Well based upon that review it's nice that Abit didn't cheap out on the RAM as they used 3.6ns. Unfortunately the core didn't get any higher than any other 4200 card and the RAM should have hit 600mhz at least, most 4.0ns as used on 99% of std 4200's hit 540-560mhz. Considering the added noise level I don't think it would be worth it, even if it was the same price as any other 4200. As for Gainwards 'Golden Sample' ... this was great in the GF3 days, but their 4200 cards use the same RAM as almost every other 4200 manu and as such doesn't o/c any better. As for the core, all GF4TI cards tend to hit 280-320, all the cooling methods pretty much even out so once again the Golden label is more 'Run Of The Mill' IMHO. Gainward GF3 cards had image quality a cut above the rest, all GF4TI cards are almost identical so once again it stands for nothing. 4200-128MB cards have std clocks of 250/444 and 99% use 4.0ns RAM this should hit about 300/550, 4.5ns will hardly o/c at all. 4200-64MB cards have std clocks of 250/500 and 99% use 3.6ns RAM which should hit about 300/600. However, the 64MB isn't very wise when you consider that the 13% higher RAM clock should which ahould boost perf almost 10% only gives the 4200-64MB cards a 2-3% advantage. Once again when both types of 4200 are o/c'ed the 9% faster RAM clock of the 64MB cards isn't going to give noticably better perf BUT if a game uses more than 64MB the 64MB card will be hit VERY hard no matter what the speed. When more than 64MB is needed expect a 4200-64MB @ 300/600 to give perf about equal to a 4200-128MB @ 250/444. Well worth paying the extra $20 for 128MB and you'll have a much easier time selling any 128MB card. The 4200 cards known to cheap out on the RAM which can effect stability, std clock speed and will definitely hinder o/c are; AOpen 4200-128MB, EVGA 4200-64MB and from what's been said above Gainward's non-Golden Sample 4200 cards. Stay well away! The ns rating of the RAM is a very good guide to how well it should o/c, but if it's anything other than Hynix or Samsung take it with a pinch of salt. I've never heard of a Gainward 4200-128MB using 3.5ns RAM, but I have heard of a few less than reputable suppliers that's for sure! I don't for one second believe Gainward hand-pick their chips, and I haven't heard or fread anything to suggest Gainward o/c any better than any other 4200 card. If Gainward really gave a crap about their 'Golden Sample' tag then they'd have spent the extra $4 and stuck 3.6ns on their 4200-128MB rather than cheaping out on their std 4200 to make their GS 4200 look better. If you want 4600 clocks you're going to have to pay for it. Look out for the enhanced 4200 cards using the longer 4400/4600 board design incorporating 3.3ns BGA RAM (all of which adds to the price) as these o/c to 4600 speeds and often surpass it o/c'ing as well as a 4600 o/c's! These cards are the Asus Deluxe, Suma Special Edition and Albatron P Turbo. I'd advise you get a std GF4TI4200 not from the manus listed previously for $140ish or else buy an enhanced 4200 or true 4400 for $180ish or get the top 4600 for about $210. So the diff types of GF4TI4200 cards break down 3 ways: You will always get a variance with cards of the same type and manu, but even so as an accurate guide the above is pretty good. Exceptions are (as pointed out previously) AOpen 4200-128, EVGA 4200-64 and Gainward non-GS 4200 as these all decided to save $4 on the manu'ing process which is NOT a good thing at all. So the things left uncovered which vary between the diff manus are cooling solutions which all pretty much even out, RAM HS which are almost completely pointless, colour, packaging, sw bundle, inclusion of DVI-to-CRT adapter, TVout vs VIVO and of course price! So essentially so long as you avoid those cheap-out manus any 4200 of the same type is almost identical to the next, decide on the RAM size and whether you are willing to pay for the 'enhanced 4200' cards and then see what features you want and then let price be the key factor.
Originally posted by: AnAndAustin
$120-140 Std 4200-64MB 250/500 3.6ns_o/c=300/600
$140-160 Std 4200-128MB 250/444 4.0ns_o/c=300/550
$180-200 Enh 4200-128MB 270/550 3.3nsBGA_o/c=300/700
MrDudeMan just because a Rad8500/9000PRO run at 275/550 (retail) doesn't mean they can out do a 4200 even at 250/444! In almost every game and benchmark the 4200 at standard clocks is the better card, once the 4200 cards are o/c'ed the Radeons (non 9700 LOL) are left in the dust, esp if you want to play with AA. Not only that but the Rad8500LE (250/500) is clocked 10% slower than the Rad9000PRO yet still beats it by 15%. Of course the exact diffs do cvary depending mostly upon settings and CPU used, but anything over an Athlon 1.4ghz and you're certainly better off with a 4200. Rad8500 cards are very good, and for under $100 you should find a 128MB version which will play all modern games very well, but labeling it 'best bang for buck' is certainly stretching things a bit just like saying 8500 is on par with 4200! As for '2D' image quality that's a falicy, GF4TI cards are every bit as good as ATI Radeons, certainly up until 1600x1200x32 @ 75Hz but even then since most of us don't have superb 19" monitors we aren't likely to be doing much 2D (working and reading text) at that setting. Yes, TVout and DVD playback the Radeons are clearly better, however GF VIVO is VERY cost effective, the TVout can be tweaked and DVD playback is a very small issue on any CPU above 600mhz! Bearing in mind these bm's are at default clocks, Rad8500 and GF3TI500 hardly o/c at all while 4200 usually exceed 4400 perf. Anyway take a look: I run UT2K3 MAX EVERYTHING and no chop. Well it seems the Gainward 4200-128MB definitely uses 4.4 - 4.5ns limiting o/c to about 300/460 while almost every other 4200-128MB uses 4.0ns making 300/550 more than possible. Most 4200-64MB use 3.6ns and using that an o/c of 300/600 should be doable, EVGA 4200-64MB uses the incorrect lower clock of 250/440 and 4.0ns RAM meaning 300/550 type o/c. Standard clocks for the other GF4TI are; 4400=275/550 4600=300/650 and a 4200 at the same clock with the same RAM size gives the exact perf. Of course even at stock speeds both the 64MB and 128MB are excellent cards, but it is a shame when a manu effectively kills the potential of the card and even worse affects stability and clock speeds by saving $4 on the cost of the RAM chips. As for which to go for 128MB or 64MB, esp on 4200 cards where the clocks and RAM type often differ is a long running question. Both are fine cards that's for sure. The 4200-64MB come at 250/500 while the 4200-128MB come at 250/444, a 13% diff in RAM speed which should bring about 7% advantage for the 4200-64 actually turns out to be a 2-3% advantage, and a LOT less if more than 64MB becomes necessary. The same is true when a 4200-64 hits 300/600 to a 4200-128 at 300/550 and benchmarks have shown that when more than 64MB of RAM becomes necessary a 4200-64 at 300/600 drops to about the perf of 4200-128 at 250/444. The same stands true for the Rad8500LE-128MB vs the 10% faster clocked Rad8500-64MB, perf diff is 2-3% but the hit from lacking adequate RAM is very significant. Since we're dealing in small costs going for 128MB for all cards it does make sense, it will help to ensure games can be run without sacrificing details/settings and that when the card is ready to be sold on that you'll sell it quicker and for more if it is a 128MB card. It seems a 64MB card isn't going to hurt you significantly in all modern games inc UT2003 so anyone with a 64MB GF3, Rad8500 or GF4TI certainly don't need to upgrade, but considering the price diff, perf diff and long term perspective anyone buying a new card really should plump for 128MB. Some benchmarks:As AnandAustin stated (though not in direct response to your comment), the reason that the 128MB cards get better benchmarks in most applications is because they come clocked lower. However, if you look at this review that AnandAustin pointed to before, you'll see a decent performance gain in Codecreatures, and an incredible difference in performance in Jedi Knight II, which, by the way, is out now, not a future game that you have to prepare for.Originally posted by: wizdum
64MB gets better benchmarks than the 128MB and you don't really need 128MB anyway.
Regarding the 'cheap out' versions of the 4200 cards in general ... using 4.5ns which has a technical limit of 444mhz is cutting it too near the line and that is never a good idea for long term stability or lastability, the same goes for 4200-64 cards using 4.0ns as again 500mhz is the technical limit of the RAM and in the case of the EVGA the default RAM clocks are 14% lower at 250/440 instead of 250/500 thus hindering perf even without o/c'ing. Anybody buying a 4200 with no intention of o/c'ing a little really should consider the Radeon or GF3 alternatives. It makes sense to o/c the 4200 cards a little just like the GF3TI200 cards they are intentionally clocked below their true abilities in order to promote the sales of the faster more profitable cards. All cards of the same model type/version (eg GF4TI4200) even from the same manu will vary in o/c, add this to the fact that the review sites are very likely to get the 'best' rather than the 'random' sample that us punters get and it makes little sense to get too pedantic over precise o/c'ing numbers, when o/c'ed a 5-10% clock variance with the exact same model+manu card has never been uncommon. So 2 'identical' 4200 cards manu'd by BRAND-X will prob vary from 280/520 to 320/580 when o/c'ed, so I wouldn't take any review or individual user's experience as anything more than a rough guide to what the exact same card should do. Anyway ... Those are all priced very evenly, the 4200 cards using 4400/4600 design and 3.3ns BGA RAM come in at TI4400 pricing: There's no wrong or right, only opinions, experiences and advice! The core of 4200 seems very little to do with manu as all reviews seem to agree that cooling solutions don't make a much diff, heat isn't the prob it is just luck of the draw off the factory conveyor belt / nVidia. I don't believe any company are going to send cores back to nVidia, you really expect them to hook them up, test them at higher clock speeds and then attatch a note saying 'sorry but these cores don't o/c well enough so swap these for ones that do!' As for the RAM heat is never the issue and HS's are only decorative as once again they tend to all use the same spec RAM (4.0ns for 128MB and 3.6ns for 64MB) and either Hynix or Samsung, so again it is a luck of the draw and no manu is going to send the chips back, if 4.0ns can hit 500mhz and 3.6ns can hit 555mhz there is no cause for complaint as these are the technical limits! So brand matters very little for GF4TI except for the odd manu who skimp on the RAM to either save $4 or to make the 'super' version of the card look better than it really is! Just MHO In the order I gave: Then seperated and in ascending order: As I mentioned before there will be a natural variance even between exactly the same cards from the same manu and reviewers are more likely to get 'friendlier' cards than the average punter too. Anyway the Gainward cards certainly differ, the 4200-128 is near the bottom while the 4200-64 is near the top, even so there's only about 10% diff between the absolute best o/c'ers and the absolute worst which isn't far away from how you'd expect cards from the same manu to vary ... so I'd say price is one of the key factors in the decision, certainly more so than brand. I think it's clear that Gainward's 'Golden Sample' labelling is certainly more of a marketing gimmick now unlike in the GF3TI200 days it actually meant something.Originally posted by: Yield
I just got an MSI 128MB Ti4200 card and it runs rock at the end of the slider with coolbits. I'll grab RivaTuner and give it some more, but as it sits right now, it's at 315/560 maxed right to the end.. doesn't seem to have a single problem.. no exotic cooling, just what it came with. I love this card.
PCMarine, it's worth finding the limits of you card, that way you know how fast you can go. Best to back off a few notches from the absolute max in terms of not overly stressing/ageing the card. You may find 275/500 could actually be past the limit, or else the limit could easily be 330/580. So I'd suggest you find the limit (point at which visual glitches appear) and then back off about 20-30 clocks as that should be VERY safe, who knows 300/550 could be a VERY conservative o/c for your card ... but at the same time 275/500 could be pushing too far as well LOL!
Facebook
Twitter