Update on 4890...

[4745454b]

Paper numbers are the numbers you'd see in a "normal" review. 25FPS at XXX by XXX resolution for example. Strange was talking about the way Hard reviews, where they talk about the experience. Sometimes they will mention that although X card was on paper faster, there is no real way to tell the difference between the two unless you could look in the case, or see the benchmark results.

Synthetics are fine as long as you are using the same hardware every where else in the rig and are just comparing the performance of one component.

My problem with synthetic benchies is that they don't model real world usage. Take the previously mentioned 2900XT. 3DMark said it should compete really well, but in actual games it could barely take on the 8800GS. Even the 8800GT was faster. Synthetic harddrive programs have the same issue. They model what the drive is theoretically capable of, but they fail to take into account real world usage. (not all clusters needing to be read are next to each other for example.) I've seen to many times where a synthetic benchmark says X should be faster, but real world programs say Y really is.

The only time when I use them is to stress test X device. If an overclocked video card can handle many loops of 3DMark (or a new to me used video card.) then I can say the overclock is stable.

Again, saying/guessing that X will be Y% faster then Z gives us an idea as to Xs relative performance. Reading Rescawen's post gives me an idea as to how fast s/he thinks it will be. I do not understand why Strange says its "shite"

 

[JAYDEEJOHN]

Guessing is one thing, but guessing using guesstimates is another. I trhink I understand SS here. But..... heheh, if you know the previous performance of an older card, and draw a % of perf of a newer cards perf vs the older one, then its not paper, even tho its all a guess. When ATI announced their 4xxx series, they gave a % as to how much better the new 4xxx series would be over the older 3xxx series, and at the time it was a paper guesstimate, it did come thru as to its ability, one gen vs another.,Going by this line of thought, you can use a % guesstimate, not using HW differences only, as in clock gen ,tmus etc, but also common sense as to what a companies done in the past as well as what is needed currently for people to pull the trigger for a better card, as in a certain % to justify a new purchase

 

[spathotan]

Stranger, the "30%" remark was 4870 vs GTX 285. The GTX 285 is allegedly like 15% better than the GTX 280. Its really simple math.

According to your charts, its a 20% difference in Fallout 3, 15% difference in Crysis, and a 25% difference in Stalker. That's close enough for some rough estimation done by some folks in this thread.

On a different note, one thing I do notice on those charts is buying a GTX 295 over a 4870X2 is a little questionable.

 

[mactronix]

May i suggest an addition to your Sig S/S

Bottlenecking is an evil word.
Generalizing is an evil practice

mactronix

 

[L1qu1d]

I still don't get it, this just seems like rambling to me.

Other than that, nothing. Clearly stated 30% when talking about the 285 GTX.

Then again I just don't care enough to read the post properly after % have no meaning.

This could completely change the p/p results then.

To me 10% isn't a epenis its basically insurance. For example that is

It would mean the the 20-30% the 4870 X2 has over the 280 GTX is meaningless which puts the 280 GTX ontop? And yes I mean 280 GTX not 285 GTX to clear that out.

Anyways I'm pretty finished with this particular thread since it did get off topic from what the 4890 will be, and apprerntly its going to be nothing since % don't matter.

 

[rawsteel]

GTX285 is about 20% faster than 4870 1GB on average (few exceptions on 2560x1600 4xAA, but also few exceptions where 4870 beats even the 285)

My prediction is that the 4890 will be at least 20% faster thus coming VERY close to 280/285 if not beat them. ATI might pull another bomb here.

P.S - GTX285 is exactly 8-10% faster than GTX280 as only diff is clock speed and the clock speed is about 10% faster (which doesnt always scale linearly with performance)

 

[L1qu1d]

yes your right, clock speed memory speed and shader speed, each varying from 8%-15%.

But do go on.

The 285 GTX right in the above charts shown by Strangestranger, shows the 285 GTX 25% faster, right in the first 1.

You won't see 20% from the 4890, not with 100 mhz increase and soo little SP added.

Strangestranger its not that my brain can't comprehend, its that your opinion shouldn't be treated as fact.

Though percetages mean nothing to you, doesn't means its not a good way to judge the brand, I mean it does sound alot more appealing right?

So this really isn't a matter of being right or wrong, its a matter of personal preference.

 

[Nica Guy]

according to those benchmarks the stock clocked 285 is about 18% faster than the 1GB 4870... i thought it was much more... i actually think an 18% increase is possible... I think the 4890 will perform just the same as a stock 285, if not it'll be a little less... probably like 5%...

From what i think, the 4890 will outperform the 280, which according to those benchmarks is 7-15% (10% in average) faster than the 1 GB 4870...

 

[Nica Guy]

I think even an OC 275 will beat an OC 280...

 

[JAYDEEJOHN]

My thoughts as well. How a small bump in clocks can be stretched to 15% , and 20% stretched to 30% is beyond me. A clock for clock increase isnt a guess, its accurate, and nothing scales 100% no matter what, and we dont have more than 10%. Im also not sure why some people are sooo confident about exactly what ATI is going to release here. I guess I need more links, as the best info Ive found has it bouncing anywheres from 30% better than the 4870 down to 12% better. If we take an average or in between those numbers, and there wont be a difference between the 285 and the 4890. All the 275 rumors etc, all the 295 unavailabilities, and only the 285 as a true common seller available, Im not so sure nVidia does have an answer to this one, and again, not sure why some people are insisting they know what ATI is releasing

 

[Nica Guy]

and to those saying 160sp and a 100mhz increase won't affect performance...
look at the performance difference on a 4850 and a 4830 (that's a 160sp difference)
and the 4850 is clocked approximately 100mhz higher on all clocks...

does that makes sense to you?

 

[Nica Guy]

yeah jaydee is right... we are all assuming we know for a fact all these new cards' specs. If those specs were like the ones mentioned, then that's what i think it would happen.

 

[spathotan]

and to those saying 160sp and a 100mhz increase won't affect performance...
look at the performance difference on a 4850 and a 4830 (that's a 160sp difference)
and the 4850 is clocked approximately 100mhz higher on all clocks...

does that makes sense to you?

An overclocked 4830 can match a 4850 in performance btw.

 

[invisik]

I think ati is just tricking us again the 4890 will have over 1000+ sps. xD
The card sit between the gtx280-gtx285 in my opinion.

 

[Nica Guy]

An overclocked 4830 can match a 4850 in performance btw.

cool ...
making my point even clearer on how clock speed increases can affect performance... add 160sp to that and
it'll probably increase those 20% in performance...

i agree with invisik... probably between those 2 cards at a much better price...

 

[spathotan]

The 4890 will be a great card no doubt. I think the price is more important than where it fits in line.

hmmm

 

[spathotan]

cool ...
making my point even clearer on how clock speed increases can affect performance... add 160sp to that and
it'll probably increase those 20% in performance...

i agree with invisik... probably between those 2 cards at a much better price...

Keep in mind that's a pretty high OCed 4830, the ones with good coolers can do it pretty easily. But youre still better off just buying a 4850.

While I think Nvidia is just flooding their lineup even more with the GTX 275, they are probably dead on with their apparent guesstimation as to where the 4890 will fall, between the GTX 260 and GTX 280, and im gonna have to go with that. If it comes in cheaper than the GTX 280 (which im sure it will) its a win.

 

[JAYDEEJOHN]

Exactly. Its the price that counts here. Theres tons of room for pricing with this card. Everyone involved will be winners here, and nVidia is going to suffer for it. All the partners, ATI and consumers, all win, except nVidia

 

[spathotan]

ATI is just really missing a STRONG partner, somebody that does their own development for drivers and software and even hardware, like EVGA does. EVGA has even released a SLi Enhancement Patch that adds SLi support to games that arent normally supported.

 

[JAYDEEJOHN]

I wouldnt bet on nVidia knowing anything here. Or why did they release their cards, the original G200s at such high prices? I mean, ATI had announced its pricing before NDA lifting. No, I cant say nVidia knows a thing of what ATI is doing here

 

[Nica Guy]

Keep in mind that's a pretty high OCed 4830, the ones with good coolers can do it pretty easily. But youre still better off just buying a 4850.

While I think Nvidia is just flooding their lineup even more with the GTX 275, they are probably dead on with their apparent guesstimation as to where the 4890 will fall, between the GTX 260 and GTX 280, and im gonna have to go with that. If it comes in cheaper than the GTX 280 (which im sure it will) its a win.

im not planning on buying a 4830, i already got a 4850... next buy probably something in the 5000 series or 4890 if its a good deal... my 4850 is still performing very well with new games...

 

[L1qu1d]

What I would like to see is a 4670 revision.

That card still impresses me with the amount of power it can bring to the table, and it doesn't even need a connector.

If ATI (doubtful we'll see it from Nvidia) made 1 that can match the 4850 or the 4870 (not likely anytime son), with out a power connector, I just found my gf's new card

Though that would really be looking at a new gpu and not a revision

 

[JAYDEEJOHN]

Even tho we saw some leaked specs and perf on the 740, its possible it may just be the card youve just described. The one G3D had was engineering sample, not completely best silicon,clocks, or drivers. Maybe better perf, less power yet from the real one

 

[L1qu1d]

thank god, I've tried the 4850 in her system, and that card ran hot as hell in there, but she has a Vaio that has no ventilation.

 

[4745454b]

^_____ Drills, saws, and tin snips can fix that...