:toast:Quote:
Originally Posted by Overmelted
Printable View
:toast:Quote:
Originally Posted by Overmelted
just a question about gpu's here
the TMU's = raw pixel performance right?
the shaderunits (ALU)= calculations on pixels renderded by the TMU's?
so if an ATI card has too few TMU's arent the ALU's a lot idle?
because there arent pixels to do caculations for?
its just a silly question, and i dont know enough about graphic cards for this stuff, but i am sure some guys here do
http://www.inquirerinside.com/?article=29152
but then this whole thread is all pure speculation.
typical gf4 post I guess(don't read often and looks like gf4 hasn't changed a bit)
seriously,
why would ati release a card thats around the same 3dmark scores as the x1800xt
TMUs are texture mapping units; directly analogous to pixel performance when shader operations aren't present. The primary ALUs do the calculations on the pixels rendered from the TMUs. In less shader heavy titles any performance bottleneck will therefore be the TMUs (in X1900's case).Quote:
Originally Posted by GoThr3k
Both ATI and nV have stated they believe a 3:1 ALU:TMU ratio is the way forward in respect to future games, and thus hardware; ie as shaders become more prevalent. This is the ratio we see in X1900s; whether or not we actually see games in this series lifetime that fully use this shader power is unknown, although UT2007 and UE3 based games could be good candidates.
And perkam, thanks for the welcome :) :) :)
Those tests were run on a stock Opteron @ 2.2Ghz, with average memory timings. VR-Zone showed you what it will do in a real benching environment. Thats a big difference in CPU power.
The 9688 score was attained again with stock card clocks on a 2.97Ghz 148
http://www.hardforum.com/showpost.ph...6&postcount=20
Would be a honour to see the results... ;)Quote:
Originally Posted by hipro5
A fair comparison, between the 3 most talked about cards around at the moment by a Overclocker! We cant ask for a better comparison than that, can we...
Good luck with the testing :toast:
People really shouldln't bother replying to this guys threads.
Im just wonder,which driver u use on x1900 card.On 6.1 i cant see nowhere 1900 and 580 core.
FYI with a bunch of crap running and my X1800XT stock PE BIOS(700/800) I got 9994. I get over 11300 at the higher overclocks. Thats my 24/7 settings at 2.6Ghz.
What cpu is that with ?
What system specs / speed?
those "scores" are bs:cussing:
How have you come to that conclusion?
Is good saying that they are BS, but not if you dont explain why...
wait 2 days and you'll seeQuote:
Originally Posted by Bennah
people don't have to post anything or even post a theory as to why these scores are wrong.
simple facts here, Gf4ti is saying these cards are crap because they are 9000k 2k5 stock,and the 7900 scores 13k. the ONLY place there are 7900 rumoured scores also say the x1900xtx WILL get 11.5k STOCK, they also show the 7900 was run on a fx60@2.8Ghz and the x1900 was on a FX57 at same clocks, we all know for a fact the dual core will give a point boost so they won't be THAT far apart.
GF4, how does going from a 24pipe 7800 to a 32 pipe 7900 mean its got 100% pipe increase, its 8 pipes extra, its a decent increase but no where near 100% increase in raw juice.
the x1900xtx will be(looking at clocks((not being OTT)) ) VERY AVAILABLE, ON LAUNCH. that means rrp prices within days(after few stores piss about with pre order prices and trying to convince everyone stocks will be low so pay 20% more). in the uk that means within a week of launch i can get a x1900xtx for £370 or a 7800 512MB, and much slower card, for £480, if i can find one in stock(most gtx 512's in the uk, also no stock show as between £500-530ish). they are NOT in the same price bracket, officially they are on paper, unofficially there are very much not, and very much not the same kinda availability. there is simply no question the x1900xtx will outperform and be cheaper than the 7800 series, which is now dead, get over it.
the 7900 looks like a freaking beast, but looking at jump in clock speeds, and using same impossible to get memory that 7900 scoring 13k(assuming its accurate(and it probo is as its a very scaled card, its faster and has similar number of extra pipes so it should perform like that. ie over the 7800 is has 33% more pipes, 33% more speed, but it has more shaders on those pipes, its an even increase across the board so 33% faster scores are likely). sorry long bit there getting away from the point. that 7900 13k card will be the 7800 gtx 512mb, it will highly likely be VERY limited, end up costing more than rrp due to stocks(like the 7800 512). i would expect a 32 pipe maybe 600Mhz card , lower mem clocks too but still 512mb just under the x1900xtx cost, probo $580 or something but due to lower clocks will be closer to the x1900xtx in performance, but personally i think it will be faster than the 1900 in all but a few extreme shader games(ut2k7 is the only one i can think of thats coming, fear aswell).
however, assuming nvidia and ati stick to what they've done i will be pretty surprised if we don't see firstly the x1900xt hitting the xtx speeds easily and beyond. also the 7900gt will to keep costs down and give failed 32pipe cores a chance to sell, have less pipes, 24, 28? dunno. but when used in games a 24/28 pipe cheapish card will have some serious power, obviously won't compete at the top end in benching but in games i'd doubt it would be easy to see the difference.
for the love of christ, http://www.vrforums.com/showthread.php?t=53164
same card, same clocks, the slower card scoring pretty much as the leaked info(that said 7900 13k, x1900xtx 11.5, x1900xt 10.5k) were saying. also the fx to the dual core gave a 700 point bump in score alone.
as people have said all through this thread 9000k is a x1800xt stock score, the x1800xl will give around the 8.2k mark IIRC, the xt is clocked way higher than the xl at stock.
H thread scores are BULL, VRzone thread scores are MUCH more on target.
but again we really haven't seen what either cards7900/1900 can do overclocking wise. ati is on 2nd 0.9nm process, and had issues with first cards pcb's and getting high stable speeds, i'll assume they've fixed that (or done a bit better) on the newer cards. will they overclock further than the 7900, maybe.
tbh, as per usual, despite what 3dmark says if you pay for either top end or the model below card in games, all but one or two big exceptions, you will generally fail to notice a difference between the cards. they tend to compete well as always, no matter what people say the 9800 was a touch faster than the 5900 but play on either IN GAMES and it was hard to notice a huge fps difference. x800 vs 6800 series, again not that much in it at all, 7800 vs x1800, again, had both, both play very well and very similarly.
get the card that is better value or has a feature or IQ level you prefer, everything else is just nitpicking.
Hze it's not possible man...Who have make this bench in first page...
at 2.9Ghz i have 10'070 with Pe bios without i m close to 9900... impossible thing... and this is with mipmap in quality setting.. just no AA no AF..
For each review i have see make, they give the bench of X1900 series at more of 12K..OC have pass the 13'500
the x1900s were a few months late?
I wonder how my 4400+ at 2.5 ghz crushed your opty at 2.97 ghz in the CPU dept???
Im unsure what to think atm. So much results going round with different speeds etc.
I just want to see what hipro5 results come out like...
When is the X1900XT arrving in the states anyways?
X2 @ 2.6Ghz. The PE BIOS puts the card at 700/800. My 11350 score had the CPU at 2.8Ghz and video at 775/875.Quote:
Originally Posted by Bennah
Actually the X1900 line is a refresh of the X1800 where the 7900 is a die shrink which means new technology.Quote:
Originally Posted by Willis
AgreedQuote:
Originally Posted by G H Z
your math skills are lacking, a 100% increase in pipes would be 48 pipes.Quote:
Originally Posted by Geforce4ti4200
32-24=8 8/24 = 33% increase in pipes :rolleyes: Also, it has around a 33% boost in clock speeds, so we have to conclude that it will be less than 33% faster than the gtx 512. We're just using common sense here
Yes keep using 3dmark even though 3dmark doesn't take advantage of what separated the 1900 from the 1800, SHADER PERFORMANCEQuote:
"Don't worry........On Monday/Tuesday I'll get an X1900XTX to test.......I already have an 7800GTX 512MB and an X1800XT in my hands......I'll compare all three of them together with the EXACT SAME config........I think that this will be fair enough......."
you are my hero! :woot: :toast: be sure to test in all four versons of 3dmark as well as whatever games you have! Keep the cpu, ram, etc constant and the video cards stock then overclocked.
Get your mind out of the pipe mindset, and btw, it competes with whatever is out, which is the gtx 512. Just like the 7800gtx when first out competed with the x850xtQuote:
"It's not X1900XT vs 7800 it's vs 7900."
Then why did ati release the x1900xt now and with only 16 pipes?
you could also play it nvidia fanboy style and say the fastest out *available* which is the gtx 256...Quote:
Originally Posted by sabrewolf732
anyway, i dont see why some (most ?) people who got a 6600gt/x1300 give 2 :banana::banana::banana::banana::banana:s on who is faster between a 7800gtx and a x1800xt