There were two upsets in the graphics card sector this year: at the beginning of the year, Nvidia launched the 6 GB version of the Geforce RTX 3050 on the market. However, the cards were not only cut in terms of memory expansion, but also in terms of the entire architecture – in the end, around 20-30% gaming performance was lacking compared to the 8 GB versions. In August, things continued with the Geforce RTX 4070, which is now also available with GDDR6 instead of GDDR6X memory. Although the difference in speed is only marginal, the differences between the two variants are not obvious at first glance.
We can actually still count ourselves lucky these days, because around 20 years ago, graphics cards were much more likely to be fiddled, cheated and lied about. For this article, I have collected two particularly brazen attempts, from which not even the buyers of high-end graphics cards have been spared.
I’ll start with the Radeon 9800 Pro from AMD’s predecessor ATI, which changed hands for just over 400 euros when it was released in the first half of 2003, gave the competition from Nvidia a run for their money at the time and has gone down in hardware history as an icon. On Nvidia’s side, the Geforce 6800 is used, which came onto the market around a year after the Radeon 9800 series and, in contrast to the mediocre previous generation Geforce FX, was generally convincing. With a starting price of around 300 euros, the Geforce 6800 represented the entry into the upper class, as the top models 6800 GT and 6800 Ultra were positioned even higher with a fully unlocked NV40 graphics chip and GDDR3 graphics memory.
Incidentally, the two cards are not directly comparable, as the Geforce 6800 competed against the X800 series from ATI, which in turn represents the next generation of the Radeon 9000 series. One of the reasons for this is the more modern feature set of the Geforce 6800, which already supports DirectX 9.0c with Shader Model 3, while the Radeon 9800 still has to make do with Shader Model 2.0. In addition, the Geforce 6 series was already available for the then newly introduced PCI Express slot, while the Radeon 9000 series was only available for the AGP slot.
Originals vs. cheat packs
| Radeon 9800 Pro “Cheat pack” | Radeon 9800 Pro “Original” | Geforce 6800 “Cheat pack” | Geforce 6800 “Original” | |
| GPU | R350 | R350 | NV40 | NV40 |
| Pixel/vertex shader | 8 / 4 | 8 / 4 | 8 / 3 | 12 / 5 |
| TMU / ROP | 8 / 8 | 8 / 8 | 12 / 12 | 12 / 12 |
| Graphics memory | 256MB DDR | 128MB DDR | 512MB DDR2 | 128MB DDR |
| Interface | 128 bit | 256 bit | 128 bit | 256 bit |
| Clock rate chip | 378 MHz | 378 MHz | 325 MHz | 325 MHz |
| Clock rate memory (bandwidth) | 250 MHz (8.0GB/s) | 338 MHz (21.6GB/s) | 300 MHz (9.6GB/s) | 350 MHz (22.4GB/s) |
Curious: While a sufficient amount of graphics memory seems to be in short supply in today’s generation of graphics cards, there is more than enough of it on the cheat packs – in the Geforce 6800 even four times as much as the original. However, savings have been made in the crucial areas, as a halved memory interface or reduced clock rates have an enormous impact on gaming performance. In the Geforce cheat pack, the graphics chip was even cut by additional pixel and vertex shaders.
The tactics of the cheat packs are clear to see: A large amount of memory was intended to lure inexperienced buyers 20 years ago; after all, the misconception that “more memory = faster graphics card” was still widespread. Many buyers will probably never have noticed that memory bandwidths were cut or clock rates reduced. Incidentally, the blue PCB of the Geforce 6800 cheat pack represents the ultimate recycling of leftovers and was manufactured in calendar week 46 of 2008 (red arrow) – over four years after the graphics chip was released. The Geforce GTX 280 was already available at this time. The free areas of the graphics memory, which explain the halved interface, are also clearly visible.
There are no free areas to be seen on the Radeon cheat pack, and overall the model manufactured by Etonda is quite close to the reference design from a visual point of view. However, the halved interface can also be seen here in detail, as the graphics memory is installed in a straight line on the PCB and not in an “L” shape as in the original.
48 Antworten
Kommentar
Lade neue Kommentare
Veteran
Urgestein
1
Mitglied
Urgestein
Moderator
Neuling
Urgestein
Veteran
Veteran
Veteran
1
Mitglied
1
Urgestein
Urgestein
Urgestein
Veteran
Veteran
Alle Kommentare lesen unter igor´sLAB Community →