The GT300/Fermi Thread - Part 2!

[annihilat0r]

well if one is 1fps and the other is 3 then 2 fps means much

683/734 means a 10% increase... not too bad

[eric66]

if that happens (prices dropping to $200) then it's the only time 5830 could be considered for purchase. A $240 card can't be counted as a good card because "it's going to drop in price". The 5830 release sucks and as far as I know the only site to actually praise this launch was HardoCP. "perfect performance perfect price" lolz

performance is there but price sucks (i blame useless expensive 5870 pcbs) thats true and amd cut too much may be but that card will be only dx 11 card with acceptable performance and affordable price

[eric66]

Yeah, that was probably the worst graph ever. Looks like a 100% increase from 5870 but in reality it's less than 10%.

yeah just saw mads name on graphs so that explains great looking green bars

[annihilat0r]

performance is there but price sucks (i blame useless expensive 5870 pcbs) thats true and amd cut too much may be but that card will be only dx 11 card with acceptable performance and affordable price

"performance is good but price is not" is complete nonsense. The only case this could be true if it was the fastest card ever built or something like that.

When its performance is in the middle of somewhere, performance and PRICE are the two variables in the kick-ass-card function. Its priced $240 so its performance is not good at all, it is bad. If it was priced $200 its performance would be "good" and even then not extremely so

[informal]

I think it's much more interesting to see what the frame rates will be like in an in-game benchmark run with the use of FRAPS for example. Do that run for a minute or so, start at around a minute into the game and repeat 5 times to take the average. That's what I want to see.

Exactly. Those min fps values are often I/O related (loading scene) and not strictly GPU related. The scenario you describe above is good starting point.
Looking at the posted numbers for the 470 model I'm really not impressed. The chip is huge,hot and clearly not designed with gaming in mind(but GPGPU). It will be inefficient in gaming from the perf./watt/$ POV and AMD scored a big win with the "small&scalable" approach. AMD will have little trouble competing with top fermi model even with single turbo clocked Cypress... And the 6xxx series refresh is not even here...

[Piotrsama]

I was thinking "wow, ownage"...
Then I realized what a curious graph that unigine one is

[Dimitriman]

Exactly. Those min fps values are often I/O related (loading scene) and not strictly GPU related. The scenario you describe above is good starting point.
Looking at the posted numbers for the 470 model I'm really not impressed. The chip is huge,hot and clearly not designed with gaming in mind(but GPGPU). It will be inefficient in gaming from the perf./watt/$ POV and AMD scored a big win with the "small&scalable" approach. AMD will have little trouble competing with top fermi model even with single turbo clocked Cypress... And the 6xxx series refresh is not even here...

You mean 5000 series "refresh"

6k series will not be a refresh, as stated many times by AMD since last year or maybe longer than that. New architecture, most likely Q4 launch with Q1 availability.

Now a "5890" with 1Ghz clock should compete with the Fermi just fine although I'm afraid of what AMD will charge for that one.

[Yvese]

Really, comparing two cards that don't have the same mem size at high resolutions is just silly. Of course a 1GB card is going to lose vs a higher memory card at 2560 res.

Come back with a 2GB 5870 comparison then we can see realistic results. Sadly, both cards don't exist for consumers atm

[Chumbucket843]

Exactly. Those min fps values are often I/O related (loading scene) and not strictly GPU related. The scenario you describe above is good starting point.
Looking at the posted numbers for the 470 model I'm really not impressed. The chip is huge,hot and clearly not designed with gaming in mind(but GPGPU). It will be inefficient in gaming from the perf./watt/$ POV and AMD scored a big win with the "small&scalable" approach. AMD will have little trouble competing with top fermi model even with single turbo clocked Cypress... And the 6xxx series refresh is not even here...

where did you get the idea fermi was designed without gaming/graphics in mind? it has one of the biggest firsts for graphics: parallel setup. ATi has some patents on this too so we might see that in their future gpu's. double precision isnt as fast on geforce and the caches are used for graphics. there isnt much gpgpu hardware holding back graphics. they are not that different anyways.

read the graphics whitepaper.
http://www.nvidia.com/object/IO_86775.html

[Marios]

GTX 470 with 220W TDP doing 10% worse than 5870? No thanks.

[jam2k]

LOL @ these cherry picked benchmarks @2560x1600 where the 5870 is obviously running out of memory.

[mibo]

Some guy at DH posted these, have you seen this before?

I would call this "betraying your readers". The GTX470 is less than 10% faster in Unigine what is not represented by the graph at all.

[annihilat0r]

yeah, it looks like the performance slides from both AMD and Nvidia. An independent reviewing site shouldn't do things like this

[Marios]

The reviewer gave numbers, not graphs.

[zerazax]

It's only the Unigine one that's off, and the scores are misleading: 29 vs 27FPS is not 734/683

Anyways, min fps on one Crysis Warhead run? He should have run it at least 3 times, to eliminate the possibility that it was a single hitch or I/O error for example. Then again, PHK got in trouble for biased benching the GTX2xx's vs the RV770 back in the day./shrug

[annihilat0r]

the scores are misleading: 29 vs 27FPS is not 734/683

Hahahahaha what? 29/27 is equal to 734/683. Who cares if its 2900 to 2700 or 29 to 27? The difference is the same.

[jmke]

Some guy at DH posted these, have you seen this before?

of course I've "seen" them before, made them from the data posted in this thread...
http://www.madshrimps.be/vbulletin/f...website-70508/

I love the Unigine graph. It almost looks like it's done by nV

love how they used score instead of fps same chinese numbers but with awesome bars difference was 2 fps as much i remember

Yeah, that was probably the worst graph ever. Looks like a 100% increase from 5870 but in reality it's less than 10%.

yeah just saw mads name on graphs so that explains great looking green bars

do I detect sarcasm? ;p

I was thinking "wow, ownage"...
Then I realized what a curious graph that unigine one is

I would call this "betraying your readers". The GTX470 is less than 10% faster in Unigine what is not represented by the graph at all.

blame Excel 2007
for the other charts it auto-scaled the values correctly, starting at 0, that's how I like all my charts (check www.madshrimps.be charts if you don't believe me);
for the last one it didn't start from 0 for some reason

ah well... thing to take away from those chinese benchmarks is that GTX 470 is slower than HD 5870, no matter what synthetic benchmark output says

[LiquidReactor]

Some guy at DH posted these, have you seen this before?

What...GTX 470 is worse then GTX 295 by those numbers.

[annihilat0r]

What...GTX 470 is worse then GTX 295 by those numbers.

This is getting ridiculous. At launch the highest Cypress chip (5870) was slower than the x2 version of the latest generation. (4870x2 > 5870) Now it's a problem that 470 (not even 480) is slower than a GTX 295?

[mapel110]

This is getting ridiculous. At launch the highest Cypress chip (5870) was slower than the x2 version of the latest generation. (4870x2 > 5870) Now it's a problem that 470 (not even 480) is slower than a GTX 295?

Yes, because ATI launched a mid range chip with 2.15 Billion Transistor and nvidia is releasing a real high end monster with 3 Billion Transistors.

[annihilat0r]

Since when a 5870 is midrange and a GTX 470 is a "real high end monster"? That's funny

[LiquidReactor]

This is getting ridiculous. At launch the highest Cypress chip (5870) was slower than the x2 version of the latest generation. (4870x2 > 5870) Now it's a problem that 470 (not even 480) is slower than a GTX 295?

Yes, because ATI launched a mid range chip with 2.15 Billion Transistor and nvidia is releasing a real high end monster with 3 Billion Transistors.

Well thats cause ATI plainly screwed. In three weeks we will know if same happened with nVidia.

[Andrea deluxe]

edit

[zerazax]

Hahahahaha what? 29/27 is equal to 734/683. Who cares if its 2900 to 2700 or 29 to 27? The difference is the same.

Oops, you're right, I screwed up my math

Well thats cause ATI plainly screwed. In three weeks we will know if same happened with nVidia.

Yes, because ATI launched a mid range chip with 2.15 Billion Transistor and nvidia is releasing a real high end monster with 3 Billion Transistors.

Uh, and how do you guys know that games aren't reaching the limits of shading/texture power? Neither side might have "screwed up", it might be that hardware is reaching a saturation point

[kemo]

Uh, and how do you guys know that games aren't reaching the limits of shading/texture power? Neither side might have "screwed up", it might be that hardware is reaching a saturation point

Saturation point is when you switch from a 5870 to a 5970 in a game supporting crossfire and gets the same fps , then overclock your CPU and keep getting the same fps