Sapphire HD 4650 OC Edition

[W1zzard]

To read this review go to: http://www.techpowerup.com/reviews/Sapphire/HD_4650_OC/

Please Digg the review to help us promote it.

[eidairaman1]

Primarily a Replacement to a Dead IGP.

[mdm-adph]

Originally Posted by eidairaman1

Primarily a Replacement to a Dead IGP.

Really? For $70, that card's pretty damn fast -- as fast as an HD3870 when overclocked, I'd say that's a bit better than an IGP.

[newtekie1]

Originally Posted by mdm-adph

Really? For $70, that card's pretty damn fast -- as fast as an HD3870 when overclocked, I'd say that's a bit better than an IGP.

No, it doesn't even match an HD3850 when overclocked, forget an HD3870. I don't know where you got that from.

[wahdangun]

Originally Posted by newtekie1

No, it doesn't even match an HD3850 when overclocked, forget an HD3870. I don't know where you got that from.

yes, hes right look at the graph below :
http://i4.techpowerup.com/reviews/Sa...es/perf_oc.gif

it's way faster than hd 3850, and yes it's was also faster than 3870.

this litle card realy rock

[newtekie1]

A single test in COD4 is hardly evidence that it outperforms the HD3870 once overclocked.

http://i1.techpowerup.com/reviews/Sa..._1920_1200.gif

The non-overclocked results for this card in COD4 are already close to the HD3870 and have it beating the HD3850.

But when we look at overall:

http://i5.techpowerup.com/reviews/Sa...es/perfrel.gif

The HD4650 isn't even close to the HD3850.

So, yes, if you look at one single test(COD4) the card at stock speeds is already faster than an HD3850, and almost matches an HD3870, and overclocked it is faster than the HD3870. But personally, I play more than just one game.

That is why reviews test more than just one game.

Overall, even overclocked, this card isn't going to be faster than an HD3870.

[mdm-adph]

Originally Posted by newtekie1

A single test in COD4 is hardly evidence that it outperforms the HD3870 once overclocked.

http://i1.techpowerup.com/reviews/Sa..._1920_1200.gif

The non-overclocked results for this card in COD4 are already close to the HD3870 and have it beating the HD3850.

But when we look at overall:

http://i5.techpowerup.com/reviews/Sa...es/perfrel.gif

The HD4650 isn't even close to the HD3850.

So, yes, if you look at one single test(COD4) the card at stock speeds is already faster than an HD3850, and almost matches an HD3870, and overclocked it is faster than the HD3870. But personally, I play more than just one game.

That is why reviews test more than just one game.

Overall, even overclocked, this card isn't going to be faster than an HD3870.

Well, who knows what the overall chart would look like if the overclocked speeds (for all games) were factored in. I'd still think it would at least be on par with the 3850/3870.

Either way, the whole point was that the card is far more than just a "replacement for an IGP," which I'm sure you'll agree with.

[W1zzard]

Originally Posted by newtekie1

A single test in COD4 is hardly evidence that it outperforms the HD3870 once overclocked.

http://i1.techpowerup.com/reviews/Sa..._1920_1200.gif

The non-overclocked results for this card in COD4 are already close to the HD3870 and have it beating the HD3850.

But when we look at overall:

http://i5.techpowerup.com/reviews/Sa...es/perfrel.gif

The HD4650 isn't even close to the HD3850.

So, yes, if you look at one single test(COD4) the card at stock speeds is already faster than an HD3850, and almost matches an HD3870, and overclocked it is faster than the HD3870. But personally, I play more than just one game.

That is why reviews test more than just one game.

Overall, even overclocked, this card isn't going to be faster than an HD3870.

so which test would you suggest to assess overclocked performance in the review? only one test. i so knew this would happen, now i'm asking myself why even bother making another graph (like i didn't know before that this would happen)

[wahdangun]

hmmm then if that the case, i think it will be good idea if wizz tes the card again after overclock and include it int he graph just like other site

[wahdangun]

sorry : i mean include it in all graph not just one graph, so everyone will know the result after and before overclokingn,

[mdm-adph]

Originally Posted by W1zzard

so which test would you suggest to assess overclocked performance in the review? only one test. i so knew this would happen, now i'm asking myself why even bother making another graph (like i didn't know before that this would happen)

Aye -- I knew when I saw that chart the trouble you went through in picking a chart to show the overclocked speeds at.

I guess you could try and just show a 3dMark score -- if that's not enough for most people, then just go back to the way you had it before and not show any graph at all for the overclocked speeds.

[newtekie1]

Originally Posted by W1zzard

so which test would you suggest to assess overclocked performance in the review? only one test. i so knew this would happen, now i'm asking myself why even bother making another graph (like i didn't know before that this would happen)

I have no problem with the way it is currently done. There isn't a single test that will provide a good judge of overall performance, the only way to do that would be to retest the card on all the tests run, but I'm sure that would take way to much time and I certanly don't expect you to do so.

Retesting the card on COD4 when overclocked is fine, and gives a good idea of the performance boost overclocking will give you, but people are taking it wrong. It isn't a good way to judge the overall card performance compared to the other cards. Just like you wouldn't judge a cards performance one just a single test, you shouldn't judge the overclocked performance on just one test either.

Just as an example, if I picked Quake4(one of the test that the HD4650 did bad in) then the overclocked results would probably show that the HD4650 doesn't even outperfrom the HD3850, forget about the HD4870. So, I'm not really concerned with the way it is done now, I'm just concerned with the way people are taking it and using it to assume overall performance.

Originally Posted by mdm-adph

Well, who knows what the overall chart would look like if the overclocked speeds (for all games) were factored in. I'd still think it would at least be on par with the 3850/3870.

Either way, the whole point was that the card is far more than just a "replacement for an IGP," which I'm sure you'll agree with.

I doubt it will be on par with the HD3850, certainly not the HD3870. If you look at the stock results, COD4 was pretty much the best the HD4650 did in terms of performance compared to the HD3850/70.

I certainly agree, it is more than just a replacement for an IGP.

[christof139]

Being cheap, it might have been better to get two 4650's instead of the two 4670's I have for XFire. However, there weren't any reviews out on the 4650's and this is the first review i have seen. I guess some reviewing places are being cheap on the number of reviews they do.

This was a good review, and I am guessing that 2 4650's in XFire would perform with nearly the same scaling/ration increase as two 4670's.

With decent and not so decent cheaper low and midrange cards it is always nice to see how they perform in Xfire or SLI.

Another advantage of ATI is that their cards will play older games such as METW 1 etc. whereas the NVIDIA 8000 series and up cards won't. So, people like me need to know this and why ATI or anyone else for that matter never mentions it is not good and it is a good thing and a good selling point for a gfx card to be able to run older games and/or programs. This should be advertised.

Chris

[Wai_Wai]

Sorry I would like to raise one issue.
Try to compare "HD 4650" with "HD 4670". It gives readers a feeling that HD 4650 is really great, or is actually way too good for that amount of price. The overall performance is way too close, ie 11% only. FPS difference can be as close as <4 FPS in some tests.

However:
http://www.pcgameshardware.com/aid,6...=659355&page=9
HD 4650 performance is only 57% of HD 4670. The other results are similar. The gap is obvious.

http://www.bjorn3d.com/read.php?cID=1411&pageID=5798
~63% on average.

http://xtreview.com/addcomment-id-62...ce-review.html
~70% on average.

Based on your result (the gap being 11% only), HD 4650 is better price/performance-wise.
Based on others (the gap being 30-40%), HD 4670 is better price/performance-wise.

I believe some readers who have read those reviews will have such confusions.
So would you explain why it makes such a big difference?
(I probably know why. But I want to ask the author of this review first)