-
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.
It looks like the 'real' /affordable RDNA3 + next gen NV desktop launch won't launch until September. Thoughts?
- Thread starter g67575
- Start date
Yup, the RX 6800 seems reasonably priced for the performance it offers atm. But I'd rather wait just a little longer.
I sort of agree about waiting, except that used prices do seem to be coming down. It's not easy to get a good deal though, as stock is limited.
If you have the cash to spend, I'd say it's not worth waiting (if it's going to be too frustrating).
I sort of agree about waiting, except that used prices do seem to be coming down. It's not easy to get a good deal though, as stock is limited.
If you have the cash to spend, I'd say it's not worth waiting (if it's going to be too frustrating).
Last edited:
Lets assume the rumoured specs of the RTX 4070 are false. For one thing, 5,888 shaders would be the same amount as the RTX 3070...
If Nvidia released a version of the top AD104 chip, but without the GDDR6X, would anyone here buy it, if it was priced £100 lower?
With GDDR6X, the total memory bandwidth for AD104 is 504.2 GB/s (assuming 192 bit memory bus). Less than the RTX 3070 TI!
With GDDR6 (clocked at 18gbps) it would be 432 GB/s, similar to the RX 6750 XT:
www.techpowerup.com
Or, they could use even faster GDDR6 (which would probably put it too close to the TI, at 480 GB/s.
Either way, not much of a decrease.
The idea is basically exactly what they did with the RTX 3070 and RTX 3070 TI (one interesting thing about these cards, is that the RTX 3070 has a TDP that is 70w lower), the only difference being that with the RTX 4000 series, the TI was released first.
So, you'd see perhaps just over 5% difference in performance (on average).
EDIT - The RTX 3070 had 2 fewer SMs than the RTX 3070 TI, so perhaps the RTX 4070 will have 2 fewer SMs also (58). So, that would result in 7,424 shaders.
Seems as though the RTX 4060 TI is going to use GDDR6 VRAM (apparently 18gbps), so there's no reason at all why this wouldn't be the case for the RTX 4070.
If they released this card, it would likely be very competitive (in terms of performance) with any Navi32 based cards released by AMD...
EDIT 2 - 7,424 happens to be the same number of shaders used in the AD104 based RTX 4080 mobile GPUs.
Both of these mobile chips use GDDR6 VRAM.
The number of cores (58) seems to have been confirmed here by a Geekbench result:
https://videocardz.com/newz/nvidia-...ature-7424-cuda-cores-42-faster-than-rtx-3080
I think we will basically need to wait and see what they do with RTX 4080 mobile chips (obviously these mean mega bucks for Nvidia), shouldn't be long before these launch.
Nvidia says "From February 8th, our critically acclaimed GeForce RTX 40 Series GPUs come to laptops with the launch of GeForce RTX 4090 and 4080 laptops" with other GPUs in the series coming on 22nd Feb.
Link:
https://www.nvidia.com/en-us/geforce/news/geforce-rtx-40-series-laptops-available-february-8/
If Nvidia released a version of the top AD104 chip, but without the GDDR6X, would anyone here buy it, if it was priced £100 lower?
With GDDR6X, the total memory bandwidth for AD104 is 504.2 GB/s (assuming 192 bit memory bus). Less than the RTX 3070 TI!
With GDDR6 (clocked at 18gbps) it would be 432 GB/s, similar to the RX 6750 XT:
AMD Radeon RX 6750 XT Specs
AMD Navi 22, 2600 MHz, 2560 Cores, 160 TMUs, 64 ROPs, 12288 MB GDDR6, 2250 MHz, 192 bit
Or, they could use even faster GDDR6 (which would probably put it too close to the TI, at 480 GB/s.
Either way, not much of a decrease.
The idea is basically exactly what they did with the RTX 3070 and RTX 3070 TI (one interesting thing about these cards, is that the RTX 3070 has a TDP that is 70w lower), the only difference being that with the RTX 4000 series, the TI was released first.
So, you'd see perhaps just over 5% difference in performance (on average).
EDIT - The RTX 3070 had 2 fewer SMs than the RTX 3070 TI, so perhaps the RTX 4070 will have 2 fewer SMs also (58). So, that would result in 7,424 shaders.
Seems as though the RTX 4060 TI is going to use GDDR6 VRAM (apparently 18gbps), so there's no reason at all why this wouldn't be the case for the RTX 4070.
If they released this card, it would likely be very competitive (in terms of performance) with any Navi32 based cards released by AMD...
EDIT 2 - 7,424 happens to be the same number of shaders used in the AD104 based RTX 4080 mobile GPUs.
Both of these mobile chips use GDDR6 VRAM.
The number of cores (58) seems to have been confirmed here by a Geekbench result:
https://videocardz.com/newz/nvidia-...ature-7424-cuda-cores-42-faster-than-rtx-3080
I think we will basically need to wait and see what they do with RTX 4080 mobile chips (obviously these mean mega bucks for Nvidia), shouldn't be long before these launch.
Nvidia says "From February 8th, our critically acclaimed GeForce RTX 40 Series GPUs come to laptops with the launch of GeForce RTX 4090 and 4080 laptops" with other GPUs in the series coming on 22nd Feb.
Link:
https://www.nvidia.com/en-us/geforce/news/geforce-rtx-40-series-laptops-available-february-8/
Last edited:
I'm fairly sure that RTX 4070 will have almost the exact same spec as the mobile RTX 4080 (AD104) GPU:
www.techpowerup.com
The same 58 cores, ROP and TMU count, but with the GPU clocked much higher, probably 2600Mhz.
Also, GDDR6 memory too, probably the same 16Gbps spec?
In that case, the memory bandwidth would be 384.0 GB/s, so not great, not terrible. Something that Nvidia can cheap out on, I suppose.
It was a similar story for the RTX 3070 and RTX 3070 TI Mobile, same slightly cut down spec, both with the same GDDR6 VRAM spec (vs the full 48 core GA104 die):
https://www.techpowerup.com/gpu-specs/geforce-rtx-3070.c3674
https://www.techpowerup.com/gpu-specs/geforce-rtx-3070-ti-mobile.c3852
The high tier laptop GPUs take priority though, because they are basically like 'white gold', in terms of profitability.
NVIDIA GeForce RTX 4080 Mobile Specs
NVIDIA AD104, 1665 MHz, 7424 Cores, 232 TMUs, 80 ROPs, 12288 MB GDDR6, 2250 MHz, 192 bit
The same 58 cores, ROP and TMU count, but with the GPU clocked much higher, probably 2600Mhz.
Also, GDDR6 memory too, probably the same 16Gbps spec?
In that case, the memory bandwidth would be 384.0 GB/s, so not great, not terrible. Something that Nvidia can cheap out on, I suppose.
It was a similar story for the RTX 3070 and RTX 3070 TI Mobile, same slightly cut down spec, both with the same GDDR6 VRAM spec (vs the full 48 core GA104 die):
https://www.techpowerup.com/gpu-specs/geforce-rtx-3070.c3674
https://www.techpowerup.com/gpu-specs/geforce-rtx-3070-ti-mobile.c3852
The high tier laptop GPUs take priority though, because they are basically like 'white gold', in terms of profitability.
Last edited:
Out of interest, is anyone going to buy a laptop with an RTX 4080 mobile GPU?
It's looking like it will perform faster than a desktop RTX 3070 (but it has a bit lower memory bandwidth apparently).
These should be launched on 08/02/2023.
It's looking like it will perform faster than a desktop RTX 3070 (but it has a bit lower memory bandwidth apparently).
These should be launched on 08/02/2023.
Last edited:
I suspect Nvidia is thinking that they don't even need to bother with FE / Reference models for the 'Mid end' GPUs.
It seems clear so far, that Nvidia is going to prioritise laptop GPUs over the mid end RTX 4000 series desktop cards (and these are very similar to desktop models, some of which could have probably been adapted for use as reference model GPUs instead).
They will sell all of the RTX 4070s easily. Because the numbers of dies produced at the mid/lower end should be greater, they can probably try to set prices close to MSRP just by having a healthy supply of AD104 cards.
It's not the mid end cards they are concerned about selling, I think they only worry about selling the super expensive cards like the RTX 4070 TI, 4080 and 4090.
No issue with RTX 4070 TI sales from what I hear.
AMD doesn't seem to be putting much effort into producing reference models (none at all in the UK), so not much competition there. If AMD seriously wanted to compete, this is how they would do it, by fixing prices and releasing lots of reference model cards. Never happened with RDNA2, but I suppose not surprising because they never managed to produce enough. It's disappointing that AMD were patting themselves on the back for even producing any reference models at all for RDNA2.
Last edited:
Quite a significant difference in warranty claims for graphics card producing companies here:
AsRock, Gainward and Inno3D all have similarly low warranty claims within 2 years.
Palit does a bit better than I'd expect (some of their designs look a bit cheap imo).
The return rate on AMD, Sapphire and XFX cards doesn't reflect that well on AMD GPUs.
No data on how long it takes AsRock to process and Inno3D to process claims though (but it does take 13 days on average for AsRock to process motherboard claims).
From here:
https://www.techspot.com/news/97425...ble-graphics-cards-motherboards-warranty.html
The data is from a Swiss retailer.
I'd guess that many RMA claims would be related to cooler failure /faults. Even for expensive graphics cards, companies don't always use fans with a decent design and build quality.
AsRock, Gainward and Inno3D all have similarly low warranty claims within 2 years.
Palit does a bit better than I'd expect (some of their designs look a bit cheap imo).
The return rate on AMD, Sapphire and XFX cards doesn't reflect that well on AMD GPUs.
No data on how long it takes AsRock to process and Inno3D to process claims though (but it does take 13 days on average for AsRock to process motherboard claims).
From here:
https://www.techspot.com/news/97425...ble-graphics-cards-motherboards-warranty.html
The data is from a Swiss retailer.
I'd guess that many RMA claims would be related to cooler failure /faults. Even for expensive graphics cards, companies don't always use fans with a decent design and build quality.
Last edited:
0.3 for the lowest return rate vs 2.5% for the highest, I'd call it significant.
It's more likely that they are selling well over 300, especially over 2 years.
Anything above 1% over 2 years I'd probably avoid.
With the exception of reference models, average reliability of these might be reduced just because of a 'bad batch'.
GPU failures don't seem to be common, I don't think there'd be much (or any) difference between Nvidia and AMD here.
I doubt that the introduction of GDDR6X has helped in terms of reliability, due to the higher temps.
Fans seem to fail a lot on graphics cards too, you would have thought this would be an easy problem to solve, as you can buy decent case fans that are rated for 5-10 years.
It's more likely that they are selling well over 300, especially over 2 years.
Anything above 1% over 2 years I'd probably avoid.
With the exception of reference models, average reliability of these might be reduced just because of a 'bad batch'.
GPU failures don't seem to be common, I don't think there'd be much (or any) difference between Nvidia and AMD here.
I doubt that the introduction of GDDR6X has helped in terms of reliability, due to the higher temps.
Fans seem to fail a lot on graphics cards too, you would have thought this would be an easy problem to solve, as you can buy decent case fans that are rated for 5-10 years.
Last edited:
Definitely something to watch out for then, when buying a card.
The RTX 4070 is probably coming out in April:
https://videocardz.com/newz/nvidia-geforce-rtx-4070-expected-to-launch-in-april
I think they've got the spec wrong, as it seems to dissimilar to the RTX 4080 mobile (also based on AD104).
Also, they say it will use GDDR6X, even though the RTX 4080 mobile is using GDDR6.
https://videocardz.com/newz/nvidia-geforce-rtx-4070-expected-to-launch-in-april
I think they've got the spec wrong, as it seems to dissimilar to the RTX 4080 mobile (also based on AD104).
Also, they say it will use GDDR6X, even though the RTX 4080 mobile is using GDDR6.
Last edited:
RTX 3080s (and higher tier cards) apparently have been out of stock since December in the US:
www.techspot.com
Wonder if that means the RTX 4070 will be available in about a month?
But RX 6000 series cards still in stock, (except for the 6900 XT).
Also, still no sign of Navi32 chips in laptops, which doesn't bode well.
Will GPUs Ever Get Cheaper? GPU Pricing Update
After a period of consistent new GPU releases from Nvidia and AMD, things have cooled down in February, but it's still worth taking a look at how...
www.techspot.com
Wonder if that means the RTX 4070 will be available in about a month?
But RX 6000 series cards still in stock, (except for the 6900 XT).
Also, still no sign of Navi32 chips in laptops, which doesn't bode well.
Last edited:
I think the prices of the 6700 XT are very good (actually below the MSRP price of £419). I would have bought one, but there are a few games it struggles with at 1440p, like Cyberpunk 2077, WH3 and no doubt the new Hogwarts game.
We basically need the performance of the RTX 3080 but for less money, to play very smoothly at 1440p (or perhaps 1440p upscaled to 4K).
There's also ray tracing to consider, but I think this subject tends to distort people's views about graphics performance. It seems like something to consider only when used together with other features like frame generation, which is still probably 1-2 years away from seeing wider adoption in games.
We basically need the performance of the RTX 3080 but for less money, to play very smoothly at 1440p (or perhaps 1440p upscaled to 4K).
There's also ray tracing to consider, but I think this subject tends to distort people's views about graphics performance. It seems like something to consider only when used together with other features like frame generation, which is still probably 1-2 years away from seeing wider adoption in games.
Last edited:
I don't think we will see Navi32 GPUs until FSR3 /frame generation is ready, which AMD has kept quiet about.
I don't believe that the GPUs themselves aren't ready, but it would make for a more impactful launch if they launch alongside FSR3.
The key thing, is that AMD is still somewhat behind Nvidia in 3D graphics performance with their Navi31 cards (also in ray tracing, but FSR3 can help here).
It's probably just not ready yet, perhaps in a few months.
Nvidia has their frame gen. tech ready to use now, so it means there's really nothing that should hold back the release of cards like the RTX 4070 (except that they want to focus on releasing and producing laptop GPUs first).
I don't believe that the GPUs themselves aren't ready, but it would make for a more impactful launch if they launch alongside FSR3.
The key thing, is that AMD is still somewhat behind Nvidia in 3D graphics performance with their Navi31 cards (also in ray tracing, but FSR3 can help here).
It's probably just not ready yet, perhaps in a few months.
Nvidia has their frame gen. tech ready to use now, so it means there's really nothing that should hold back the release of cards like the RTX 4070 (except that they want to focus on releasing and producing laptop GPUs first).
Last edited:
This is basically what you need to play games at 1440p (ultra, RT off), in the latest titles:
RTX 3080 or RX 6800 XT level of performance.
The RTX 4080 does well with RT enabled:
RTX 3080 or RX 6800 XT level of performance.
The RTX 4080 does well with RT enabled:
Last edited:
There's been a 22.2% increase in inflation since late 2020. The MSRP of the RTX 3070 FE was £470 in October 2020, and there has been approx. 22.3% inflation (RPI) since then, so a similarly priced RTX 4070, taking inflation into account, would cost ~£575. If it uses GDDR6X this will increase the cost a bit (so will using 12GB of VRAM instead of 8GB). Nvidia says their new fab process is expensive also, and the cooler design is improved too (arguably GDDR6X does need this), so a 4070 could end up being quite expensive.
Maybe they could make 2 versions, with different kinds of VRAM?
Maybe they could make 2 versions, with different kinds of VRAM?
Last edited:
I tend to agree, 12GB of VRAM would be a sensible amount to include. My preference would be two 12GB cards, with the difference being whether GDDR6 or GDDR6X is used.
Alternatively, just use the fastest GDDR6 memory available, which would be a new variant called 'GDDR6W' (upto 22gbps). It definitely looks like GDDR6X is due to be superseded by other technologies fairly soon.
Alternatively, just use the fastest GDDR6 memory available, which would be a new variant called 'GDDR6W' (upto 22gbps). It definitely looks like GDDR6X is due to be superseded by other technologies fairly soon.
Last edited: