So it’s about 3 weeks ago, one day after I
published my RTX 4070 vs 3070 Ti comparison when I find out that there might be a problem with my
4070 laptop, meaning the results could be wrong. I get this message from XMG, who sent the 4070
laptop I used, showing that at the factory their 4070 laptop hits 185 FPS in Shadow of the Tomb
Raider at 1080p highest settings, which is 18% higher than what I originally got! A fairly big
difference. I only got this 185 FPS number after I’d already tested a few games on this laptop
and they’d told me I was good to proceed. Unfortunately their first batch of early
mass production samples had two issues, which were high unit-to-unit
variance in CPU cooling, and higher than expected GPU
memory temperatures on air cooling. I actually discovered performance problems on both
the 4080 and 4090 configurations, which is why we haven’t tested them yet. We went through days of
troubleshooting and eventually were told that the 4060 and 4070 laptops should be good to go. So we
started the testing. But as it turns out, there was a problem with both the 4060 and 4070 laptops.
It just wasn’t quite as obvious. I just want to make it super clear that this only applies to the
early samples that were sent to reviewers like me, any customers that pre-ordered these laptops
have been kept up to date and the laptops were delayed while they fixed the problems. So just
to be clear, no customers were affected by this. Being able to provide you with
accurate results and be a trusted source for benchmark data is one of my top
priorities. So after hearing that the 4070 laptop GPU might be 18% faster than what
I originally showed in the first video, my first panicked thought was that I need to
delete the video and redo days of testing as soon as possible. Except it would take a couple
of weeks for the root cause to be fixed and it would take time to send me a replacement
laptop, which is what we’ve got here. But hang on, after 4070 reviews went live,
others like Hardware Unboxed said that the RTX 4070 was so boring that they’re not
even going to make a video about it, so the 4070 can’t be that much better than what
I had in my video, right? I actually spoke to Tim before the 4070 embargo ended, because I
thought my results were suspiciously close. There was no real difference between the 3070 Ti
and 4070, which I thought was a bit strange. But we had similar results, and at that time I thought
great, someone else has confirmed what I’ve seen. But then a few days later it hit me. Hardware
Unboxed were also using an early 4070 laptop sample provided by XMG. So if my RTX
4070 laptop wasn’t performing properly then it’s possible that theirs wasn’t
either. It could have the same problem. I also remember a conversation I had
with someone from one of the bigger laptop brands. I asked them what they thought
about the 4070 performance and they kind of laughed and said something along the lines of
“yeah I guess there’s a reason that Nvidia are launching the 4080 and 4090 first”. So this
big laptop brand also didn’t think that the 4070 was anything too special, so again
I thought that my results were correct. After thinking about it, I decided to keep the
original video up until I could get the new 4070 laptop replacement to retest everything. It
seemed kind of silly to take down the video based on one game test that was provided to
me. Especially as their testing was done with different memory. They used DDR5-4800 while
I used DDR5-5600 but with slower timings, and Shadow of the Tomb Raider is known
to be memory sensitive, so it’s possible that some of that 18% performance gain they
showed me was due to the memory difference. Anyway, we've got the new 4070 replacement
laptop here which has been rigorously tested at the factory, so it will provide us with accurate
results. We could have just retested a few games, but I decided to spend the extra
time and re-test all 25 games at 1080p and 1440p resolutions, as this will let
us fairly compare against the last comparison. Alright these are the 1080p results from the
original RTX 4070 vs RTX 3070 Ti comparison video. In 25 games, the 4070 appeared to be 3%
faster. Basically the way to read this graph, is these games were this percent faster on the
4070, while these ones ran slower on the 4070. Now here are the new results after retesting
the RTX 4070 in the same games. The 4070 is now 7% faster compared to the 3070 Ti on
average, with just 3 of the 25 games running slower on the 4070. So a better result for the
4070, but it’s not really a major difference, at least when averaging the results from all 25
games, and pretty much all the games towards the top are CPU heavy. Which makes sense considering
the major limitation before was on the processor. Now these are the old results for the higher
1440p resolution. In the original video, the data showed that the RTX 4070 was
almost a percent slower compared to the RTX 3070 Ti. Basically no real difference
once you average a bunch of games together. After the retesting, the RTX 4070 was now about
2% faster compared to the RTX 3070 Ti, however 9 of the games were still performing better on the
3070 Ti laptop. So the 4070 is better than before, but there’s definitely less of a difference
at 1440p compared to 1080p. Even at this higher resolution, many of the games that were
doing better on the 4070 laptop are CPU heavier titles. And based on my conversations with XMG,
it seemed like the 4070 in the old laptop was working perfectly fine. The problem was that
the CPU could hold it back in some workloads. You can see all the results with actual FPS
numbers from the 25 games with the link below the video. I’ve uploaded them for free to my Patreon
page, you do not have to pay to see them, but if you’d like to join our discord community and get
behind the scenes details like this before the video, then consider supporting our independent
testing. Or in this case, our re-testing. So was I wrong about the RTX 4070
laptop? No, I don’t think so. The data collected from the first
4070 laptop certainly wasn’t perfect, as we’ve seen there was a clear improvement
retesting with the fixed 4070 laptop, but the difference really wasn’t big
enough to affect any of the conclusions or anything else I said in the original
video, especially at the higher 1440p resolution where the GPU change makes less of
a difference, and this is a GPU comparison. So after completing this re-testing, I don’t
think that it’s worth deleting the original video, but I will update it and leave a link to this
video with the new results. The 4070 still isn’t really that much better compared to the 3070 Ti,
at least without features like frame generation. I really love the idea of getting next gen
hardware like this before it’s available so I can thoroughly test it out to let you know
how it performs, but after this experience maybe it’s not worth the hassle. I’d rather
be accurate from the start and not have to redo everything. Even if it means waiting a
month or two. Fortunately, it wasn’t that big of a deal this time around, but regardless I
am glad to have that correct data out there. I also want to say that I don’t
blame XMG for any of this, I really appreciate that they were willing
to send me the same laptop with a 4060, 4070, 4080 and 4090. We’re going to have some
really awesome comparisons coming! We’re still waiting to get the 4080 and 4090 replaced,
but I’ve already got the 4070 and 4060, so I’m hoping next week we’ll have that 4060 vs
4070 comparison, make sure you’re subscribed! Although there wasn’t really much
difference between the 4070 and 3070 Ti, there actually is a much bigger difference
between the 4080 and 3080 Ti. You can check that comparison over here next. I’ve
compared them in 25 games at 1080p and 1440p, but also the higher 4K resolution as those GPUs
are more powerful, so I’ll see you in that one!
