News Best Graphics Card Deals for Prime Day October: Cheap Nvidia and AMD GPUs

bigdragon

Distinguished
Oct 19, 2011
1,142
609
20,160
A 3090 for $870... Is this a good deal or not?

Benchmarks say it'll cut Blender renders in half compared to the 2080 Super. I'd assume the 3090 would also perform well in the UE5 editor. Still, that's a lot for a GPU that appears to perform on par with the 4080 12gb.
 
Last edited:
  • Like
Reactions: artk2219
When I click on your link for AMD cards at Amazon, it brings up about 90% Nvidia cards.
Yeah, that's Amazon for you. Search for any GPU, a bunch of the results will be from completely different GPUs because maybe you want one of those instead? I hate it. Newegg's search tool is 100x better.
A 3090 for $870... Is this a good deal or not?

Benchmarks say it'll cut Blender renders in half compared to the 2080 Super. I'd assume the 3090 would also perform well in the UE5 editor. Still, that's a lot for a GPU that appears to perform on par with the 4080 12gb.
If you do a lot of "professional" work, I'd say it's a decent deal on a 3090. At the same time, if you're doing professional work that leverages a GPU, you probably should have bought a 3090 already and would now be eying the (sold out) 4090.

The 4080 12GB will probably only be significantly faster in games that use DLSS 3. So you'd be paying a bit less than the 4080 12GB base price for double the VRAM and a card that will do nicely in certain non-gaming tasks. We'll see how the 4080 12GB (aka shoulda-been-4070) performs in independent benchmarks next month.
 

Udyr

Honorable
Mar 3, 2021
256
107
9,690
Yeah, that's Amazon for you. Search for any GPU, a bunch of the results will be from completely different GPUs because maybe you want one of those instead? I hate it. Newegg's search tool is 100x better.
Not really. Whenever I search for a, lets say, 6900 XT, the first 25+ results are for the specific unit I searched, and only after the 30th+ I get mixed results for other AMD based units and whatever Amazon considers the Nvidia equivalent.

It could probably be your specific user and it's search history or TH's commission based links affecting the results.
 
  • Like
Reactions: artk2219
Not really. Whenever I search for a, lets say, 6900 XT, the first 25+ results are for the specific unit I searched, and only after the 30th+ I get mixed results for other AMD based units and whatever Amazon considers the Nvidia equivalent.

It could probably be your specific user and it's search history or TH's commission based links affecting the results.
It looks like it depends a lot on his the availability of whatever you’re searching for. If you look for a popular graphics card, your results will be better than if you look for something that’s either new or old. However, Amazon‘s search is almost completely useless if you try to do something like sort by price, low to high. Then searching for something like RTX 3090 will show dozens of results for crappy low end cards before you ever see a single 3090.

I know I’ve searched for various GPUs many, many times at Amazon and routinely get terrible matches. Maybe it’s Amazon storing cookies and trying to figure out what to show me, and certainly some of it is the terrible product descriptions some companies post. “GT 730, get this while you wait for RTX 3090!” I don’t know, but I do know Newegg is way more helpful in that it doesn’t fill my results with unrelated garbage.
 
  • Like
Reactions: artk2219

bigdragon

Distinguished
Oct 19, 2011
1,142
609
20,160
If you do a lot of "professional" work, I'd say it's a decent deal on a 3090. At the same time, if you're doing professional work that leverages a GPU, you probably should have bought a 3090 already and would now be eying the (sold out) 4090.

The 4080 12GB will probably only be significantly faster in games that use DLSS 3. So you'd be paying a bit less than the 4080 12GB base price for double the VRAM and a card that will do nicely in certain non-gaming tasks. We'll see how the 4080 12GB (aka shoulda-been-4070) performs in independent benchmarks next month.
I did decide to get that Zotac 3090. I'm in the prosumer / indie category so there's a limit to how much money is available for tech gear. 24GB VRAM and 10k+ CUDA cores is hard to ignore at that price plus Amazon stacking another 10% discount on top for using their credit card. The final price is very close to what I would have paid for a 3080 10GB 2 years ago.

A lot of the early info about the 4080 has been disappointing. I know the independent reviews and benchmarks aren't out yet. However, the price for performance seems very unreasonable. I'm worried Nvidia is going to keep reaching for the moon while letting AMD and Intel fight for the mainstream and low-end markets.
 
  • Like
Reactions: artk2219
The 4080 12GB will probably only be significantly faster in games that use DLSS 3....
@JarredWaltonGPU, check out Hardware Unboxed's review of DLSS 3. Very interesting, to say the least.
It doesn't bode well for those looking to enable DLSS 3 just to get over 60FPS in high quality games with RT on, at least in its initial, release version. Be interesting to see of Tom's comes to the same conclusion.

View: https://www.youtube.com/watch?v=GkUAGMYg5Lw


Edit - I just found your pre-release report of DLSS 3, from 15 days ago. Definitely a much softer conclusion, from a quality standpoint. Perspective is necessary, I guess. DLSS 3 is a 'freebie'. If looked at from that point of view, you can either use it if it works, or stick with DLSS 2 or just native frames if it doesn't.
 
Last edited:
  • Like
Reactions: artk2219

LokkenJP

Distinguished
Jun 17, 2014
28
3
18,545
Hello.

I'm not sure if this GPU has been released on the US or not, but here in Europe one killer deal nowadays is the AMD Radeon 6700 non-XT (retailed mostly from Sapphire and XFX brands), which is not even mentioned on this article.

It's about 96% performance of the full fat 6700 XT, but almost at 6600 price levels, and usually lower than 6600XT on most cases.

As an example here on one of the biggest spanish retailers:
https://www.pccomponentes.com/buscar/?query=amd 6700&or-relevance (AMD 6700 in the 420-480€ range depending on brand and model)
And in particular:
https://www.pccomponentes.com/sapphire-pulse-amd-radeon-rx-6700-oc-10gb-gddr6 (the Sapphire Pulse model, now in "offer", but in fact it has been in this "offer" for months now).

Meanwhile:
https://www.pccomponentes.com/buscar/?query=amd 6600&or-relevance
The cheapest brand new 6600 (non-xt) is about 380€ and most models are sold on the lower 400-ish.
And the 6600XT/6650XT all go into the high-400/500/600€ range.

Note that overall, prices here in EU are way higher than on US, even on the same brand and model GPUs.
 
Last edited:
  • Like
Reactions: artk2219

mickrc3

Distinguished
Mar 20, 2016
87
23
19,865
Still showing 'Renewed' cards like the RTX 2060 on the chart, not a new card. If you're going to show non-new GPUs then EVGA's B-Stock RTX 2060 is $189, their 3060 ti is $349.
 
Last edited:

rluker5

Distinguished
Jun 23, 2014
901
574
19,760
IS THE 6800XT the best bang for your buck?
Depends on your use case, but for most, yes.
For someone who just wants no frills fast for cheap it is one of the best deals for at least the last 5 years.
And I have a 3080, 6800 and A750 and the 3080 is my favorite, but no way is a 3080 worth 40% more than a 6800XT. Maybe 10% if you just prefer Nvidia and -10% if you prefer AMD. These 6800XTs are almost twice as fast as a console for about the same price (if you already have a pc to put it in).
 
  • Like
Reactions: artk2219 and gg83

tamalero

Distinguished
Oct 25, 2006
1,226
241
19,670
A 3090 for $870... Is this a good deal or not?

Benchmarks say it'll cut Blender renders in half compared to the 2080 Super. I'd assume the 3090 would also perform well in the UE5 editor. Still, that's a lot for a GPU that appears to perform on par with the 4080 12gb.

Id just wait until the 7000 series of AMD is out then select your options.
 
  • Like
Reactions: artk2219

Elusive Ruse

Estimable
Nov 17, 2022
452
586
3,220
@JarredWaltonGPU, check out Hardware Unboxed's review of DLSS 3. Very interesting, to say the least.
It doesn't bode well for those looking to enable DLSS 3 just to get over 60FPS in high quality games with RT on, at least in its initial, release version. Be interesting to see of Tom's comes to the same conclusion.

View: https://www.youtube.com/watch?v=GkUAGMYg5Lw


Edit - I just found your pre-release report of DLSS 3, from 15 days ago. Definitely a much softer conclusion, from a quality standpoint. Perspective is necessary, I guess. DLSS 3 is a 'freebie'. If looked at from that point of view, you can either use it if it works, or stick with DLSS 2 or just native frames if it doesn't.
Digital Foundry came to similar conclusions as Hardware, for now DLSS 3.0 is basically fake frames unless you are driving a car in a straight line for a long time. If that was not enough, it increases your input latency by a wide margin since it holds two frames to compare and come up with that frame in the middle that it's supposed to generate.

View: https://www.youtube.com/watch?v=92ZqYaPXxas
 
  • Like
Reactions: artk2219
Digital Foundry came to similar conclusions as Hardware, for now DLSS 3.0 is basically fake frames unless you are driving a car in a straight line for a long time. If that was not enough, it increases your input latency by a wide margin since it holds two frames to compare and come up with that frame in the middle that it's supposed to generate.

View: https://www.youtube.com/watch?v=92ZqYaPXxas
I actually played Miles Morales with and without DLSS 3. Without pixel peeping and doing a bunch of extra stuff, I can honestly say I never noticed anything that made me feel like FG was causing rendering errors. Did it look better? No. Did it look or feel at all smoother? On the 4080, at maxed out settings and 4K, yes. I think it did. I imagine there's a sweet spot between a minimum real FPS requirement and a maximum where DLSS 3 feels best.

That's important to understand. In normal use, without going to great lengths, it's difficult for even a professional graphics card reviewer to detect the errors. It's not that they don't exist, but it may very well be that they don't really matter.

If you listen to the lengthy explanations given on the problems with DLSS3, what it all devolves to is that you generally have to carefully look for rendering problems, sometimes going so far as to intentionally do things to create the problems (like moving the mouse around really fast). It's a pain in the rear to test, as you need to capture 4K at 120 Hz basically. Some of the rendering errors are quite bad, but the inconsistency of the problem means the actual gameplay experience may not be affected all that much.

I'm still not sure what to actually think, having tried DLSS3 in several games now. For me, it was fine. I could enable or disable it and not really notice much of a difference — again, without putting in a lot of time to test and try and figure out precisely what's happening on a frame by frame basis. But the 4090 and 4080 are already so fast that perhaps it's less of a problem. How will Frame Generation feel on a future RTX 4050? I suspect a lot of the problems discussed by others will become far more noticeable. That's not going to be a good experience, in other words.

But ultimately, I do like having choice. Maybe some games I decided I like the FG effect, others I don't, some run fast enough without it, etc. It's the same reason I like that Miles Morales also supports FSR 2.1 and XeSS alongside DLSS 2/3. The more the merrier! I mean, I'm never going to be upset that a game has MSAA, TAA, FXAA, DLAA, DLSS, FSR 2.1, XeSS, etc. Maybe I never use some of the options, but maybe someone else cares enough to use a specific setting.