I’m sure y’all remember the GTX 970 controversy, where the card’s bandwidth would take a hit if the VRAM consumption crossed 3.5 GB. This was because to access the last 500 MB of VRAM, the card was dependent upon the same ROPs as the ones meant for the initial 3.5 GB. Something similar, albeit less serious has been discovered in case of the RTX 2070. Turns out that the card has two versions, one of which relies on a relatively poor silicon lot.
While the ramifications aren’t nearly as serious, it is still worth taking note of. The main impact is on the card’s operating frequencies and the overclocking potential. The higher quality SKU is named the TU106-400A while the standard version is TU106-400. NVIDIA is selling the TU106-400A for a higher price to board partners for use in special versions like the “Gaming OC” or “XC Ultra”. These come with factory overclocks of around 100+ MHz, and the performance is usually 10-15% better than the stock cards. As for the catch, these higher quality cards are usually $50-75 more expensive.
The silicon quality isn’t the only differentiating factor. This higher end SKU usually utilizes the same SKU as the RTX 2080 as well as better VRM and thermals. This allows them to benefit from increased voltages and TDP which is what eventually also converts into higher operating clocks.
Now, whether the extra 50-70 bucks make it worthwhile, that is highly debatable. The value for money score of the RTX cards is after all not so shiny. For a better idea of the performance deltas between the two cards, have a look at the benchmarks below:
The EVGA RTX 2070 is the stock version while the EVGA RTX 2070 XC Ultra is the special SKU. As you can see, the XC Ultra on an average is approx 10% faster than the stock 2070. Even after overclocking the cards, the XC comes out on top. In some cases, the difference is of more than 5 FPS which is at least noticeable.
This is probably nothing new. All the “gaming edition” or factory overclocked cards have been known to boost past the stock frequencies, but the fact that the SKUs are now being named differently, based on merely the binning process is glaring. As for whether the extra 10% performance is worth the extra $70- I’d say NO,just NO. The RTX cards are already selling past the MRP prices and paying another $70 on top of that is far from justified. You’d be better off buying a 10-series card, preferably a 1080 or a 1080 Ti if you can afford it.
- AMD Threadripper 24 Core 2970WX, 12 Core 2920X Benchmark Roundup
- $400 AMD Radeon RX Vega 56 Beats $600 NVIDIA RTX 2070 Using Registry Mod
- 2nd Gen AMD Epyc Processors Could Feature 64 Cores
Benchmark credits: Gamersnexus