Although the performance of NVIDIA's GeForce4 Ti 4600 is tempting enough to justify its $320+ street price tag, the market has eagerly been awaiting the release of the Ti 4200. We already showed the incredible performance of the Ti 4200 in our Sub-$200 Video Card Roundup and its stellar act has granted it the title of the best video card under $200. Only one problem remained, the card we reviewed was a NVIDIA reference board and the GPU wasn't shipping to card vendors at the time of publication.
Now that we're half-way into the month of May the first board vendors are shipping their cards. Enthusiastic users have noted the appearance of the Ti 4200 on search engines such as Price Watch as an indication that the cards were finally here, and indeed they are. We received our first retail GeForce4 Ti 4200 card late last week from Gainward; sold under the GeForce4 PowerPack! Ultra/650XP name, the Ti 4200 from Gainward is one of the first to hit the online vendor market.
Before we actually head into the review itself there's the on-going issue with Gainward's naming of their graphics cards. As we mentioned in our recent GeForce4 Ti 4400/4600 Roundup, Gainward has deliberately stuck to a misleading naming system for all of their NVIDIA based graphics cards. Under this naming system all GeForce4 MX and GeForce4 cards are referred to under the GeForce4 PowerPack! name. The rest of the card's name is determined by whether it is a MX or a regular GF4; in the case of an MX the card is referred to as a "Pro" otherwise it is an "Ultra." The number following the Pro or Ultra is only related to the overclocked clock speed the card ships at provided that you use Gainward's drivers, if you use NVIDIA's reference drivers your card will not run any faster than stock (you can use a separate utility to overclock it again).
I like 43.45 - 45.23. That range works well with a lot of games. Those drivers are from 2003 so will be solid for just about any game you'd want to run on a GeForce 4. Newer drivers break some older games. The 50 series was about GeForce FX Direct3D 9 level optimizations and speed cheats that impact quality. 60 series brought in GeForce 6 support. Etc.
64Mb Ti4200s are the cheapest "full" GeForce4s you can buy. The various models of GeForce4 MX are cheaper still, but they're a GeForce4 in name only; they don't have a full DirectX 8 feature set, and so are really more of a high-clocked GeForce2-and-a-half.
In the real world, the restricted feature set of the GeForce4 MX doesn't actually matter much, and they're undeniably good value for money. But if you want a proper GeForce4, the 64Mb Ti4200 is the entry level card.
The 64Mb Triplex Ti4200 doesn't have the over-the-top presentation of their Ti4400 and Ti4600 cards; it doesn't have a swoopy cooler on the main chip, and it comes in an ordinary cardboard box, not a preposterous little aluminium briefcase with a window. It's still got Triplex's signature silver circuit board, though, and so it's quite cool looking, if that sort of thing matters to you.
There's also a driver disc containing Triplex-branded drivers for their whole Nvidia-chipset card range, and a selection of not terribly compelling utilities. You can safely leave this disc in its plastic sleeve, and just install the Nvidia reference drivers instead. That's what I did.
Along with the software, you get an RCA composite video cable for connecting the card to a TV, VCR or other composite device, and a little adapter lead that lets you plug the RCA cable into the S-Video TV out connector on the back of the card. There's no S-Video lead, and there's no adapter for theDigital Visual Interface (DVI) connector on the back of the card, either. So you can plug a normal VGA-connector monitor into the 14 pin socket on the back of the card, but if you want to do dual monitor computing (which the Ti4200 supports, just like the Ti4400 and Ti4600), you'll either need your second monitor to be a flat screen with a DVI connector, or you'll have to shell out for a DVI-I to VGA adapter so you can connect another analogue-cable screen.
The normal stock speed for 64Mb Ti4200 cards is 250MHz core, 500MHz RAM. 128Mb Ti4200s get slower RAM, and run at 250/444MHz. Ti4400s are meant to be 275/550MHz, and Ti4600s are meant to be 300/650MHz. So, at stock speed, the fastest full GeForce4 is only 20% and 30% higher clocked than the slowest one.
The Triplex 64Mb Ti4200 is clocked at 250/550MHz by default, giving it almost Ti4400 performance out of the box. The higher RAM speed's accounted for by the non-standard memory chips Triplex have put on the card; they're rated at 3.3 nanoseconds (ns), which means their specified ceiling speed before DDR doubling is 300MHz. After doubling, 600MHz.
I'm happy to say, though, that this card doesn't have any such problems. Well, the one I got for review doesn't, anyway. Its memory was happy at 625MHz, and I managed to wind its core up to 315MHz, which is as far as the standard Nvidia clock speed slider will go. You can push further if you use third party utilities, but core speed has a small enough influence on performance that there's not a lot of point bothering. 315MHz from a Ti4200 is a perfectly good result.
The 3.3ns RAM on this card doesn't seem to give it a significant advantage over normal 64Mb Ti4200s.Anand managed to get a couple of 64Mb Ti4200 boards with the standard RAM to 600 and 610MHz; 625 isn't enough faster that you should care.
Fully overclocked, this cheapo 64Mb Ti4200 is faster than a 128Mb Ti4600 running at stock speed. Well, it is unless you need more than 64Mb of video memory, anyway. Run out of graphics card memory and you'll be using AGP texturing, which'll cause your 3D frame rate to plunge.
The Leadtek Ti4400 defaults to 275/552MHz; its reliable ceiling seems to be about 285/680MHz. At that speed in the default 3DMark2001 SE benchmark, it manages 9035 3DMarks, a trivial amount below the score of the overclocked Triplex Ti4200. In higher resolutions, the Leadtek card's faster RAM would outweigh its slower core and let it get ahead by a little, but the clock speed differences are less than 7% for core and less than 9% for RAM, so there's never going to be enough between the two for it to matter. Newer Leadteks may have more overclockable cores - the cooling system certainly isn't the limiting factor - and end up ahead of the Triplex Ti4200s across the board. But the difference still isn't going to be enough to justify the extra money.
What all this means is that 64Mb Ti4200s are a good deal in general, and the Triplex 64Mb Ti4200 is a good deal in particular. If you must have 128Mb of video memory then the somewhat more expensive 128Mb Triplex Ti4200 will suit you; its RAM is slower, but it really won't make enough difference to matter to anybody who's not pathologically picky.
I've got a GeForce 4 Ti 4200 graphics card with AGP 8x, 128 MB of memory, and I feel like my score in 3DMark 2001 SE is a little bit low. I've seen reviews from back in the day with people getting around 12500 points, and my barely gets 10000 with 53.06 Forceware. Clocks are all correct, the cooling is fine, capacitors are brand new, all parameters seem to match. What could explain that perofrmance drop?
True, CPU is a thing when it comes to benchmarks. Main problem back in the days was, that both nv and ATI tuned their drivers to perform best in synthetic benchmarks aka 3D-Mark xxxxx . Maybe you'll reach a higher score with a lower driver version. 55.xx seems very high for a GF4, propably nv removed the "cheats" in the code...
I tested it on different installs of Windows XP and got similar results so that's not the culprit. My CPU is an Athlon XP 2600+ at 2.1 GHz. Gonna check older drivers, might be the problem because I remember that with 93.71 I barely got 9000. I'll post an update once I check other driver versions.
Alright, I got better results with older drivers. Detonator 45.23 gave me around 10800, and 43.45 hit about 11000. Might drop even lower, like 28.32 should probably do 12000, unless I hit a CPU bottleneck.
This is with 56.73 drivers which come "stock" with Windows XP SP3. No overclocking was done.I prefer not to use older drivers that cheat to make results comparable with other (usually newer) cards from both Nvidia and ATI.
Pretty low score - the Athlon is starving on low memory bandwith. If only my SB-Live would work under plain dos, but it's a known limitation with newer boards. SB-Live and DOS are no problem with an Abit KT7A. Correct drivers are needed.
The following set of Nvidia GPUs are no longer supported in Debian Jessie, MX-15 or antiX-15 by the 173xx series proprietary nvidia driver (or any other series of the proprietary nvidia driver). They are still supported by the nouveau and vesa drivers.
The following set of Nvidia GPUs are no longer supported in Debian Jessie, MX-15 or antiX-15 by the 96xx series proprietary nvidia driver (or any other series of the proprietary nvidia driver). They are still supported by the nouveau and vesa drivers.
NVIDIA chip name Device PCI IDGeForce2 MX/MX 400 0x0110GeForce2 MX 100/200 0x0111GeForce2 Go 0x0112Quadro2 MXR/EX/Go 0x0113GeForce4 MX 460 0x0170GeForce4 MX 440 0x0171GeForce4 MX 420 0x0172GeForce4 MX 440-SE 0x0173GeForce4 440 Go 0x0174GeForce4 420 Go 0x0175GeForce4 420 Go 32M 0x0176GeForce4 460 Go 0x0177Quadro4 550 XGL 0x0178GeForce4 440 Go 64M 0x0179Quadro NVS 0x017AQuadro4 500 GoGL 0x017CGeForce4 410 Go 16M 0x017DGeForce4 MX 440 with AGP8X 0x0181GeForce4 MX 440SE with AGP8X 0x0182GeForce4 MX 420 with AGP8X 0x0183GeForce4 MX 4000 0x0185Quadro4 580 XGL 0x0188Quadro NVS 280 SD 0x018AQuadro4 380 XGL 0x018BQuadro NVS 50 PCI 0x018CGeForce2 Integrated GPU 0x01A0GeForce4 MX Integrated GPU 0x01F0GeForce3 0x0200GeForce3 Ti 200 0x0201GeForce3 Ti 500 0x0202Quadro DCC 0x0203GeForce4 Ti 4600 0x0250GeForce4 Ti 4400 0x0251GeForce4 Ti 4200 0x0253Quadro4 900 XGL 0x0258Quadro4 750 XGL 0x0259Quadro4 700 XGL 0x025BGeForce4 Ti 4800 0x0280GeForce4 Ti 4200 with AGP8X 0x0281GeForce4 Ti 4800 SE 0x0282GeForce4 4200 Go 0x0286Quadro4 980 XGL 0x0288Quadro4 780 XGL 0x0289Quadro4 700 GoGL 0x028C 2b1af7f3a8