[Review] Say no to “single” GeForce 9800GTX!
In NVIDIA’s name scheme, GTX means “GeForce Technology eXtreme”, performs better than a GT card. April 1st, NVIDIA rolls our their 9800GTX, a 9 series “GeForce Technology eXtreme”. Is it really a extreme card which can it feed gamers rising graphic demand? or only a April Fool’s Joke?
Page 1: Introduction
Page 2: 9800GTX: born for 3-way SLI
Page 3: Hybrid Power Supported
Page 4: 9800GTX is huge
Page 5: a closer look at 9800GTX
Page 6: 9800GTX’s cooling
Page 7: 9800GTX video playback test
Page 8: GeForce 9800GTX power consumption test
Page 9: 9800GTX overclock test
Page 10: Test Bed and remark
Page 11: 9800GTX vs 8800GTS 512
Page 12: 9800GTX vs 8800GTS 512 OC
Page 13: 9800GTX vs Radeon HD 3870X2
Page 14: Say no to single 9800GTX
April 4th, 2008 at 1:39 am
Thank You ! That is exactly what I’ve been asking for in a post embedded to previous article. Your work is great and noticed
April 4th, 2008 at 10:48 pm
[...] 9800GTX early tests Couple of new articles out this morning on the subject: [Review] Say no to GeForce 9800GTX! G80 VS G92: Med to high end GeForce GFX cards round-up! That last article above is what I have [...]
April 4th, 2008 at 11:59 pm
FACKERSSS !!! THOSE BENCHMARKS ARE FAKE !!!!
FUCKERSS
April 5th, 2008 at 12:22 am
[...] na rynku. Niestety w przypadku modelu 9800GTX, okazauje się to być dalekie od prawdy. Według testów przeprowadzonych przez Expreview 9800GTX jest szybszy od referencyjnej wersji 8800GTS zaledwie o kilka procent, natomiast od wersji [...]
April 5th, 2008 at 7:00 am
Silly thing goes of benchmark. Totally fake.
April 5th, 2008 at 8:47 am
Wots wrong with the benchmarks?
April 5th, 2008 at 1:10 pm
What the fuck? Other websites are showing the 9800GTX is fucking great. Sure you guys didn’t do something wrong?
http://www.tomshardware.com/2008/04/01/nvidia_geforce_9800gtx_review/page5.html
April 5th, 2008 at 1:45 pm
Maybe most of the readers are misled…
1stly LOOK AT THEIR DRIVERS-
Expreview -
Radeon 3870X2 - Catalyst 8.3 WITH HOTFIX
9800 GTX - Forceware 174.74 beta
Tomshardware -
Radeon 3870X2 - Catalyst 8.3
9800 GTX - Forceware 174.53
If you can ACTUALLY READ, then you realise that the 9800 GTX is doing better in Expreviews benchmark. There is misunderstanding, since Radeon 3870X2’s new hotfix has improved performance so there IT IS FASTER THAN 9800 GTS USING THE HOTFIX. Tomshardware and MOST OF THE OTHER SITES DID NOT USE THE HOTFIX…. so thats why it seems that the 3870X2 is getting pwned, when as an actual fact, if people are using the LATEST DRIVERS, then it will perform magnificently.
Hats of to Expreview for another great review, 9800 GTX is indeed A WASTE OF MONEY.
and NO lol i am (or was) a NVIDIA fanboy until they released the SH*T that was the 9000 series… ive never used ATI before… but maybe people should actually think and see that AMD/ATI are coming back and NVIDIA are going down for once.
April 5th, 2008 at 10:18 pm
The benchmarks are totally fabricated. For one, the FEAR bench is about 1/3 of what the scores would actually be on even the GTS (I averaged 249 fps at the same res with EVERY SINGLE OPTION MAXED) on one GTS and a Q6600 @ 3.7, card clocks same as theirs). I ran the same tests last night, and can say from these numbers that they didn’t run ANY benches. Also, if you look at the OC Test page, it shows the 9800GTX 3DMark score (running on a 3ghz Q6700, NOT a QX9650) scoring almost 15k, which is outstanding. The scores are bullshit, the benches were never even run, and I doubt this “testbed” they show even exists. BULLSHIT HAS BEEN CALLED.
April 5th, 2008 at 10:25 pm
Oh, and, why in their GPU-Z screenshot of the GTS does it show that the “Nvidia Reference” GTS they are using (650mhz core) have default clocks that match the “Overclocked” settings? It is because they photoshopped a screenie of the 9800GTX GPU-Z readout to make it look like a GTS. Look closely at these scores and screenies before you go thanking them for doing this. They are full of shit. Every single score on here is incredibly low, too low for them to have even run these tests, since they aren’t in the same BALLPARK as actual scores from their described “testbed”. Next time you try to BS people, make sure you don’t… (A) Contain scores so low that people know you are lying, (B) Show that you blatantly photoshopped the result pages by being sloppy and forgetting to change the default clock speeds on your “overclocks”, (C) Leave in pics of your REAL test machine (3ghz Q6700, not the QX9650 they show on the testbed page) in the 3DMark menu in the background, and (D) Give scores that are INCREDIBLY low for either card. You have actually convinced me that the 9800GTX is a GREAT card, since it can hit 14.6k on a 3ghz Q6700, something that takes me almost 3.6-3.7ghz to do on a GTS 512. Back to the lab again, idiots…
April 5th, 2008 at 11:56 pm
to spAde:
about GPUZ:
we use NiBiTor change the 8800GTS 512 frequency to 675/2200(Overclocking) ,and nvflash the VGA BIOS.
about 9800GTX OC:
we use another platform for OC ,Q6700 OC 3GHz and test VGA Frequency extreme.sorry for the article didn’t show the “special” platform, it’s editor’s foult.
normal platform can found out in the testbed page.
April 6th, 2008 at 12:28 am
lmao 512 bit
April 6th, 2008 at 2:50 am
to luckyboy and kensoi:
What we reviewed is seriously.We never and never make fake report in expreview.
April 6th, 2008 at 5:34 am
You are full of shit.
April 6th, 2008 at 8:14 am
I’ve read a ton of your reviews, all BS. In one review, the 3870X2 scores 20k in 3DMark, when the review is intended to make an Nvidia card look bad, but then in another, it scores only 15k. Sorry guys, you aren’t clever. Maybe when you hit the big time you’ll have more than one email address for your site, and it won’t be a Gmail.
April 6th, 2008 at 11:10 am
Funny that you say you use a BIOS flash to overclock, when the page clearly states that you used RivaTuner 2.07. So, to OC the card by 150mhz/280mhz on the core/vram, you used RivaTuner, but for a tiny 5mhz OC, you edited and reflashed the BIOS?
April 6th, 2008 at 2:20 pm
@spAde:
this is the first NVIDIA product that Expreview suggests that we DONT buy… and im sure every one else agrees, its a WASTE of money. Really, its obvious that they recommend NVIDIA over ATI, since ATI has failed to deliver so far. The info they come up with is mostly accurate, ie they were the first to announce the existance of 8800gs, before any other site.
However for the benchmarks, you raised a good point…
WHY DOES A GRAPHICS CARDS PERFORMANCE FLUCTUATE IN MOST BENCHMARKS YOU HAVE DONE, EVEN IF SAME DRIVERS ARE USED?
April 7th, 2008 at 12:36 am
It isn’t accurate AT ALL. Their numbers change from review to review, for one. For two, if the GTX sucks so bad, how does it hit almost 15k in 3dMark06 IN VISTA ON A 3GHZ QUAD. My Q6600/GTS takes about 3.6-3.7 to do that. Dude, their “bench” results all depend on what they’re tryin to tell you to buy. You think that a machine that gets only 80fps in FEAR is getting 125 in UT3? Oh, btw, MY quad at 3.7ghz (instead of their 4.0) and GTS at their OC gets avg 255fps in FEAR, not 80. If you look on another page, they show the GTS in SLI, with 1ghz slower CPU getting 150fps. So, adding another card, and then completely strangling it with a huge drop in CPU speed will get you over double performance? Nope, sorry. Dude, they came to EVGA’s forums and replied to my posts about this (I ran all these benches, and completely destroyed this whole review, btw), and gave me their email (expreview@gmail.com- they have one fuckin email address, and it’s gmail? What pros…) so they could “explain” it to me, instead of just posting there. WTF is “VGA Frequency Extreme”, and why would you need to test it on a different machine than your TESTBED? Why, to achieve a 800/1200 overclock (that’s 150/280 over stock), would you use rivatuner, but then you say that for the 5mhz overclock to match the GTX, you used a BIOS flash? For 5mhz? Oh, wait, that was just their excuse for why the “Default Clocks” match the “Current clocks” in their GTS GPU-Z screenie, when it’s really a Photoshop job of the GTX readout. Oh, and, they wanted me to email them about this, instead of just posting on my forum thread. Since when is there a legit timedemo in CoD4? They said if I emailed them, they’d show me how to “get it to record”. Meaning, they are using console commands to draw FPS on-screen, and running FRAPS, and going through a level. THAT’S NOT A BENCHMARK, DON’T PRETEND THAT IT IS. A timedemo/bench is a preset route that doesn’t change, spends a certain amount of time looking at certain things, so that stuff like the SKY doesn’t skew your score with high points for lookin at shit that isn’t there. TOTALLY BOGUS. Oh, and notice how on the 4ghz “testbed”, they only score 355 points higher in 3DMark06 with the GTX than they do with the 3ghz Q6700 that they MISTAKENLY LEFT IN THE PICTURE. NOT POSSIBLE. A 1ghz increase in speed would have undoubtedly raised even the CPU score itself far more than 355 points, let alone how much it would’ve alleviated the bottlenecks on the cards and let them run at their full potential, raising the score overall FAR higher. I know I get about 5-6k increase overall in score from a 1ghz OC on my Q6600 with one GTS. Oh, gee, expreview guy that found me on my home forums, still waiting for that explanation. lol, he said he was happy I was confused. ?:P?
April 7th, 2008 at 1:57 am
@ Boz
It’s because they aren’t actually running any fuckin benches most likely. They pull scores off of BS sites like Fudzilla and other random places and pass them off as their own. I’m getting stepped-up by EVGA from 2 GTS 512’s to 2 GTX’s, head over to the EVGA forums in about a week or so for some REAL benches.
April 7th, 2008 at 1:59 am
Oh, and who couldn’t have predicted a GS version of the 8800. They do it with every popular series- release a nerfed version with a GS designation at a lower price. I guarantee it wasn’t “expreview” getting back-alley info from inside sources that led them to that “news flash”.
April 7th, 2008 at 5:56 am
Serivce Pack ade, the Great Warrior.
What about page 14 ? I getto a “rss feed page”…
April 7th, 2008 at 10:08 am
At whitehousewhore,
it is a bug of the site’s plugin, please reflash and it will be fine
April 7th, 2008 at 10:15 am
So, Jeff (the english editor they say, I think you mean “refresh”, not “reflash”, fuckin idiot), got anything to say about the fact that more than a couple of your reviews are now shown to be completely fake? I’m still waiting for an explanation.
April 7th, 2008 at 10:16 am
Pardon, cookies were deleted, but I think you knew who it was…
April 7th, 2008 at 10:22 am
@ Whitehousewhore,
Dude, it’s because this is can barely even be called a fuckin review site. It’s a blog hosting page. They don’t even have their own goddam emails for their “employees”. It is expreview@gmail.com…not jeff@expreview.com or anything like that, but a fucking GMAIL account. ONE gmail account. He followed me to the EVGA forums where I first discovered their nonsense and tried to give me these excuses and make me look like I was the one who was uninformed (I think we see his folly there), and then when I put the final nail in the coffin with the rest of what I discovered, he hasn’t replied since. Not on there, not on here. They’re trying to pretend like I didn’t just yank their bullshit out from under them and slap em across the mouth with it, but I diiiiiiiid. Yea, that RSS feed that is fixed by a refresh (not reflash, moron) is just another example of how shoddy this website is. Where else do you go that a completely different page loads at random, and you can sit and refresh until you get the right one lol?
April 7th, 2008 at 5:03 pm
1、About Drivers
Test platform: QX9650 (3GHz)+Asustek Maximus Formula+DDR2-800-4-4-4-10 (Link)
1.The Dirvers
Because of NV’s reason, you have to use different drivers for different cards if .inf is not changed. Like GeForce 9 drivers have been updated to 174.74, but GeForce 8 still stayed at 169.xx.
Like the chart we provide up here. the new driver can boost its performance of the card, and 17x even updated with Video playback enhancement.
Didn’t like other site’s review, In our review, we always mod NV_disp.inf to make the new driver to support dated cards. In our view, this will be more accurate than using different drivers for different cards.
NV like to release new drivers a few days before new product’s release. This time NVIDIA release 174.74beta in march 27, and the product release date is April Fool. We have to work overclock for this review.
But this time the funny thing is NV provide almost all their product imformation in the 174.74beta so this time we don’t need to mod that much. (For full product support please download this)
2. About testing number questions.
We can guarantee that all our test have been run in a fair environment. It is your call to trust it or not:
a: 3Dmark06 score issue.
From our news 9800GTX 3Dmark06 score here, don’t be too excited *Update* dated at Feb. At that time the ForceWare driver we used is 74.16 WHQL Mod. and platform is Asustek Maximus Formula with DDR2-800 4-4-4-10. You will understand what’s the card’s real protential by compare 3GHz Quad and 4GHz Quad. There is no bottleneck for the card but the score keeps simliar.
If you are going to ask why GPU-Z 0.1.7 can read GeForce 9800GTX, that’s because at those time we have provided TPU for what they want and they send us back a 0.1.7 with 9800GTX support. Do you have a 9800GTX? if you have, i can provide you that version.
b. FEAR
In our review, we used F.E.A.R: Perseus Mandate V1.0 build-in benchmark. Please notice Soft Shadows will greatly affect the final numbers. If you select soft shadows, you can’t run with anti-aliasing. You can either run with soft shadows and no AA, or with no soft shadows and AA. If you select soft shadows, you can’t run with anti-aliasing. You can either run with soft shadows and no AA, or with no soft shadows and AA.
c. COD4
It seems you never heard of COD4 timedemo? you can record like this and we will provide you our own timedemo to let you check our review.
1. Record the demo through the console with /record [name] and /stoprecord to finish it.
2. The demos will be stored in …/Call of Duty 4/main/demos
3. If you try to play back CoD4 demos directly, the game will crash. Instead, copy the demo folder into …/Call of Duty 4/Mods/ModWarfare and start Call of Duty 4.
4. Once loaded, select Mods and load ModWarefare and the game will re-start with the mod loaded.
5. Then bring down the console and type /demo [name] to watch the recording or /timedemo [name] to benchmark it.
6. The results will be shown in the extended console (Shift + `)
if you have FPS limiting, try to get rid of it:
1. change com_maxfps to xxx (max number can be 999,we use 300) instead 85(default) in the console (Shift + `)
ExpRev timedemo is here:
i will answer all your question no matter you have asked us in here or in Evga fourm in the future post.
d. UT3
We have modified UTEngine.ini, changed maxFPS from 62 to 300. Also we have run WAR-Sinkhole_fly using UT3 benchmark tools(link). We set all the graphic opinions in game instead of set it in benchmark tools.
3. Test platform:
We ran different test in different platforms. That’s because the time it is rather limited (we have less than three work days to finish this) . With three platforms we can run multiple tests in the same time.
1、basic platform：QX9650 OC 4GHz with Asustek Maximus Extreme
2、OC platform：Q6700 OC 333X9 with Asustek Strike Extreme
3、Power consumption platform：QX9650 with Asustek Maximus Formula
4. About Frequency
The biggest question you are holding right now is Page 9 and Page 12, core clock of two cards. First of all, in page 9 we use Platform 2 to test. Before our OC, the 3Dmark06 is 13770 and after OC to 800/2500MHz the number raise to 14607, raised 6%. In this page we are only focus on how high the clock can be, the 3Dmark06 run is only to make sure its stability.
In page 12, We did not use the conventional way to OC the card(like page 9). In this page we hope to set the clock more accurate so we use NiBiTor to change the card’s BIOS and flash it. Thats why GPU shows Default frequency as 675/2200MHz(DDR).
From your reply i am sure you have been confused in Page 9 and Page 12.
this message have been posted here:
Additional info:
because our email sever is facing some issue now, which can lead to rufuse foreign mail. you can try to email me at jeff # expreview.com.
April 8th, 2008 at 12:26 am
OMG, they just won’t give it up. No one said anything about different driver versions. He asked why the same cards with the same version perform VERY differently throughout your tests. And in FEAR, even with Soft Shadows, as well as EVERY OPTION AT MAX (besides AA), a 3.7ghz quad with a GTS gives an average 250fps, not 80. As far as the COD4 “timedemo”, IT DOESN’T EXIST. That is NOT a benchmark. A timedemo/bench is a predetermined run through a map set by the devs, not you. Recording yourself running through a map and calling it a benchmark isn’t accurate. If you ran through with your aim at 15% above level, your score would be higher than if you ran through with it 15% below level, because of the levels of detail varying from sky to ground. As far as all of your different “testbeds”, that’s fine and good, except it still doesn’t change the fact that (A) you show a 3ghz quad and a GTX scoring 15k in 3Dmark06, which a GTS simply won’t do, and (B) your scores ARE fake because due to the nature of 3DMark’s scoring, there is NO possible way that a 1ghz increase in CPU speed would result in only 355 points gained overall. We’ve already been over your BS a bunch of times, give it up. They are lies, you know it, and I know it. It looks like you just picked scores at random, because most of these aren’t even REMOTELY accurate for the “testbed” you supposedly ran them on. I don’t know what sort of rank amateur you think you are fooling, but it isn’t me slick. I guarantee I’ve run 3DMark and other benches more times in the last 2 months than your whole “team” has in their collective, pitiful, lying lives. I love how you try to paint me as some rookie who doesn’t know what they’re talking about, with shit like “It seems you never heard of COD4 timedemo?”, and “From your reply I am sure you have been confused”. No, I haven’t, because it is something you made up. And, I’m not confused, I’m well aware of everything being discussed here. I don’t even get how you are still trying to defend this. You are so obviously lying it is incredible. Like I said before, if you’re going to lie about scores, you might want to actually RUN the benchmarks first, and then go from there, since a lot of your scores for EITHER card are off by more than 50%, and in some cases are 1/3 of what they should be. You aren’t fooling anyone guys, anymore at least.
April 8th, 2008 at 1:02 am
1.COD4/HL2/Quakewar/Bioshock/UT3 doesnt exist build-in benchmark,so every test cancel? How can I make every personl or media use same timedemo or setting. We test hardware day and night and never make fake number.
2.about 3dmark score,I think you need find out more sample. You alway have your standard even we show you another link has the 3dmark pics,so I dont want to say anything about it.And I am pretty sure our test number is ture.
3.Why we say no to 9800gtx,because we feel about 3~4% increase is not very good option to upgrade VGA card from 8800GTS 512 and waste more money, this conclusion is up to single card test. Our 3way test coming soon.
After what jeff or my answer, we are very serious about what you ask, and I am sorry to see your comment about what expreview.com doing.
Even I am poor of english, I already try my best to answer more question.
Have a nice day!
April 8th, 2008 at 1:30 am
Interesting test, I am not surprised with the outcome however as the card is basically the same as the 8800gts 512 and good call for not recommending it. On a side note, I didn’t expect the 3870×2 to perform that well after the hotfix (just take a look at the 3D mark scores and compare it to before!), too bad it is still the worst in power consumption. Keep it coming.
April 8th, 2008 at 9:45 am
OMG, this guy just won’t give it up. heheh..funny
Go buy a 9800gtx and stop wasting time here, if you really have so much confidence on this poor product.
April 8th, 2008 at 1:24 pm
Um, isn’t it you that keeps going back to the EVGA forums and posting more and more of your nonsense, basically begging for me to pick it apart and disprove it, yet again? Like I said before, it isn’t about the GTX vs. GTS, I already own 2 GTS’ and love them. It is about calling bullshit when I see it, such as 81fps average in a 3 year old game with a 4ghz quad core and top end video card. Idiot.
April 8th, 2008 at 11:11 pm
Hello and good day spAde,
I read one of your posts and was wondering if you (and everyone else) noticed that the differences between the score marks for the 3870×2 is the resultant of the Hot Fix (as pointed out by boz)? As for unspectacular framerates for FEAR and a few games, it might be due to the newer drivers or some other form of variables, Vista perhaps. A collegue of mine has much worse framerates running UT3 on Vista than on XP, 30’s compared to 55+. Or maybe the 9800gtx is memory limited with 4xAA and 16xAF at such high resolutions… I am uncertain, though it will be revealed with further testings in other review sites with updated drivers and such. Also I would like to point out, with all due respect spAde (and everyone else), that opinions are most welcomed anywhere, but common courtesy and respect comes first, and it would unfortunate if they are forgotten and replaced with unnecessary foul languages, even annonymously in a place such as here. If common courtesy and respect are not shown here, then they will be neglected outside of it. We are far from being foul people, so let’s keep it clean here.
April 9th, 2008 at 12:28 am
Hotfixes aside, the tests on FEAR were run on as identical a setup as possible. When I talk about the 81fps, it was referring to the GTS bench, which I also ran, and got 255fps on. Same drivers, same OS, same settings (maybe, IDK if they maxed everything like I did). The Crysis run was similar. my 25fps at “Very High” compared to 21 at “High”, same drivers/OS, and my CPU speed was 300mhz lower. I would like to know how you explain a 3DMark score’s lack of gain with a 1ghz increase in CPU speed? In that situation, the card being tested doesn’t matter. The CPU score alone would change by the thousands, not less than 500. Sorry slick, not buyin it, quit sellin it. I know you’d like to paint me as some rookie who is confused (here and at EVGA), but it just isn’t the case.
April 9th, 2008 at 5:36 pm
Hello Anonymous,
I am unsure about how you managed to reach 255fps (is it max framerate or average?) on a FEAR bench with everything set to max/close to max qualities. Most of the reviews on the 8800gts 512 I recalled do not have that level of framerates, maybe the counter is off or it averages wrong? I am not sure. For Crysis, I am most certain that most private users have tweaked the configs to get better framerates and/or image quality, and thus better gameplay (sun/moon rays too), testers are just testing, with no intention on gaming so everything is left untouched. The explaination for the 3DMark score, I have seen it elsewhere so I am not surprised. We both know that ghz increase in the CPU, no matter how large, won’t affect the score by a huge significance. I could be wrong with everything and I welcome for any clairification.
April 9th, 2008 at 9:00 pm
So you guys are saying these benches are fakes?
Makes me not wanna come back here anymore…
April 10th, 2008 at 4:07 pm
you know what, We at Exp won’t give a damn about the Anonymous guy AKA spAde.
faking a 14 page review in a no-Advertisement website? what for?
if you don’t trust review @ here, run testes yourself or go read those bribed review out there!
April 10th, 2008 at 6:28 pm
lol i think spAde shouldnt even bother with this site if he disagrees. Jeff has proven both of us wrong, and the lil doubts i had have now been solved @ answered. This will still be my no1 tech site along with fudo, coz THEYRE BOTH RELIABLE.
April 10th, 2008 at 9:48 pm
It is just unfortunate that some people can be quite negative and offended such as spAde/Anonymous (and quite a few) with the results for this review, it is as if they had lost a million dollars in a fair bet and would claw in any which way possible to get them back. I would like to question the legitimacy of spAde’s (insistive nature with his) 200+ average(?) framerates for FEAR and 3DMark 06 CPU score. As for the “provacative” test, thus far, I haven’t seen any solid challenge to the accuracy for the tests done in the review, I therefore agree that they are correct and fair. And again, good call for not recommending the purchase of the 9800gtx. A higher model number isn’t necessarily a better model.
April 12th, 2008 at 4:27 am
[...] Expreview [...]
April 12th, 2008 at 4:30 am
lol, that’s the funny part. I did run the tests, at identical specs besides my quad at 3.7ghz compared to your 4.0ghz, which is the basis for saying this is all false. If you would like screenshots of the FEAR results, I can give them very easily. And, nobody has proven me wrong. Putting up graphs that you make doesn’t make it fact. Their 3DMark06 score for the GTS is reasonable, but 3DMark06 isn’t really a good bench for new cards, as well as the fact that despite what Aramint says, CPU speed plays a HUGE factor, both in the CPU score, as well as the total score, since the GPU can’t render frames without the CPU first processing them, making the GPU fully dependent on the CPU. Read up on how benchmarks actually work, and run these tests YOURSELF (since I already have), and then you can talk about it.
April 12th, 2008 at 7:48 pm
Pics please and we can all rest this out. With regards.
April 13th, 2008 at 12:02 pm
@spAde:
All talk and no proof
April 14th, 2008 at 3:24 am
Hello boz,
I hope he has proof otherwise I will discredit all of his posts here. And to spAde… Here is one of the few reasons why I say CPU high overclock has no real significance with the final 3DMark06 score, though I am entirely wrong as you mentioned: http://www.overclock.net/amd-general/219099-old-ocn-top-40-3dmark06-amd.html Real users with everyday computers. I trust average useable computers than a singular result from a singular computer or groups of computers that are made to break 3DMark06 scores. I have no worries if I am wrong entirely, but I have seen enough to conclude and defend what I have said. Most sincerest of apologies to Expreview if I am dragging this too long, if it has become a bothersome, then I will end.
April 14th, 2008 at 9:59 am
Well, I don’t know about the numbers but I’d say that they were bang on in the GTS/512 comparison(s). When you factor in other benchmarks and games the % difference is more like 2% overall.
That’s not to say that the 9800GTX isn’t a good card because it is. However, under current pricing and with the 8800GTS/512 still alive in the channel at much lower prices … it’s not a sound buy. I agree with their conclusions to stay away from it.
Had the 9800GTX been a 512bit part .. it would be a different story. But it’s not. Nividia should have just called it a 9800GTS as the GTX branding is misleading.
April 14th, 2008 at 10:29 am
this is a load of crap, i’m gonna agree with spade, check out extreme systems, they have benched it and have gotten about 18k with about the same setup. boz, this site and fudo are both very UNRELIABLE, if you want real info visit EVGA’s forums or extreme systems, they’ll tell you straight up how this stuff is.
April 14th, 2008 at 10:50 am
to lckyby517:
show me the link of 3dmark06 with 4GHz(400X10) with DDR3-1066 6-6-6-18-1T / 9800GTX default/Windows Vista SP1/174.74beta driver. Please make sure what you say is similar of correct.
No matter in Evga forum or XS.
April 14th, 2008 at 10:52 am
to lckyby517:
How this stuff is? what are you talking about?
Show me the link of 3dmark06 with 4GHz with DDR3-1066 6-6-6-18-1T/9800GTX Default with 174.74/Windows Vista SP1.No matter in evga forum or XS.
April 14th, 2008 at 11:05 am
After what we doing with lots of VGA card test,special first 8800gt review,3450,3650,8800gts 512,8800gs,9800gx2 and…
I think we have enough data can sure that what we test was serious and nothing fake,So I dont want to make another thread about T of F discuss.I know people alway have question.
In this article, few guy may dont like the title “Say no to Single 9800gtx”, but there is no reson to setup up of expreview. That make me upset.
Our team always doing our best to reader, and we will going on.
April 16th, 2008 at 9:38 am
@ Aramint
Note though, how those are all AMD CPUs, not Intels. The performance clock for clock is highly different, and so using this to compare to an Intel testbed is moot.
April 16th, 2008 at 8:33 pm
Sorry spAde, but I am done here. Keep going though, someone will sure pick up after this.
April 17th, 2008 at 9:43 pm
Spade, Intel or AMD pure CPU clocks dont really affect the final 3Dmark06 scores… You might gain a couple extra hundred points here and there, but the reason why you get huge increases is typically because the CPU is still bottlenecking the GPU. In this case the score will continue to scale well with every incremental increase of the CPU core clocks. However, with stock default clocks, the 9800GTX is expended at about 14.4k, so any extra CPU clock increase wont yeild any significant results. Now if you overclock the 9800GTX to about 850ish core and such, you should be able to max it out to 17kish… For this, youll need a C2Q running at roughtly 3.7Ghz… As for your FEAR scores, with soft shadow I can ganrantee your not getting 250 avg., just get the latest version max, everything out and play the game with fraps and youll notice your avg. frames are around the 80-85fps mark…
April 18th, 2008 at 4:01 am
Well, I can guarantee that I do. I can youtube a fraps demo if you like.
April 18th, 2008 at 5:17 am
@ Zer
What I mean is, bottleneck or no, there’s no way that the CPU score alone (Yes, 3DMark assigns separate scores that are ADDED TOGETHER for the total) would not be greatly affected by a 1ghz increase, having nothing to do with the GPU in that regard. I will say I do stand somewhat corrected as far as FEAR. It seems when benching this one, SLI was not properly disabled (apparently these drivers require a reboot), which accounts for the score (which, to the naysayers of SLI, all depends on the game/settings). However, running just now GTX clocks on one card (other one is in a box atm), my max was around 210, average 132, low 68. Erroneous or no, 132 is still a far cry from 80, considering my CPU is a full 300mhz slower.
@ Aramint
You might consider doing some research on this stuff before speaking about it. Telling me that a 3ghz AMD and a 3ghz Intel QX9650 will give roughly the same score in 3DMark using the same cards is laughable.
@ Zer
They proved your theory wrong with their screenshots. To say that CPU clock increase won’t yield results is fundamentally wrong because of the fact that the CPU is tested independently. You all seem to be under the impression that 3DMark assigns one score on average for all of the tests, instead of it ADDING SCORES FROM EACH INDIVIDUAL TEST TO DETERMINE THE TOTAL. If at 3ghz it does 14.6k, there is NO POSSIBLE WAY that at 4ghz it would be less than 400 points higher. Email FutureMark, they’ll tell you the same thing. This isn’t an argument about whether or not it is bottlenecked by the CPU. It is about whether the results are accurate, which if you know how 3DMark scoring works, you would know they are not. Tell you what. Put a quad core at 3ghz, and run ONLY the CPU test in 3DMark, and then OC to 4ghz (or as high as you can get), and take a look at the score difference. At this point, the GPU isn’t even a factor.
April 18th, 2008 at 5:27 am
No need, I did them for you.
Q6600 @ 3.7ghz CPU Score 5143
Q6600 @ 2.7ghz CPU Score 3860
A difference of 1283, not 355. This is how it affected the CPU score alone, without GPU bottlenecking coming into play.
April 18th, 2008 at 5:29 am
Wish there was an edit button…
This also is from the same CPU, not a Q6700 @ 3ghz vs a FAR SUPERIOR QX9650 @ 4ghz, in which case the score difference would be even higher.
April 18th, 2008 at 7:42 am
Read the text in its enteity, the [/b]FINAL score[/b]…
April 18th, 2008 at 7:44 am
Do a complete run with your CPU running at 3.3Ghz then try 3.7Ghz, the score will only vary a few hundred points…
April 18th, 2008 at 7:47 am
Ah I just read your first reply… The scores arent added linearly… Jeez, they are added into a formula which then gives you a final global score:/
April 18th, 2008 at 8:04 am
Heres the formula… 3DMark Score = 2.5 x 1.0/(( 1.7/GS + 0.3/CPU Score )/2)
April 18th, 2008 at 5:14 pm
This is my real final comment. spAde, you were wrong with your FEAR bench as you have admitted, until I see pictures (not numbers posted in the comments section), then I will still discredit all of your posts. Only reply my comment with a photo, or a link, no @Aramint or usage of my name. I am truely done commenting but I will keep reading. With regards.
April 19th, 2008 at 7:24 am
What does the difference between 3.4 to 3.7 have to do with anything? It was a question of the difference in score between 3.0 and 4.0, which is what they show. Where are these numbers from? And, the difference is, when I make a mistake, I admit it, not try to cover it up with random bs and lies. And it doesn’t matter how the scores are added with only the CPU test run, since there’s nothing to add.
April 19th, 2008 at 11:25 am
“And it doesn’t matter how the scores are added with only the CPU test run, since there’s nothing to add.”
We were always taking about the effect of the CPU clock on the FINAL score, sigh…
Why change the story now? One day your saying the CPU score is the end all because it should make the final score jump 1200points, when in reality it doesnt since with the Formula 3Dmark06 uses to calculate the FINAL score, which is what this review displayed, it has much less impact that the GS(Gaming Score i.e: SM2.0/3.0 etc..). Your failure to understand even how 3Dmark06 works, as in when the GPU is no longer bottlenecked by the processor, hence showing barely any additional increase in the GS score comparatively to the CPU score, creates very small gains because the bulk of the 3Dmark06 score lies within the GS score…
All Im goin to say here is; that you attacked a completely legit review because YOU tought it was wrong, this due to your misundertanding of how 3Dmark06 works and the principles behind GPU to CPU relationships in that particular benchmark.
April 20th, 2008 at 6:39 am
I’m curious, at what point is the GPU no longer bottlenecked by the processor, according to you?
April 20th, 2008 at 7:07 am
Well, the best way is to test:)I mean, I cant just tell you a fixed number because there way too many variables present that could affect this such as system setup, GPU model and clocks, SLI or Crossfire configurations, drivers etc..
However, I can explain what usually happens when benching a CPU limited system (default settings, once AA and AF are used, this goes back to beig GPU limited, given the CPU is clocked decently) In my experience and from what Ive gathered from many other individuals, the moment your GPU is no longer bottlenecked by the CPU is when small increamental gains yeild poor results… Typically if a CPU is bottlenecking a GPU, you could see as much as 300-450extra 3Dmarks points on the final score for every 100Mhz jump of frenquency. Once the final score stops scaling well, is when youve pretty much hit the GPUs max rendering power and now are relying on brute CPU scores to yield any additional results…
April 24th, 2008 at 4:56 am
Here’s a second opinion for anyone who wishes to look. Believe or disbelieve, up to you.
http://www.guru3d.com/article/content/516/1/
April 29th, 2008 at 2:01 am
[...] This is absolutely not true. Best bang for the buck? How could you possibly reach that conclusion? Expreview has clearly shown us the 9800GTX performs only around 4-5% better than the 8800GTS G92, overall. [...]
May 4th, 2008 at 3:47 pm
First off spAde, try to keep it polite eh? Slagging somebody off for their written English is real lowest-common-denominator stuff. No need to resort to racism to make your point.
Secondly, drivers (and other variables, as a number of commentators have said) can make a huge difference. I only stumbled on this site because I’m searching for a FEAR fix. I’m currently getting 13 (yes, thirteen) FPS in FEAR on a 3ghz Q6600 with a 3870×2 - latest patch, latest drivers. I used to get wildly varying FEAR results from one test to the next on my 8800GTX, between about 60 avge and about 240 avge. For comparison’s sake, my Crysis FPS are consistently up around 35 at 1680×1050 on high, DX10.
At worst, Expreview may need to refine their benchmarking techniques to ensure consistency of results; regardless, why don’t you lay off your barrage of vitriol and go and enjoy your two GTSs?
May 9th, 2008 at 9:16 am
[...] is slightly better then the GTS. But there is no doubt they are very similar. Yep. The numbers (Expreview) show an 9800GTX performs between 4-5% better, on average, than an 8800GTS 512MB in games. Just for [...]
May 10th, 2008 at 6:25 am
lol this is still going on…
May 21st, 2008 at 4:49 pm
[...] the memory actually performs better on the GTS at the same frequency. This has been proven by Expreview. To demonstrate, they overclocked the 8800GTS G92 and 9800GTX to the EXACT same clocks - and the [...]
June 16th, 2008 at 6:50 pm
[...] sper sa va suparati ca ma bag si eu in seama ..este vb de placile mai vechi mai exact despre 9800gtx!! care dupa cum se poate vedea si in linkul de mai jos, nu merita cumparate http://en.expreview.com/2008/04/03/g…gtx-review/11/ [...]
July 9th, 2008 at 11:26 pm
I’ve given up reading through all these comments, as it seems that the mouthy, ill-informed, and obnoxious spAde has rendered this space a waste. For what it’s worth, spAde, while you are obviously not an idiot, you type like one. You have all the subtlety of expression one might observe in chimpanzees on heat, and compound some very fundamental errors in your understanding of the facts with an attitude that only serves to embarrass you and all those you might be related to.
There is a way to air one’s opinion - in your case, a little discretion may have served you better than to keep spouting your crap as loudly as possible. Were you the sort of kid who used to cover your ears and hum so that you couldn’t be told off by your teacher? This might be why you have a) failed to learn manners, b) failed to learn at all.
While I agree that there are issues in these benchmarks that need to be addressed if they are to be taken more seriously, you have issues that need to be medicated if you don’t want to end up a gibbering twat… scratch that - you evidently already are.
July 10th, 2008 at 12:39 pm
OK, having just acquired a new rig, here are some unbiased scores for reference
Can’t comment on any other cards etc. Not in it for bragging, just found that is was difficult to find examples on the net. Hope this helps.
I found that overclocking the CPU got the majority of the performance increase, I think that the 9800 is an average card, but all that was available at the time I bought this kit. The processor appeared to “bottleneck”before the card, and disk speed may also be a limiting factor.
Rig
GigaByte GA-EP35-DS3P Rev 2.1 F3 Bios
Intel Core2 Quad Q9450 2.66
Gigabyte 9800GTX 512MB 675/1100/1688
4GB Kingston 1066Mhz DDR2 RAM
Samsung 500GB HDD
3DMark06
XP Service Pack 3
NVidia 175.16 drvers
13825 3DMarks
XP SP3 O/Clocked CPU FSB 400Mhz VGA 800/1200/2000
16501 3DMarks
Standard cooling, VGA fan @100% no temp problems and rock solid stable
Vista Ultimate
NVidia 175.16 drivers
13322 3DMarks