinkBoost was enabled on other reviews it only helped the 2900XT.

The problem with the more general issue around optimised chipsets is that it's not good business sense for a manufacturer to make a board that only works well with certain cards, even if they're their own. Otherwise if ATi had come out with a killer card 680i/650i sales would go way down.

Nevertheless, I would like to see a 2900XT benched on RD600, ATi's own chipset, but I don't expect stellar results. If for some reason such results were great, I still wouldn't buy it since RD600 has no dual 16x Crossfire support and I want two cards in my next rig.

This is OT and not focusing on the issue at hand. Truth of the matter remains the same. I have found no documentation that provided which bios was used on some of these reviews.

 

[Dan_D]

As I read through reviews of the HD 2900XT around the web, I began noticing a pattern. Before I comment on it, I'd just like to remark that this post is for the ATI camp, so it'd be much apperciated if the knuckle dragging, frat-boy mentality, spelling-challenged lot from across the aisle just not post. If you find anything I say noteworthy, please discuss it amongst yourselves over in the Nvidia threads. Thank you for your consideration.
- Next, I was never expecting this card to take on the 8800GTX. Anyone who has followed its development should also not have expected it, so from the outset it seemed aimed squarely at the 8800GTS.
- I don't give a rat's ass about power consumption or heat. With the guys running two 8800GTX's in SLI using enough to power a NY city block jumping and hooting like monkeys atop the issue of the HD 2900XT's power use, the hypocrisy is too much to allow for a discussion.

Now, having said all that, here's the observation:

Before I get started let me begin by saying I own both NVIDIA and ATI cards currently and have had plenty of each in the past. There are things I like and dislike about products from both sides. This post is in no way meant to pick on ATI fans. It is however an examination of the comments above, and the facts relating to the G80 cards and the R600 based card in question.

First off, why wouldn't you or anyone else have expected the R600 to take on the 8800GTX or Ultra? This is the only time ATI hasn't released a comparable card in the same product pice points. Why would they not have gone after that market again? The truth is that the card didn't meet expectations and ATI isn't going after that product segment because the product that was meant to do so has failed to meet that goal. So ATI is selling what they have is better than selling nothing at all. Especially when R&D costs need to be recouped. Furthermore, after all the speculation on the web, how could one NOT expect the R600 to be at least an even match with the 8800GTX?

The card has 320 Unified Streaming Processors compared to 128 of the G80. Granted at the time we didn't know that ATI counts them differently than NVIDIA does. R600 was rumored to have 1GB of GDDR3 or GDDR4 ram operating at higher frequencies than that of G80 cards. So more ram, higher frequencies and if the memory were GDDR4 it would use less power. Doesn't this sound better to you? Higher clock core frequencies than G80. Well that wasn't a surprise. ATI has always done that since the introduction of the X800Pro. It doesn't mean better performance necessarily so I'll grant you the fact that it was no indication of superior performance. So what about the rumors of R600's development led you to believe it wouldn't kick the crap out of G80?

Most reviewers used the eVGA 680i SLI. The 680i chipset includes a technology called Linkboost, that automatically detects an Nvidia GPU and assigns system resources to boost performance. Now in the Bios there is an option to disable it, but anyone who has ever used an Nvidia chipset knows sometimes it will haul off and start implimenting a feature you had disabled (1T, 2T anyone?) while telling you it is still as you'd set it.

No. Linkboost causes nothing but problems and has officially been removed as a feature from all 680i motherboards in later BIOS updates acording to NVIDIA's own website. It has also been shown many times to do next to nothing. It certainly wouldn't sway the results beyond 3%, which is an acceptable margin of error by most testing standards. You will see up to 3% variance in testing the same hardware multiple times. As far as the 1T/2T thing that is a problem with some motherboard BIOS' not being coded correctly or some other implementation failure on the part of the motherboard manufacturer. EVGA boards do not do this but the ASUS Striker Extreme does.

The few reviews that used a neutral (in relation to an Nvidia gpu on an Nvidia chipset, at least) Intel chipset almost every time had the HD 2900XT outperforming the GTS 640mb by sizeable margins, and on a few rare occasions even pulling close to the GTX.

I asked our noble leader Kyle Bennett if using a different, more neutral chipset (like the Intel P975 on the Badaxe used for the power consumption tests) would affect the results Brent Justice's review arrived at. His (Mr. Bennett's) reply was succinct: "No."

That's because using a different motherboard wouldn't help and he knows it. The differences in the outcome of some of the data has more to do with the testing methodology than the motherboards themselves. Kyle benches enough boards (and I go over this data when I write a review) that both he and I are keenly aware that the performance difference between motherboards of different chipsets is next to nothing. PCIe 1.0a compliant motherboards will be close enough in performance to keep the test results within that acceptable margin for error.

Look at the benchmarks in this review:

http://enthusiast.hardocp.com/article.html?art=MTMyNSw2LCxoZW50aHVzaWFzdA==

There are times when the Intel boards are faster across the board than their NVIDIA counterparts with NVIDIA graphics hardware installed. If you look at the data you can see that some games perform better on one motherboard chipset vs another even though they are all using the same graphics card. Why do you suppose that is? Differences in the memory controller in the case of the Intel processor compatible boards I'd suspect, and in reference to AMD boards there is no difference as the memory controller is part of the CPU. The fact is some games run better on one chipset vs. another. There is no way that a i975x chipset based board would be more neutral than another.

Data is data and differences in the games tested will lead a reviewer to reach different conclusions based on the data they gathered. Anandtech and Tom's Hardware aren't necessarily wrong and neither is the [H]. Look at the games you play and look at OpenGL vs. Direct X based games and think about the technology newer games that haven't come out yet are using and then decide what graphics card to buy based on the data from all available sources. Also look at the testing methodology and decide what type of testing you find more reliable and more credible. The [H] doesn't use canned benchmarks and scripts which use additional CPU time along side the game itself. These guys play the games and record this information as accurately as possible. Let's also bear in mind that if you are just running scripts, you don't get a feel for the game since you aren't playing it. So there is one important quality that ends up missing from the review. Granted that quality is subjective, but you will get that based on the feedback of the reviewer and it will always vary from person to person but I personally find that part of the process has value to the reader.

As for not caring about heat and power usage, you should. Sitting next to a leaf blower is no fun and heat shortens the life of components and extra heat means that it becomes less likely that reliable stock operation will be possible given certain common circumstances. Such as people using smaller cases, (especiall SFF machines) and placing their boxes in places (desks, cabinets) where the machine gets poor ventilation. As far as power usage, I agree that it doesn't make a huge difference for most people but it could very well mean the difference between having to buy a 750Watt PSU for $200 or a 1KW PSU for $350-500 to run it and the rest of your hardware. Another final point on the power issue is that two HD 2900XT's will use more than two 8800GTX's will at full load. With already enormous power requirements, why would I choose a card that is only just as good as the 8800GTS when it uses much more power, makes more noise, generates more heat and costs $70 more?

 

[siegecraft4]

Pardon me if you would, oh sage progenitor of wisdom, for indulging myself in a harmless scintilla of whimsy.

I laughed. This is the most colorful sentence I've seen in a long time.

 

[Pepsiennis]

My compliments to the Commander,

Kyle and Brent have considered the possibility in the past that the motherboard chipsets of competing graphics card vendors might offer hidden advantages to their own products. Their findings upon testing this hypothesis have offered no support to it whatsoever. The reason that [H] (and presumably, other quality review sites) use the 680i chipset is that it has proven to be the highest-performing product available for Core 2 Duo processors.

If that's the case, why didn't Kyle ofer a link to his findings in that regard? As for highest performing, that isn't an accurate statement, as the C2 revision of the P965 chipset has been consistently out-performing the 680i SLI, especially in regards to overclocking, but often in terms of memory performance, if objective numbers are admissable.

If some of the sites reviewing the 2900XT are using 975X chipsets, it could well be an indication that they do not maintain the highest standards with regard to the equipment they invest in for testing purposes, a shortcoming that, if true, could reasonably be expected to be reflected in the results of their product testing methodologies as well. After all, as I understand it, even P965 motherboards offer superior performance to those based on the 975 chipset.

The P965 would have been a neutral alternative, especially as its Crossfire is a mere courtesy and couldn't be claimed to be usable, so no fear of bias there.

It could also be a sign that the sites concerned avoid the 680i chipset because they dislike nVidia and its products. Perhaps they shun SLI and embrace Crossfire. If so, it is not unreasonable to suppose that their biases might be reflected in their review methodologies.

Point granted, but it's a double-edged sword, as it could also be pointed at our beloved [H].

Contrast these concerns with the history and reputation of the staff members of [H]ardOCP, who spare no expense in using and testing the best available equipment, and have demonstrated impartial zeal in the pursuit of the truth regarding every product of interest to their target audience. In light of such considerations, I would submit that if there was cause for concern regarding the accurate investigation of a product, they would act upon it without delay.

Indeed, I am not trying in this thread to suggest any fault, intentional or through oversight, of Brent and Kyle and all the rest of the [H] staff. It is a knee jerk reaction, perhaps, to try to make this subjective when I'm trying to keep it objective, so that no one muddies waters more than necessary. Nor is it a condemnation of their methodology which, while often unnecessarily copious with graphs to support subjective opinions (hell, just tell us what you think of it, guys. We trust your opinion or we wouldn't be reading it), is a sound one to focus on the games that we actually play...

...Your thesis... It is incumbent upon you, as the proponent of the thesis, to carry the burden of proof.

Had I broached this as a thesis, it would indeed be incumbent upon me to provide reasonable data. But as from the beginning of this thread I have suggested it is rather an observation, it is not incumbent upon me after all. All I have tried to do is bring to notice what no one, to my knowledge, has yet mentioned in the hopes of hearing other's opinions on the matter.

Now, as for the high toned manner: when people enter a discussion with comments like "Brand A kiks azz..." and other such inanities, it isn't contributing to the discussion. But when someone as yourself approaches and makes very thoughtful and interesting comments, I'm always happy to chat a topic with you, even if you manage to prove me wrong.

Your Servant,

Pepsi Ennis

 

[Pepsiennis]

Hello Dan, glad to see you here.

First off, why wouldn't you or anyone else have expected the R600 to take on the 8800GTX or Ultra?

I just think it's a mistake to live by habit. It's much more useful to get your information and make your own judgement. Reports suggested almost two months ago that ATI was going to decline to meet the 8800GTX with a *XTX, because of the leakage leading to the heat issue, and looking at the reported specs the last few weeks, it was obviously aimed at the 8800GTS. As you say, they went with what they had, and are trying for a sea race after that.

That's because using a different motherboard wouldn't help and he knows it...

Then an explanation, however brief, would have been most useful. It's odd to ask us to use our own judgement when weighing a review, but when asked about the components used in that review to not share one's experience. And I certainly have a great respect for all of the [H] staff's experience, that's why I enquired.

Kyle benches enough boards (and I go over this data when I write a review) that both he and I are keenly aware that the performance difference between motherboards of different chipsets is next to nothing.

Dan, did you mean to say that? Because if so, why bother when we don't need the information. Because if we don't need the informatin, then you're doing it for someone else. I KNOW that's not the case at all, but I don't think you want to leave that argument open like that.

...heat shortens the life of components and extra heat means that it becomes less likely that reliable stock operation will be possible ...

Component lifetime? Nah, I upgrade too often. But reliability is always a concern... hence my seven fans and full tower case with the side off. As for SFF users, I don't think a large number of them are going to be interested in what is judged to be a full-size enthusist's card, and those that are already face almost the same issue with an 8800 series in an SFF. The hardcore enthusiast is always looking to upgrade, regardless of if he can afford it at the moment or not, and most reviewers agree thePSU requirements stated by ATI are very much on the safe side for normal operation. And a recent price drop by Nvidia in time for the ATI launch just doesn't sit with me, especially as I live in a market where those won't be reflected for another two months (on average here in China).

Your arguments are helpful for the great number of users. And your explanation is very much appreciated! A question, though:

As you say different games perform differently on various chipsets, why not include a sampling of the mainstream offerings in reviews. I realize that's a question better answerd by Brent, but it would be quite useful to those long-time [H] readers who have Brand X motherboard chipset with Brand Y GPU running Game Z. A real world scenario indeed, though I realize it entails a few days more work.

 

[Commander Suzdal]

Pepsi Ennis,

If nothing else, investigation of the idea can only add to the body of knowledge that benefits us all. It would be unfortunate, however, if yet more hardware selection variables were added to an already complex upgrade planning process. Let us hope that performance differences that crop up between motherboards running the same GPU prove to be small enough to allay concern, for the benefit of us all.

When do you anticipate receiving your 2900XT, and do you have a reasonable opportunity to test it on more than one motherboard?

 

[Dan_D]

Hello Dan, glad to see you here.


I just think it's a mistake to live by habit. It's much more useful to get your information and make your own judgement. Reports suggested almost two months ago that ATI was going to decline to meet the 8800GTX with a *XTX, because of the leakage leading to the heat issue, and looking at the reported specs the last few weeks, it was obviously aimed at the 8800GTS. As you say, they went with what they had, and are trying for a sea race after that.

Then an explanation, however brief, would have been most useful. It's odd to ask us to use our own judgement when weighing a review, but when asked about the components used in that review to not share one's experience. And I certainly have a great respect for all of the [H] staff's experience, that's why I enquired.

Dan, did you mean to say that? Because if so, why bother when we don't need the information. Because if we don't need the informatin, then you're doing it for someone else. I KNOW that's not the case at all, but I don't think you want to leave that argument open like that.

Yes I meant to say that and I didn't leave it open. I have already stated the reasons why the motherboard would have made little to no difference. I can't speak for Kyle and I wouldn't presume to do so, but he is aware of the differences between boards on a technical level just as I am. Though he may not have said anything more than "no", he gave the answer for a reason. It is a question I have answered as best as I can given the circumstances.

I am quoting my post above:

"PCIe 1.0a compliant motherboards will be close enough in performance to keep the test results within that acceptable margin for error.

Look at the benchmarks in this review:

http://enthusiast.hardocp.com/articl...50aHVzaWFzdA==

There are times when the Intel boards are faster across the board than their NVIDIA counterparts with NVIDIA graphics hardware installed. If you look at the data you can see that some games perform better on one motherboard chipset vs another even though they are all using the same graphics card. Why do you suppose that is? Differences in the memory controller in the case of the Intel processor compatible boards I'd suspect, and in reference to AMD boards there is no difference as the memory controller is part of the CPU. The fact is some games run better on one chipset vs. another. There is no way that a i975x chipset based board would be more neutral than another. " -Me.
As I have tried to say all along, the so called NVIDIA optimizations are pointless and wouldn't have skewed the test either way. Using an Intel chipset board would have raised the benchmarks scores in some tests and dropped them in others but it would have done so for BOTH cards tested. PCIe 1.0a complaint motherboards will perform virtually identically in any given situation with any card designed to meet PCIe 1.0a specifications. I linked to our own benchmarks that were conducted by Kyle in a review written by me in graphs I made and uploaded myself. That data shows the NVIDIA chipsets out performing the Intel chipsets some times and losing to them at others. In almost 100% of those cases the difference between boards of different chipsets is small. It's generally less than 3%.

Performance is rarely a factor in seperating one board from another. Frequently I've seen $100 perform just as well as a $300 board. The biggest difference continues to be stability and features. In those cases the more expensive boards have something to offer over their simpler lower priced counterparts.

Component lifetime? Nah, I upgrade too often. But reliability is always a concern... hence my seven fans and full tower case with the side off. As for SFF users, I don't think a large number of them are going to be interested in what is judged to be a full-size enthusist's card, and those that are already face almost the same issue with an 8800 series in an SFF. The hardcore enthusiast is always looking to upgrade, regardless of if he can afford it at the moment or not, and most reviewers agree thePSU requirements stated by ATI are very much on the safe side for normal operation. And a recent price drop by Nvidia in time for the ATI launch just doesn't sit with me, especially as I live in a market where those won't be reflected for another two months (on average here in China).

Perhaps I wasn't clear. I appologize. My meaning is this: Extra heat in certain given circumstances will not only reduce the operational lifetime of the product but it will compromize stability and all things being equal, the hotter card in the right machine might be less stable than it would be if it were running the competitions cooler running card. Granted there are many more variables that determine the overall stability of a system and of a graphics card. Many of us here on the [H] upgrade their systems frequently and this isn't a huge issue.

In the last year or so I've had these video cards:

7800GTX 256MB SLI
7900GTX SLI
X1950XTX Crossfire
8800GTX SLI

In my secondary machine I've had:

X1800XT
X1950Pro
X1950Pro Crossfire

I will likely be pulling the Crossfire configuration out of there and installing a single 8800GTS.

Heat production can't always be overcome easily either. Some people don't want to invest in better cooling, they don't want to turn their air conditioning down in order to reduce ambient temperature where the machine is. Their AC unit might not work the best. You might also be trying to use a card like that in a small form factor PC where size, power usage and cooling are all very important factors. In those cases performance is secondary to those other considerations.

Your arguments are helpful for the great number of users. And your explanation is very much appreciated! A question, though:

As you say different games perform differently on various chipsets, why not include a sampling of the mainstream offerings in reviews. I realize that's a question better answerd by Brent, but it would be quite useful to those long-time [H] readers who have Brand X motherboard chipset with Brand Y GPU running Game Z. A real world scenario indeed, though I realize it entails a few days more work.

Again it would accomplish next to nothing. All the tests where benchmarks are conducted in the motherboard reviews are done at 640x480 to take the graphics card out of the equation as much as possible. The same CPU is used as often as possible to isolate the difference between chipsets and boards but the reality of it is that the difference between motherboards when running high end processors and video cards at high resolutions is going to be 0%. This goes back to my point that going with a different motherboard in the context of the video card review would not have changed ANY of the results. When I say that different chipsets perform differently that's true on a technical level and is shown in those tests that are designed to expose those differences. In the real world the choice of board doesn't matter in terms of performance.

 

[phide]

Perhaps this isn't the time nor the place, but I just wanted to comment on a couple of non-technical comments you've made here. Be aware that you opened the floodgates for this, so to speak.

I'd just like to remark that this post is for the ATI camp, so it'd be much apperciated if the knuckle dragging, frat-boy mentality, spelling-challenged lot from across the aisle just not post.

Firstly, I don't believe it is wise to start a thread that is intended to seek technical conclusions to a technical hypothesis by stating that a particular sect of individuals are a "spelling-challenged lot" (of which you may consider calling yourself a member: you've misspelled "appreciated" in the quotation above). You've brought up an interesting question here, and certainly one worth following through, but just asking the question in a mature, civil manner would have been sufficient.

Next, I was never expecting this card to take on the 8800GTX. Anyone who has followed its development should also not have expected it, so from the outset it seemed aimed squarely at the 8800GTS.

I agree, which why I find some comparisons to the GTX somewhat odd. As a competitor to the GTS, I myself believe that the XT is actually quite a decent product -- not quite qualifying as a "flop" by my criteria.

With the guys running two 8800GTX's in SLI using enough to power a NY city block jumping and hooting like monkeys atop the issue of the HD 2900XT's power use, the hypocrisy is too much to allow for a discussion.

I don't believe the hypocrisy exists as you understand it. The issue many of us have with the XT is not so much that it consumes a great deal of power, but that the additional consumption doesn't translate into better performance or better image quality. The efficiency of the architecture, quite simply, seems to be unfathomably poor. If the XT had delivered the performance we were expecting from its 512-bit bus, 320 SPs and high TDP, I don't believe anyone (or at least very few people) would have thought much of the power consumption.

Nobody is enraged at the potential environmental impact of the XT, nor is anyone who has ever looked a power bill up in arms about the power consumption (power is still relatively inexpensive). Our chief concern is simply that the additional power is not translating into something we value, and because we haven't been told as to why the consumption is so much higher, we see it as a glaring flaw of R600.

My aim with this post is to hearten those who allowed themselves to get caught up in some rather puerile gloating and taunting going on since yesterday, and to encourage you to pay closer attention to the hardware used in the reviews around the web and see if you notice a pattern of MB chipset and gameplay framerates.

Is there a chance you could give us links to reviews that did not use 680i boards?

 

[gljvd]

hardware.info used an intel chip

http://www.hardware.info/nl-NL/articles/amdnZGpqZGia/ATI_Radeon_HD_2900_XT_test/1

We hebben de nieuwe Radeon HD 2900 XT getest op een tweetal testsysteem. Allereerst hebben we de kaart blootgesteld aan onze bestaande reeks benchmarks onder Windows XP, gebruikmakend van een testsysteem bestaande uit een Intel Core 2 Extreme X6800, 1 GB DDR2-800 geheugen, een ASUS P5W DH moederbord (Intel 975X chipset) en een Western Digital Raptor 74 GB harddisk.

The tests except 2 tests which I assume are driver bugs , show the 2900xt ahead of the gts cards and sometimes within spiting distance of the gtx . In splinter cell its the fastest card they tested beating the gtx .

The two games with driver problems I believe are as follows

x3 - the radeon hd looses almost half its frame rate with aa and af turned on going from

96.5 fps at 1920x1200 to 56.5 fps The x1650xt looses less than 7fps in the same res jump !

stalker on vista has a driver bug for nvidia . None of the gts cards get past 15fps
while the gtx seems to perform fine with a commanding 86fps while the 2900xt can only put out 46.9 So something is up there