2900XT arrived

[Matt2]

Originally posted by: silentvois
You still having power issues with your cpu overclock and the 2900XT? Perhaps you could try rewiring some of the hardware to the other rail on that PSU. I would think that 1 of the rails which handles the power to the videocard is being overloaded.

Yup, still having the power issues when OC'd to 2.8GHZ. I disconnected the case fans, but still shuts down during 3DMark when operating at 2.8GHZ.

I would agree with you assessment that one of the rails is being overloaded.

It's an odd issue that occurs when I raise the CPU voltage. For example, 2.7GHZ @ 1.425v is ok, but if I raise the CPU voltage to 1.45v and keep the CPU @ 2.7GHZ, the system will shut down.

With the CPU @ 2.7GHZ /1.425v the system is stable and I can even overclock the video card to 775/875 with ease.

 

[Matt2]

Originally posted by: Cookie Monster
Have you fiddled around with the IQ related options? e.g CFAA? would you say the IQ have improved or got worse?

edit - or is practically the same?

Good question. I havent really had time to play a whole lot between benches, installing games, downloading patches, etc...

I plan to take a hard look at the IQ tomorrow before I have to work.

 

[Cookie Monster]

Originally posted by: Matt2

Originally posted by: Cookie Monster
Have you fiddled around with the IQ related options? e.g CFAA? would you say the IQ have improved or got worse?

edit - or is practically the same?

Good question. I havent really had time to play a whole lot between benches, installing games, downloading patches, etc...

I plan to take a hard look at the IQ tomorrow before I have to work.

Its because im the same boat. Im finally in a situation where ill build a rig that will last me 3~4 years. (AS you can see my current rig has been serving me very well for the past 4 years)

2900XT here in NZ is around the same price as majority of the GTS models.

Its either that or volt modded 8600GTS 512mb SLi ftw. (im talking clocks of core/mem 900+/1000+ MHz with simple pencil mods)

Keep up the good work :thumbsup:

 

[MarcVenice]

I really doubt that a oc-ed opty at 2.7-2.8ghz will severely bottleneck a 2900xt in games. It would be nice to see if frame rates really do go up, in a number of games, if it's paired with a c2d at 3ghz or so. And by go up I actually mean if the minimum frame rate goes up, making a game more playable. If the frame rate goes up from 110 to 125 its worthless, and not cpu limited imo.

 

[Keysplayr]

Originally posted by: MarcVenice
I really doubt that a oc-ed opty at 2.7-2.8ghz will severely bottleneck a 2900xt in games. It would be nice to see if frame rates really do go up, in a number of games, if it's paired with a c2d at 3ghz or so. And by go up I actually mean if the minimum frame rate goes up, making a game more playable. If the frame rate goes up from 110 to 125 its worthless, and not cpu limited imo.

An example would be STALKER. The scores you see on my benchmark site for STALKER are from a stock C2D at 2.13GHz. When I overclocked JUST the CPU to 3GHz and left the 8800GTS stock, there was a SUBSTANTIAL gain in min/avg/max framerates and probably could have gotten higher frames with a more significant o/c. Granted, this is just one game, but it's a game Matt is testing on a 2.6GHz opty DC. A C2D at 3GHz pretty much blows the doors off an opty at 2.6/7 GHz. I can pretty much guarantee his framerates will shoot up if he actually made it to 3GHz or more on that opty. At the very least, because he is running at higher res, his minimum framerate would come up quite a bit.

Matt, if you would, try running at a lower res, like 1680x1050 leaving the 2900XT at stock speeds. I'll send you the exact demo I used to bench STALKER (made it myself because the "buildings_timedemo" and "short_timedemos' " are a joke. At least you will have something to compare. PM me and give me a way to send you the demo file.

Thanks,

Keys

 

[MarcVenice]

Yeah Stalker is a CPU intensive like that. That's just one example though, it would still be interesting to see this kind of testing done over several games. In fact I'm starting to think that it might actually be nice to include such information in game reviews.

 

[Keysplayr]

Originally posted by: MarcVenice
Yeah Stalker is a CPU intensive like that. That's just one example though, it would still be interesting to see this kind of testing done over several games. In fact I'm starting to think that it might actually be nice to include such information in game reviews.

I would have done something like this, but time constraints didn't allow it. I have surely gained new found respect for reviewers as it is the norm to pull several all nighters in a row to run benchmarks to the extent they do. I couldn't afford that much time. But I still managed some of the cleanest benchmarks on the web for what I did complete.

 

[MarcVenice]

Heh a proffesional reviewer, like for example the people they have working for anandtech get everything shipped for free, or they buy it with money they get from ad reveneu's and the like. If I were to review videocards, I would try building 2-3 identical rigs, and then run different benchmarks on all 3 at the same time. Saves you a lot of time. In fact, I'm pretty sure this gets done anyways From all the hardware they review they could build several identical rigs.

 

[Keysplayr]

"Heh a proffesional reviewer, like for example the people they have working for anandtech get everything shipped for free, or they buy it with money they get from ad reveneu's and the like. If I were to review videocards, I would try building 2-3 identical rigs, and then run different benchmarks on all 3 at the same time. Saves you a lot of time. In fact, I'm pretty sure this gets done anyways From all the hardware they review they could build several identical rigs."



In bold above: What does this matter? In the end, hardware is hardware no matter where it comes from. I used my own pocket money. I didn't have the luxury of free hardware, but in the end, benchmarks would still have been the same if I actually did get free hardware.


It is still best to use only one rig, as I have found out. Even though the rigs are identical "hardware-wise" there can still be some deviation in benchmark scores for whatever reason. Using only one rig, and a master hard drive in which to image from, is the cleanest zero variable way to conduct benchmarks. I initially intended to use two identical rigs for my benches, but after discussions with some reviewers, it was deemed that the only way to remove all variables was to use a single rig, leaving only the graphics card the singular difference.

Reason: This eliminates the hundred questions about whether bios versions on the motherboards were the same. Or chipset versions were the same stepping. Believe it or not, as I am sure you have seen, if someone doesn't like bench results, he/she will look to these tiny deviations and call the WHOLE review into question. Ridiculous, I know, but it is commonly done. A single rig reduces the target in which to attack the reviews integrity.

 

[MarcVenice]

The thing in bold doesn't matter, it just means professional reviewers CAN build several rigs. And really, if a reviewing site simply doesn't tell about using 2 or more rigs, then who is to know? If said reviewing site builds 2 identical rigs, and runs benchmarks at stock settings, and over several benchmarks they difference is within 1% or even identical, then there is nothing wrong about using identical rigs to speed things up. I haven't tried yet, but I really doubt that a computer build with the same components, cpu batch #'s only varying 1 single digit, etc etc will vary much if at all in benchmarking results.

Using 1 rig will of course be the cleanest way, no doubt about it, but what I'm saying is that I doubt that using two identical rigs will be any less cleaner.

 

[apoppin]

Originally posted by: MarcVenice
Heh a proffesional reviewer, like for example the people they have working for anandtech get everything shipped for free, or they buy it with money they get from ad reveneu's and the like. If I were to review videocards, I would try building 2-3 identical rigs, and then run different benchmarks on all 3 at the same time. Saves you a lot of time. In fact, I'm pretty sure this gets done anyways From all the hardware they review they could build several identical rigs.

i prefer to run everything on the same rig

it is really easy if you have multiple partitions and also have everything "ghosted" except for the video drivers

also ... i don't mind doing complete fresh installs of Vista ... it is THAT quick!
... insert the DVD .. check back in a few minutes ... wait a bit more .. done!
-and don't forget to defrag

finally ... now that Best Buy is [finally] so cool about returns ... IF i ever have a HARD decision like 2900xt vs. GTS [again] ... i might also buy two competing cards, test them - and return the 'loser'
-that kinda equals out the "free" HW the "professional" reviewers get

 

[Keysplayr]

You're absolutely right. 1% really doesn't matter. But the point is not about that 1%. It's about people. The less subject matter to question, the better. The extent some video card enthusiasts will go through when their fav video card gets beaten even by a hair could be called legendary. There's a decent chance those particular folks don't put anywhere near as much diligence into their jobs/careers. hehe.

I'd just as soon remove any ammunition that can be used to "even possibly" cast doubt on the benchmarks.

 

[MarcVenice]

On a sidenote keyz, wouldn't it be in fact even better to build 3 identical rigs? You've said it yourself, not all rigs will be the same, even when identical hardware is used. Isn't this the case for consumers too ? Perhaps reviewers should review videocards in 3 different rigs, using the same exact benchmarks, and take the averages of those 3 rigs. Because if you keep by what you said, and rigs can't be identical then that would actually be the way to go ^^

 

[apoppin]

Originally posted by: MarcVenice
On a sidenote keyz, wouldn't it be in fact even better to build 3 identical rigs? You've said it yourself, not all rigs will be the same, even when identical hardware is used. Isn't this the case for consumers too ? Perhaps reviewers should review videocards in 3 different rigs, using the same exact benchmarks, and take the averages of those 3 rigs. Because if you keep by what you said, and rigs can't be identical then that would actually be the way to go ^^

i don't see any difference ... three identical rigs is a waste
--and why stop with just three - why not ten identical rigs?

*one* rig is perfect ... and it doesn't take any longer ... nor is it any less "accurate" [it takes less time and is more accurate with one, imo]

your 3 "identical" rigs might turn out to be not so identical and you have introduced even more variables and time-wasting .. then you have to reconfigure all three rigs every time you change something
--stick to reviewing games, Marc


[btw, where can i read them ? in English?]

 

[MarcVenice]

You can't, I'm Dutch remember, I write for a Dutch gaming site Maybe in the future when I work for some big english reviewing website

Btw, you're not getting my point. You are contradicting yourself you know. If three identical rigs aren't in fact identical, then reviewing hardware in different, yet identical rigs might add extra value, since your and my rig might be identical, yet not perform identical, for whatever reason. Thus it would be interesting to see if this is in fact the case, and if it is, taking the average of those 3 rigs would add extra meaning to the review.

In these 3 rigs, for example, you would have to swap out the same thing everytime you want to review something. The videocard, the cpu, etc etc. The only problem I can think of, is that a system might detoriate over time, like the PSU for example, making 1 rig less stable, or perhaps a oc-ed CPU, that becomes unstable.

And if the 3 rigs are identical, and perform the same in 3dmark06 and other games, then it will take less time to run all the benchmarks, you could have 3 time demo's running at the same time.

 

[Keysplayr]

Nope. Actually you would be tripling your benchmarking workload.

Example:

3 rigs. All identical in hardware and software except for the video card and it's driver.

In the first rig is a 8800GTS. The second has a 2900XT and the third has a X1950XTX crossfire (just for example purposes).

Run all your benchmarks. Then rotate the cards and the hard drives for your second run. Then rotate again for the third run. Why three runs? Because you have saddled yourself with three computers that could possibly yield different results. Why would you do this? To check your own benches. Is this a little extreme? Sure is, but would have to be done to get a TRUE average of benchmarks that might not even mean anything.

One rig is best. Anything more just opens up the floor to naysayers. You can't get simpler or cleaner results. And it's less work.

 

[apoppin]

Originally posted by: MarcVenice
You can't, I'm Dutch remember, I write for a Dutch gaming site Maybe in the future when I work for some big english reviewing website

Btw, you're not getting my point. You are contradicting yourself you know. If three identical rigs aren't in fact identical, then reviewing hardware in different, yet identical rigs might add extra value, since your and my rig might be identical, yet not perform identical, for whatever reason. Thus it would be interesting to see if this is in fact the case, and if it is, taking the average of those 3 rigs would add extra meaning to the review.

In these 3 rigs, for example, you would have to swap out the same thing everytime you want to review something. The videocard, the cpu, etc etc. The only problem I can think of, is that a system might detoriate over time, like the PSU for example, making 1 rig less stable, or perhaps a oc-ed CPU, that becomes unstable.

And if the 3 rigs are identical, and perform the same in 3dmark06 and other games, then it will take less time to run all the benchmarks, you could have 3 time demo's running at the same time.

how could i forget? Some Dutch publications are also in English
[pm me the link, anyway]

you haven't tried this have you?

what you are suggesting is a LOT more work ...
-for nothing ... no gain whatsoever ... no advantage ...

it would be like playing STALKER game in Russian, Dutch and English and then writing three "different" reviews and "averaging" them in Esperanto

 

[MarcVenice]

You too contradict yourself keys. You are giving me a nice review from lets say a videocard. Nice, but it has only been tested in one rig. You're saying you can't use 3 identical rigs, and swap out only the videocards, yet you say we should take those reviews done in 1 rig for granted and they are the best, even though everyone that could have any possible use of the review also have different rigs, even if they are identical in hardware.

So you can in fact use 3 identical rigs, and the results will be as clean as using 1 rig. Why, well, because you are saying that the reviews being done in 1 rig are the best possible and will apply to all different kinds of computers out there. Those different kind of computers resemble the 3 rigs your using though, to speed things up. If anything, this should even give the naysayers LESS to complain about.

 

[apoppin]

Originally posted by: MarcVenice
You too contradict yourself keys. You are giving me a nice review from lets say a videocard. Nice, but it has only been tested in one rig. You're saying you can't use 3 identical rigs, and swap out only the videocards, yet you say we should take those reviews done in 1 rig for granted and they are the best, even though everyone that could have any possible use of the review also have different rigs, even if they are identical in hardware.

So you can in fact use 3 identical rigs, and the results will be as clean as using 1 rig. Why, well, because you are saying that the reviews being done in 1 rig are the best possible and will apply to all different kinds of computers out there. Those different kind of computers resemble the 3 rigs your using though, to speed things up. If anything, this should even give the naysayers LESS to complain about.

you STILL don't get it

it is THREE TIMES the work - for absolutely NOTHING
-three "identical" results if the HW and setup is indeed identical

ALL of our "in-house" testing backs each other's benches up ... Keys, myself, Stumps, n7, matt2 ... we are all using similar but different builds and we all get similar results

i'd MUCH rather see 20 DIFFERENT rigs tested then 10 of the "same" build

why don't i just completely take my rig apart for you and reassemble it three times just to satisfy you with 3 sets of benches?





-and i want to see reviews from you in TWO languages from now on

 

[MarcVenice]

No you still don't get it !!! I don't want three different reviews, using 3 different rigs. Just use 3 frigging rigs, and plug a different card in each, as to speed up the benchmarks, it will cost you 1/3rd of the time That's my whole point.

 

[apoppin]

Originally posted by: MarcVenice
No you still don't get it !!! I don't want three different reviews, using 3 different rigs. Just use 3 frigging rigs, and plug a different card in each, as to speed up the benchmarks, it will cost you 1/3rd of the time That's my whole point.

i have more time than money

as it stands, i simply bench my current card, remove it and install another Card - or CPU, or memory, or a HSF or whatever] - into a clean OS on another partition [or HD] and do it again ... the final results comparison are truly on an "identical" machine and as close to perfection as is humanly possible
... any more "perfect" and it could be wrong

--get it now ? .. time vs. money ?

... and i have an awesome idea:
-:light:-
*you* do it on 3 rigs

 

[Matt2]

IF someone will send me some cash I'll do a review on 3 rigs.

I'm gonna go try out some CS:S with the different CFAA modes. I thought CFAA went up to 24x? I can only get it to go to 16x via wide-tent?

 

[Keysplayr]

Originally posted by: MarcVenice
You too contradict yourself keys. You are giving me a nice review from lets say a videocard. Nice, but it has only been tested in one rig. You're saying you can't use 3 identical rigs, and swap out only the videocards, yet you say we should take those reviews done in 1 rig for granted and they are the best, even though everyone that could have any possible use of the review also have different rigs, even if they are identical in hardware.

So you can in fact use 3 identical rigs, and the results will be as clean as using 1 rig. Why, well, because you are saying that the reviews being done in 1 rig are the best possible and will apply to all different kinds of computers out there. Those different kind of computers resemble the 3 rigs your using though, to speed things up. If anything, this should even give the naysayers LESS to complain about.

:sigh:

1 rig. 2 video cards. zero variables. I can't make this any simpler. We will now agree to disagree, yes? Cause it's going nowhere for no good reason.

Thanks.

 

[MarcVenice]

I know, at first sight it is perfect, thinking scientificaly and all that. But since the reviewed videocard will end up in all kinds of different pc's there should be no problem in using 3 rigs to speed things up

 

[lopri]

Topic went off-topic. :laugh:

Well, I'd think a site the size of AT has a lab with dozens of systems running automated(scripted) benchmarks for hours. You don't think someone's sitting in front of a monitor and mouse-clicking every few minutes, do you? Of course it'll depend on the parts being tested. Only when there is an outlier a person will actually try to figure out what's going on with the system.

On topic (?): What is the most reliable brand for 2900XT? Of course warranty and CS wise.