Test Setup
Please read the information on our GPU testing procedures and methodology here. Note that our GPU Testing Bench has been upgraded to the Ryzen 7 7800X3D CPU and AM5 platform.
Note that we have the 1602 BIOS installed, which uses AGESA version Combo AM5 PI 1.0.0.7c. We are using the motherboard’s current default BIOS settings on the latest release BIOS. We have enabled PCI-Express Resizable BAR in the motherboard BIOS. This means we are utilizing Resizable BAR on NVIDIA GPUs and AMD Smart Access Memory on AMD GPUs in all our testing. We have enabled EXPO I in the BIOS for the memory.
We are using a fresh install of Windows 11 Pro with the latest Windows Updates, including the 22H2 Fall 2022 update. We have the latest drivers installed including the latest chipset drivers from AMD. We are using the “Balanced” power profile in Windows Settings. Hardware Accelerated GPU Scheduling is ENABLED in Windows. VBS is DISABLED and Game Mode is ENABLED.
Discussion (7 replies)
Join Discussion →I get comparing to it's closest competitor. But I feel like I must be remembering reviews with rose colored glasses. Didn't reviews compare new cards to one step up and one step down as well to see how they fit in the stack and if they were worth while?
I mean if you're constantly reading/watching reviews and ingesting data I suppose having just an review against the other teams price parity (or closest) makes sense. But as a consumer I would love to see if the 78000 XT card stacks well against another brands work with an same model, a 7900xt, and a 7600xt, as well as a 4080, and 4070ti/non ti.
Nice review, Brent.
Man I just about had my mind set on a 4070... this is $30 cheaper (than the $550 4070 Gigabyte I was looking at) and seems to be better in most cases. HMMMMMM.
You can look at what we've already reviewed and get a good idea of just that. We have already reviewed the XFX, Sapphire Nitro and reference AMD 7800XT cards. Thats real apples to apples. If you throw in a lesser or higher tier card, then you really don't get fair comparisons. What we try to do is show what is out there in the same price bracket or aimed at the same or similar gaming experience. TBH, right now the prices are all over the place, so that can be hard to do. That said, these reviews have always been pretty much about comparing cards that are closest in price and performance. We do throw in prior generation cards when a new gen is released. Also you have to consider the time it takes to run the data, graph it all out, compile the stats and then publish it. It takes about 4 or 5 days to knock out a polished review using an overclocked card, the out of box card and then the comparison card. If we did any more this stuff would be ancient history by the time you read it.
So why not include those in the chart? Or are we dealing with a like for like hardware issue between reviews. Since recreating the data doesn't make sense. Oh.. I get it with how intrisic driver versions are to performance improvements that wouldn't make sense.
Either you bench everything with an older driver, or you bench everything again with a newer one - and if you're doing it correctly and not over-automating reviews such that you miss things like artifacting and such, or using built-in benchmarks that do not represent gameplay (and perhaps favor one vendor over another, and that is different than gameplay) so that you wind up with results that are correct. At a minimum, really.
Benchmarking well is really a throughput problem that isn't easy to solve, unfortunately.
That has alot to do with it. Everything is tested on that day, using the latest most up to date everything.
Game updates can also make a difference - some games like Cyberpunk on GOG are easy enough to time machine into a specific version, but the more connected ones can force upgrade you.
There's a lot of variables to control for that we have a stack of internal documentation on to get things set up the same way every time. When looking at a single card, it's focused on 1-2 comparison options at that point in time. For a GPU level review, you'll see we run a much larger comparison dataset (usually limited by time before launch and who has what to ship to whom).