Any of you guys skipping 5 series?

[Klemc]

So, for a 7800x3d, which will be the last that won't bottleneck too much, 6000 or 7000 series (xx70) ?

 

[Macro Device]

So, for a 7800x3d, which will be the last that won't bottleneck too much, 6000 or 7000 series (xx70) ?

Please be more specific.

If you're a 4K gamer and pursue maximum image quality then it's probably something along the lines of RTX 9070, or maybe 8070 Super/Ti.
If you're a "normal man" and play at 1440p144 and your game library consists of whatever you could ever find on sale then it's likely to be the 7070 Ti or 7080.
If your main goal is a stupid amount of FPS in competitive games then you'll be bottlenecked as soon as they roll a CPU-demanding enough game out.
You can calculate your own case based on that. I also am unsure if you're a DLSS/RT hater or not.

 

[Klemc]

Please be more specific.

If you're a 4K gamer and pursue maximum image quality then it's probably something along the lines of RTX 9070, or maybe 8070 Super/Ti.
If you're a "normal man" and play at 1440p144 and your game library consists of whatever you could ever find on sale then it's likely to be the 7070 Ti or 7080.
If your main goal is a stupid amount of FPS in competitive games then you'll be bottlenecked as soon as they roll a CPU-demanding enough game out.
You can calculate your own case based on that. I also am unsure if you're a DLSS/RT hater or not.

Oh, good, i'll answer:

1440p native
DSR/DLDSR 4K/5K - if it fixes game's aliasing (e.g: Marlow Briggs, Unity games (yeah)) or enhances IQ visibly keeping 120 FPS.

120/240 FPS, no less, no more. 120 being the wanted FPS, and actually always locking to 120 even if 240 outputs easyly (for power and heat). IDC more than 120 FPS.

DLSS when available, always set to Quality, Framegen also used, when available. IDK whch of the two got worst IQ loss btw.

RT/PT: RT always, PT never (FPS tanked)... but i want PT@120 !

//

Perhaps 7070 tho !?

 

[64K]

Please be more specific.

If you're a 4K gamer and pursue maximum image quality then it's probably something along the lines of RTX 9070, or maybe 8070 Super/Ti.
If you're a "normal man" and play at 1440p144 and your game library consists of whatever you could ever find on sale then it's likely to be the 7070 Ti or 7080.
If your main goal is a stupid amount of FPS in competitive games then you'll be bottlenecked as soon as they roll a CPU-demanding enough game out.
You can calculate your own case based on that. I also am unsure if you're a DLSS/RT hater or not.

What?

 

[Macro Device]

Perhaps 7070 tho !?

7080 rather.

 

[Klemc]

7080 rather.

80 series never... 70 yeah so 8070 in that case

//

Not sure a lot people here will wait 4 gen between two GPU

 

[JustBenching]

Ill stick to the 4090 unless they introduce some insane future again like DLSS 2.

 

[Klemc]

Ill stick to the 4090 unless they introduce some insane future again like DLSS 2.

I read today the 5090 will be big upgrade from 4090 ! 70% but it's unbielevable...

 

[Sunny and 75]

6GB for entry-level with 96-bit bus
8GB for lower mid-end with 128-bit bus
12GB for mid-end with 192-bit bus
16GB for lower high-end with 256-bit bus
20GB for mid-level high-end with 320-bit bus
24GB for high-end with 384-bit bus
32GB for enthusiast-level with 512-bit bus

Based on 2GB memory chips.

Micron has said that there will be 3GB memory chips in the future so we could see an xx60 class GPU (or an x600 XT class) with 160-bit bus width and 15GB total VRAM capacity.

And there're already 4GB memory chips available (8GB on 64-bit bus, that means two 4GB memory chips). It's just that GPU manufacturers are NOT willing to ship their midrange to enthusiast segments with 4GB memory chips just yet.

That could change however with the RDNA4, the 50 series and the BMG new graphics card launches by RGB, most likely starting in 2025 and going forward.

Spoiler: If I were the GPU manufacturer

Would release the 40 series lineup like this (memory configuration adjustments):

RTX 4060/Ti with 192-bit 12GB VRAM

RTX 4070/Super/Ti/Ti Super with 256-bit 16GB VRAM

RTX 4080/Super with 320-bit 20GB VRAM

(2GB memory chips)

 

[Macro Device]

80 series never

Why so? I used every single NVIDIA GPU tier (albeit 7+ y.o. generations only) and can't say xx80 is any worse than xx70. My previous GPU was a GTX 1080 Ti and it had been a top-notch experience. As of today, be me in the market for a GPU, I'd buy a 4080 over a 4070 almost regardless of the situation just because it's so much more potent. Higher peak wattage is pretty handily mitigated by the fact it provides higher FPS if limited to 200 W. No idea what else could stop you. The price? It's equally terrible in both products.

 

[Ruru]

Based on 2GB memory chips.

Micron has said that there will be 3GB memory chips in the future so we could see an xx60 class GPU (or an x600 XT class) with 160-bit bus width and 15GB total VRAM capacity.

And there're already 4GB memory chips available (8GB on 64-bit bus, that means two 4GB memory chips). It's just that GPU manufacturers are NOT willing to ship their midrange to enthusiast segments with 4GB memory chips just yet.

That could change however with the RDNA4, the 50 series and the BMG new graphics card launches by RGB, most likely starting in 2025 and going forward.

Spoiler: If I were the GPU manufacturer

Would release the 40 series lineup like this (memory configuration adjustments):

RTX 4060/Ti with 192-bit 12GB VRAM

RTX 4070/Super/Ti/Ti Super with 256-bit 16GB VRAM

RTX 4080/Super with 320-bit 20GB VRAM

(2GB memory chips)

Yeah, I just mentioned something I'd personally put to the market if I could decide.

Interesting if VRAM will see something similar as RAM since now 24GB and 48GB DDR5 sticks exist.

 

[Sunny and 75]

Interesting if VRAM will see something similar as RAM since now 24GB and 48GB DDR5 sticks exist.

Indeed.

 

[Solaris17]

I have a 4090 and am totally thinking about skipping 5 series. Anyone else?

Nah I’ll get 5 series and battlemage. Probably even AMD or pick up a 7900xtx those were pretty sick.

 

[Sunny and 75]

12-16GB 5060/5070, 24-32GB 5080

This^

Micron's roadmap indicates 3GB GDDR7 32Gbps memory chips arriving in 2025 (most likely second half).

Spoiler: DRAM density

The capacity of a single memory chip (32-bit bus width each chip):

16Gbits=2GBytes

24Gbits=3GBytes

32Gbits=4GBytes

And so forth.

 

[eidairaman1]

... Still sitting here with an RX 580 and I'm still not that interested in a GPU upgrade? Granted, I'm on 1080p 60Hz, so it's fairly manageable still, but...

My two strongest arguments to buy a new GPU are:

1-If I'm gonna leave the machine doing Folding@Home, I want it to actually make the work in short time and make it worth some good points

2-My current GPU is old enough to the point where I have to start considering that it might spontaneously fail one day, so it'd be nice to have a replacement ready.

So... not much of an argument to buy a new one currently.

You do any bios mods on the RX580?

 

[windwhirl]

You do any bios mods on the RX580?

Nope, it runs stock. I've never been a fan of FAFO-ing with the VBIOS unless it's necessary to fix an issue. I did try pushing the clocks a little higher a few years ago (1400 instead of the 1365 it comes with by default), but the performance increase was practically non measurable, plus Folding@Home really disliked it (it threw errors at me right away), so it just runs at factory configuration.

 

[chrcoluk]

For me it was either 50 series or jump in on the price dropped 4080 Super to get my VRAM up to a sane level. Ultimately what made me go 4080 Super and probably now skip 50 series, is it looks like the 50 series is going to have even bigger TDP and maybe even bigger cards. I might have called it wrong and they released something at 200w outperforming previous 300w parts but I dont see it,

I think the xx80 card will have a better value compared to xx90 on launch though vs the 40xx launch. Nvidia do seem too have at least realised they got the launch differential wrong. Although it doesnt mean it wont be more expensive as they might jack the prices up for the entire line up.

 

[64K]

For me it was either 50 series or jump in on the price dropped 4080 Super to get my VRAM up to a sane level. Ultimately what made me go 4080 Super and probably now skip 50 series, is it looks like the 50 series is going to have even bigger TDP and maybe even bigger cards. I might have called it wrong and they released something at 200w outperforming previous 300w parts but I dont see it,

I think the xx80 card will have a better value compared to xx90 on launch though vs the 40xx launch. Nvidia do seem too have at least realised they got the launch differential wrong. Although it doesnt mean it wont be more expensive as they might jack the prices up for the entire line up.

My guess is only the 5090 will be impressive but it depends on how much Nvidia decides to cut down the big chip. Possibly for a 5090 Ti also or a Titan. From the rumors the 5080 doesn't look like a winner but I guess we have to wait a while to see now with the talk of delays and of course the prices which I'm hoping aren't too outrageous.

 

[JustBenching]

My guess is only the 5090 will be impressive but it depends on how much Nvidia decides to cut down the big chip. Possibly for a 5090 Ti also or a Titan. From the rumors the 5080 doesn't look like a winner but I guess we have to wait a while to see now with the talk of delays and of course the prices which I'm hoping aren't too outrageous.

Yeap, I can't possibly see the 5080 matching the 4090. But who knows what they are cooking

 

[jarablue]

Yeap, I can't possibly see the 5080 matching the 4090. But who knows what they are cooking

Hey good lookin! Whatcha, got cookin! Hank Williams.

 

[rusty caterpillar]

Will skip if they skimp on VRAM in the mid range again.

I believe they will, specially the bandwidth.

 

[chrcoluk]

My guess is only the 5090 will be impressive but it depends on how much Nvidia decides to cut down the big chip. Possibly for a 5090 Ti also or a Titan. From the rumors the 5080 doesn't look like a winner but I guess we have to wait a while to see now with the talk of delays and of course the prices which I'm hoping aren't too outrageous.

I do think they will keep a big performance gap, they dont want another 3090 situation.

But I also think the dollar performance ratio will be better than the 4080 launch, I think will be closer to the 4080 super, if that makes sense.

 

[Zazigalka]

8GB entry level, 12-16GB 5060/5070, 24-32GB 5080, 32-48 4090, and finally 48-64GB for the Titan.

5060 will be a 170w card iirc from then leak we got a couple od weeks ago. That means we finally will get a proper xx60 card that od actually an upgrade over previous gen. But if that's the case, it absolutely needs 10gb at the very least, otherwise you'll get a card that can easily run 1440p max on the GPU side but doesnt have the vram capacity for it. 10gb 160 bit on gddr7 should be same or faster than what's on 4070. 12gb on 5060 would be nicer though

 

[Crazybc]

Price is what determines if i skip the 5 series. If its stupidly high again I'll pick up a 4080 super and call it a day. Prices for graphics cards in Canada aren't exactly cheap Not compared to down south anyways.

 

[QcRMoD]

If it has to make sense an eventual 5060Ti would be a fine upgrade over a 3070Ti or 3080 for conscious budget 1440P/4K. Because if it has to make sense it would have to be 16GB and 256 bits vs. the existing 4060Ti 16GB. Then if it has to make sense, the whole 50xx series could be a very very nice one.

Alas it obviously won’t be making sense. Mostly generational/feat upgrade over existing configs it will be. Nvidia has no compelling argument but ONE to make sense: that’d be to further weaken AMD’s market while they’re not in the competition/recalibrate the general mood to something like when they launched Pascal. But they don’t have to. A thousand reasons to be conservative, primarily profits and distribution of assets over more lucrative segments of their production ie AI.