It’s the fastest GTX 970 we’ve seen out of the box, with a 1,203MHz GPU core clock and 1,355MHz boost clock, as well as memory overclocked to 1,800MHz. In terms of customisations, Zotac’s GTX 970 AMP! Extreme edition goes even further than the MSI and Asus cards, replacing the reference cooler with a massive three-fan system. The Asus STRIX remains our favourite GTX 970 card, however, and AMD's top-end cards are still better value. As with the Asus STRIX GTX 970 card above, there’s enough headroom to increase the resolution to 3,840x2,160 in some games, and 2,560x1,440 in others - the MSI card is very slightly slower, but only by a frame or two. 115.5fps in Dirt Showdown is almost twice what we would deem playable, while 80.3fps in Tomb Raider and 46fps in Metro: Last Light were also excellent results. MSI’s GTX 970 Gaming 4G has a slightly less aggressive overclock, running at 1,140MHz and boosting to 1,279MHz. However, you can get similar performance from the XFX AMD Radeon R9 290 for around £40 less, or a quicker card for similar money with the Sapphire VAPOR-X R9 290X. If you're after an Nvidia GTX 970, the Asus STRIX is a quick model and is worth the extra fiver over MSI's GTX 970 for a modest performance increase.
The card managed around 32fps in Metro after we'd dropped detail levels to High.
Dirt Showdown was even playable at 3,840x2,160 with Ultra quality settings, with 48fps, and we saw a just-playable 29.9fps in Tomb Raider at this resolution once we swapped SSAA for FXAA. The card had no problem with Dirt Showdown or Tomb Raider once we increased our test resolution to 2,560x1,440, and we saw a smooth 49.7fps average in Metro once we disabled SSAA. This helped it score 116.8fps in Dirt Showdown, 77.8fps in Tomb Raider and 44.7fps in Metro: Last Light Redux at 1,920x1,080 with Ultra and Very High detail settings.Īll three games are easily playable at these frame rates. Asus has used the extra cooling headroom to overclock the Asus STRIX GTX 970 DirectCU II to 1,203MHz, and increase the boost clock to 1,355MHz. Both Asus and MSI have used twin-fan designs that are highly effective at cooling the GTX 970. CardĪs with the rest of Nvidia’s range, it’s next to impossible to buy a GTX 970 with a reference cooler, or one using the reference clock speeds. As with other cards that use the Maxwell architecture, the GTX 970 is surprisingly power-efficient, only drawing 145W when playing the most demanding games. The underlying GM204 GPU core has much in common with the GTX 980, except here it consumes less power as part of the chip is disabled. The 4GB of GDDR5 memory runs at 1,750MHz, operating on a 256-bit memory bus. The reference design has 1,664 CUDA cores running at 1,050MHz, boosting up to 1,178MHz when thermal limits allow. Although the final 512MB is much slower than the high-performance segment, it is still fast enough for the graphics card to use it rather than rely on swapping data with system RAM.ĭespite this specification stumble, however, the GTX 970 is still a seriously powerful GPU.
#GRAPHIC CARD BENCHMARK 2019 GTX970 DRIVER#
At this point, Nvidia’s video driver attempts to put the least important data in the slower segment. Only after the first 3.5GB is requested, to fill up the entire high performance segment, will the 512MB segment get used.
#GRAPHIC CARD BENCHMARK 2019 GTX970 FULL#
In practical terms, however, applications see the full 4GB of RAM and the video driver balances out the load. The low-performance segment is significantly slower to access than the high-performance segment, hampering performance. As a result, the video memory is split between a high-performance segment containing 3.5GB and a low-performance segment with the remaining 512MB. Instead, one of the GTX 970’s four memory controllers is only partially enabled, meaning only 1.75MB of the L2 cache is switched on. However, it turns out an error in the specification tables that went out to reviewers and board partners means this isn’t the case.
The GTX 970 was understood to have three fewer streaming processors than the GTX 980, meaning fewer texture units and a slower core clock speed, but otherwise retain the same 4GB of GDDR5 memory and 2MB of L2 cache for accessing it. Nvidia originally pitched the GTX 970 as a die-harvested derivative of the flagship GTX 980 - this means the company would re-use chips that weren’t able to run at full GTX 980 speeds for the less powerful model, rather than let them go to waste. The most powerful Nvidia-based card we’ve tested is the GTX 970.
0 kommentar(er)
