• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

CES 2025 AMD & Nvidia Keynote Live Chat [06/01/25]

I'll look at the reviews the day before I guess. If a 5080 is mostly 4090 performance with a grand in my pocket, that may be the better option after all.
Or 5070Ti when the arrives as it seems to be just below 5080, with same vRAM and should be considerably cheaper. That said,I highly doubt 5080 will be on 4090 level of performance but independent benchmarks should confirm - as long as they are done properly and it's not just dlss4 advert :p
 
I'm starting to think there might be a genuine opening for AMD here, under certain narrow conditions:
1. The 5070 is around a 4070ti in raster.
2. The 9070xt is around a 7900xt in raster.
3. The 9070xt is 400 quid.

That would make the AMD card 10% more powerful, with 16gb of VRAM (as opposed to just 12gb) and 100 quid cheaper. For me, a 500 quid card with only 12gb of VRAM in 2025 is DOA. If AMD can still make a profit at 400, they have a real opportunity for success.
 
Also if the FE is £1939 then you can pretty much guess that the cheapest AIB is going to be £2050-2100 and more expensive ones up to £2500... if i remember correctly it was £1699 for an FE 4090 and i ended up paying £1910 for a gigabyte gaming OC from overclockers on launch day
 
Last edited:
Also if the FE is £1939 then you can pretty much guess that the cheapest AIB is going to be £2050-2100 and more expensive ones up to £2500... if i remember correctly it was £1699 for an FE 4090 and i ended up paying £1910 for a gigabyte gaming OC from overclockers on launch day

If it's like last time I think AIBs have to offer one sku (nearly always the none oc) at MSRP for launch.
 
wait what? 5090rtx has a 30% increase in rasta performance vs the 4090? Link? source?

I don’t think anybody knows for sure, it’s just extrapolation from the spec sheet:

GPURTX 4090RTX 5090
CUDA Cores16,38421,760
Core Frequency2.23 GHz2.01 GHz
Boost Core
Frequency
2.52 GHZ2.41 GHZ
Memory
Capacity / Type
24 GB GDDR6X32 GB GDDR7
Memory
Bandwidth
1008 GB/s1792 GB/s
Memory
Interface
384-bit512-bit
Memory Clock21 Gbps28 Gbps
TGP450W575 W

Ignoring all the DLSS stuff the 4090 is still absolutely dominated by the 5090, other than the 5% drop in boost clocks. I’d be very surprised if it’s only a 20% gain, but you never know!

By the time we’re seeing above 24g vram usage as standard and the 5090 pulls ahead significantly, it’ll probably be 4 years time and these cards may both be outdated compared to the new top end.
 
Last edited:
It may be "dominated" by the numbers in the table but many of these numbers have little to no impact on performance.

Memory clock, bit rate and power draw don't matter, what does matter is bandwidth, and yes it's hugely higher but I'm not convinced it's needed unless you are a 8k gamer. Memory capacity also doesn't matter

Then for core count, it's a 30% increase and a 5% decrease in clocks, for a net 25% improvement. As we saw with the 4090 each extra 1 core you add on Nvidia cards doesn't add 1 unit more performance, so this net 25% won't be 25% more gaming rasterisation
 
Last edited:
Until tested side by side you can't really account for any architecture changes, optimisations and so on even at the same terraflops, etc. numbers effectively utilisation of those stats can change.
 
Until tested side by side you can't really account for any architecture changes, optimisations and so on even at the same terraflops, etc. numbers effectively utilisation of those stats can change.


I think if there was changes Jensen would have announced it. All he mentioned was changes to AI and RT cores, he never said they improved Cuda cores or Rasterisation performance

In previous generations when they improved the Cuda cores they announced proudly and the last time this happened was in the RTX3000 series and Jensen was proud to tell you how they improved rasterisation by updated Cuda cores architecture
 
Last edited:
Back
Top Bottom