The GT300/Fermi Thread

[NapalmV5]

one thing i know a3 revision must be some miracle to make those claims true

i knew id bring the positive out of you

fermi vs 5970
512 vs 3200
384bit vs 2x256bit

512 fermi shaders beat 3200 radeon shaders.. just as the actual 384bit mem bandwidth beat 2x256bit actual mem bandwidth and thats even if they stick to 4.2gbps

so much for radeon shaders hyped as "more efficient"

Let's play what if since you guys are so fond of these mind games

@Sam oslo...
and what if nVIDIA releases a dual GF100 in April ?

mind games.. oh that 400GB/s (5gbps) fermi ??

[zerazax]

i knew id bring the positive out of you

fermi vs 5970
512 vs 3200
384bit vs 2x256bit

512 fermi shaders beat 3200 radeon shaders.. just as the actual 384bit mem bandwidth beat 2x256bit actual mem bandwidth and thats even if they stick to 4.2gbps

so much for radeon shaders hyped as "more efficient"

You do realize that the "3200" shaders are just how ATI counts it (marketing speak) right? You've got to divide that by 5 to get the equivalent count for Nvidia... so 640 to 512, and that's not counting the fact that Nvidia has a hot clock meaning the shaders on the Nvidia part are overall more, equivalently

As for your bandwidth talk, that's hilarious seeing as how you once were championing 512-bit as a necessity, when 256-bit + GDDR5 did just fine when it came to the RV770 vs. GTX285

Slow down on the kool-aid thar

lol well my guesstimate is 400-500$

though still keep expectations low/leave room for surprise



spill what.. my glass is empty

I don't see Fermi selling for 400-500$ at all unless it's performance really is under the 5970's by a decent margin. No way has Nvidia EVER sold a top end card without a premium. See: GTX 280 vs. 260 prices at launch.

And I'm pretty sure people have gotten in trouble for claiming to know stuff on this forum without substantiating it

[NapalmV5]

You have to realize that both architectures are very different, you can't compare ATI's shaders to Nvidia's Cuda Cores on a one-by-one basis.
It's similar to the old Athlon VS Pentium 4 debate.
It's not about quantity of working units, but how they work, and ultimately how good performance is.

And BTW, you are supposing that fermi will have the upper hand on said comparison, but most people don't share that thought.
Think about this: Why did Nvidia compare GF100 vs the 5870? (instead of the 5970). That should give you an idea.



Considering what Nvidia said about XFX, I wouldn't expect them to be on the first batch of fermi cards.

hmmm and i thought you had some inside info seems you just based your claims on your flawed logic. shader comparison is funny though if you divide atis shaders to 5 that means atis 160 lower clocked shaders matched highly clocked 240 shaders and unless fermis shaders are some sort of miracle its really hard to believe that only 512 shaders can match 3200. Fermi has few undeniable advantages though higher 'memory and bandwith' and being single gpu are important ones

You do realize that the "3200" shaders are just how ATI counts it (marketing speak) right? You've got to divide that by 5 to get the equivalent count for Nvidia... so 640 to 512, and that's not counting the fact that Nvidia has a hot clock meaning the shaders on the Nvidia part are overall more, equivalently

As for your bandwidth talk, that's hilarious seeing as how you once were championing 512-bit as a necessity, when 256-bit + GDDR5 did just fine when it came to the RV770 vs. GTX285

Slow down on the kool-aid thar



I don't see Fermi selling for 400-500$ at all unless it's performance really is under the 5970's by a decent margin. No way has Nvidia EVER sold a top end card without a premium. See: GTX 280 vs. 260 prices at launch.

And I'm pretty sure people have gotten in trouble for claiming to know stuff on this forum without substantiating it

maybe you guys need to "Slow down on the kool-aid thar" may i suggest some guinness to balance you guys out a bit ??

let me put it this way.. transistor per transistor nvidia has higher efficiency/performance than ati whether you agree or not like it or not.. you guys can turn blue it aint gonna change

oh btw top 512 fermi could go up against a single 3200sp radeon.. thats why it cuts thru 5970 like knife thru butter

[FischOderAal]

oh btw top 512 fermi could go up against a single 3200sp radeon.. thats why it cuts thru 5970 like knife thru butter

You either like to play with fire or you're 100 percent sure about that, like in "I hold it in my hands" sure.

There've been others who were very sure about their claims (can't blame them if they just said what sources told them) like the whole "Fermi will be released in november/december" thingy.

[NapalmV5]

You either like to play with fire or you're 100 percent sure about that, like in "I hold it in my hands" sure.

There've been others who were very sure about their claims (can't blame them if they just said what sources told them) like the whole "Fermi will be released in november/december" thingy.

am i not napalm ??

[zerazax]

Neliz also called out 64 TMU's in this own thread when everyone automatically assumed 128

maybe you guys need to "Slow down on the kool-aid thar" may i suggest some guinness to balance you guys out a bit ??

let me put it this way.. transistor per transistor nvidia has higher efficiency/performance than ati whether you agree or not like it or not.. you guys can turn blue it aint gonna change

oh btw top 512 fermi could go up against a single 3200sp radeon.. thats why it cuts thru 5970 like knife thru butter

For someone who admitted that they have no actual clue about the real performance in this thread, this post is gold

[Chumbucket843]

You do realize that the "3200" shaders are just how ATI counts it (marketing speak) right? You've got to divide that by 5 to get the equivalent count for Nvidia... so 640 to 512, and that's not counting the fact that Nvidia has a hot clock meaning the shaders on the Nvidia part are overall more, equivalently

you dont divide anything by 5. they count their alu's the same way as nvidia does. games today barely need this much shading relative to other parts of graphics.

[eric66]

i knew id bring the positive out of you

fermi vs 5970
512 vs 3200
384bit vs 2x256bit

512 fermi shaders beat 3200 radeon shaders.. just as the actual 384bit mem bandwidth beat 2x256bit actual mem bandwidth and thats even if they stick to 4.2gbps

so much for radeon shaders hyped as "more efficient"



mind games.. oh that 400GB/s (5gbps) fermi ??

hmmm and i thought you had some inside info seems you just based your claims on your flawed logic. shader comparison is funny though if you divide atis shaders to 5 that means atis 160 lower clocked shaders matched highly clocked 240 shaders and unless fermis shaders are some sort of miracle its really hard to believe that only 512 shaders can match 3200. Fermi has few undeniable advantages though higher 'memory and bandwith' and being single gpu are important ones