RTX 3080 trumps 2080ti by a whopping margin

Used GPU market gets flooded with Nvidia's RTX 2080 Ti cards immediately after RTX 3000 announcement, high-end Turing prices could drop to around $300 notebookcheck.net

Theoretically speaking, if the RTX 3080 priced at $699 is 70-80% faster than the RTX 2080 Ti, the high-end Turing cards should be selling for around $300 tops in used condition. We are not quite there yet, as the lowest prices in the US are hovering around $550, while in the EU, brand new RTX 2080 Ti cards are selling for 450 Euro. The race to the bottom is still on, nevertheless.

No wonder why many RTX 2080 TI owners are now rushing to sell their cards on ebay and Amazon. The lowest price on the US ebay for a used RTX 2080 Ti is around $550, while prices in Europe already dived to 465 Euro (for a brand new card). Knowing that the RTX 3080 is going to cost $699 and is expected to be around 70-80% faster than an RTX 2080 Ti in ray traced content, prices for the used Turing cards should probably bottom at around $300.

Alternatively, we have the RTX 3070 priced at $499 which is expected to have similar performance compared to the RTX 2080 Ti on the raster side, while being considerably faster in ray traced content, so $300 or even lower for the top Turing cards in used condition is really not that far fetched.

 

GTX 980 was 165W, not 275W, and it had 33% more cores than 980M.

 

Y'all lack imagination lol.

With a repaste the GE75 can already handle a 200w 2080 and an i9 10980HK. And the heatsink design is far from optimised.

If laptop manufacturers are forced to, they will come up with something that can handle 300w+ in a medium sized chassis.

Heck my GS75 handles 200w just fine and it's just a 2.3kg thin and light. Imagine quadrupling the size of the heatsink. It's well within reach guys. Laptop manufacturers just want to cut costs wherever they can.

 

I have to agree with seanwee. Are we going to see laptops with mobile 3080s coming close to the desktop counterparts...I think so and history has shown they will come reasonably close unless your headspace is in the late 90s early 2000s

 

Ok

 

Maxwell Pascal Turing all reasonably close..not like one was double the other.

Also the 3090 has over 10 000 cores vs 5000 how do you figure 20 percent gains on over double the cores.

 

3080: 8704 CUDA cores
3090: 10496 CUDA cores (+20%)

Performance also doesn’t scale linearly with core count.

 

With Ampere (RTX 3080) we will see the gap increase further...

 

Either this or to make themselves look better, nvidia will crank up the mobile tdps accordingly which is what I'm hoping for.

Ie
2080 (215w) -> 2080 laptop (150w) 70% -> 2080 Max-q (80w) 37.2%

3080 (320w) -> 3080 laptop (225w) 70% -> 3080 Max-q (120w) 37.5%

 

They did nothing for Super. Why you think they will do it for Ampere? Because its a new arch? Nope.

2080 Super (250w) -> 2080 Super laptop (150w).

 

Precisely because its a new architecture. How is that so hard to believe? Even if its not the values i suggested nvidia has to be incredibly stupid to not change tdps for a new architecture.

 

When Ampere goes mobile I'mma be right over here with my 10 pound laptops that an handle an actually decent TDP.

 

That’s the upside to having a beefy burrito. You have thermal headroom to work with. People can’t expect the same level of performance between the Area51M and a thin and light gaming notebook ala Razer Blade or MSI GE. There is only so much copper and fans you can stuff into something of that size.

RTX 3080 laptops (beefy bois excluded) should be ecstatic if they can match desktop 2080 performance.

 

I'm already within 10% of desktop 2080 performance on my thin and light so if thats all they can do it will be sorely dissapointing.

If the laptop 3080 doesnt match desktop 3070/2080ti performance its a failure.

 

Nvidia releases more RTX 3000 spec details through Reddit Q&A session

Regarding the confusing double CUDA cores, Nvidia states that the new Ampere microarchitecture was designed “to achieve twice the throughput for FP32 operations compared to the Turing Streaming Multiprocessors (SM). To accomplish this goal, the Ampere SM includes new datapath designs for FP32 and INT32 operations. One datapath in each partition consists of 16 FP32 CUDA Cores capable of executing 16 FP32 operations per clock. Another datapath consists of both 16 FP32 CUDA Cores and 16 INT32 Cores. As a result of this new design, each Ampere SM partition is capable of executing either 32 FP32 operations per clock, or 16 FP32 and 16 INT32 operations per clock. All four SM partitions combined can execute 128 FP32 operations per clock, which is double the FP32 rate of the Turing SM, or 64 FP32 and 64 INT32 operations per clock.”​

 

Which means?

 

I think he's saying 10 000 is really 20 000 or even 40 000 cores compared to Turing..which means this is the first time I buy a flagship since the 6970m days...you simply put are outta your mind not to buy the 3090 since it will probably have a 4k longevity of 4 years and 10 years at 1080p 1440p being the sweet spot.

I'm usually right when new cards come since the 8800 days..but this time I have my doubts...hard to tell what Nvidia will do with such a high tdp. I think you guys are blatantly wrong comparing apples to apples...Nvidia will figure a way to close the gap

 

3090? Nah. 3080 is where its at.

 

Nvidia chose to market the 3090 to gamers, but unless a 3090 will benefit you professionally, or you're related to Jeff Bezos, don't buy it. It makes very little sense to spend $1500 on a GPU just to play games. Get a 3080 at almost half the price, and then upgrade again a few years down the road. The best way to future proof is to save some cash now so you have more to spend on upgrades in the future.

 

Yeah actually you guys are probably right about the 3080 but then again the latest and greatest has come at a premium

 

Latest and greatest scam that is

 

There's a lot of people with the cash to burn that just want the best. But those people are the minority. I could easily buy one many times over but there's no way I'm spending that much on a graphics card, I want to retire at some point lol. TBH a lot of people are going to just be waiting until October for the 3070 or even looking forward to 3050 and 3060. I'm guessing we probably won't see a card without RT in this gen.

Even the 2080Ti is going to be a good used buy if they end up staying lower than the $499 they are asking for the 3070. By the looks of EBay they are still going for $600-750 which seems crazy, you'll be able to get a 3080 for about that much soon.

 

Yup, that's me, waiting for a 3070 because that looks like its price-to-performance ratio will make it the best value. I could buy something higher, but I don't see there being enough benefit. And I'm still keeping my fingers crossed that AMD might soon announce an even better deal than the 3070.

 

3080 has better price-to-performance ratio than 3070. It has 48% more CUDA cores, 70% more memory bandwidth, and 2GB more VRAM while costing 40% more.

 

Redesign as I linked. Doubled Cuda core counts (I call it marketing) doesn't mean double performance. Neither do HT for processors.

NVIDIA GeForce RTX 3080 Flagship Is 68% Faster On Average Than RTX 2080 In OpenCL & CUDA Benchmarks, Up To 2X Faster In Some Cases

 

Do you guys think the reason there is no 3080ti is because NVDIA is reserving it just in case AMDs BIg Navi beats 3080? It seems that team green is following the release path of the GTX series.

 

Enjoy... http://forum.notebookreview.com/thr...ers-welcome-too.810490/page-819#post-11044272

 

It's definitely a strategy. They can't launch too many products at once. There's a few reasons for that. I'm guessing the 3090 is priced more than 2x the cost of the 3080 simply due to yield. Any chip that doesn't make it as a 3090 is just used with disabled parts of the arch as a 3080. I'm guessing 3070 yield is probably decent and they will launch a 3060 or 3070 Super/Ti when they build up enough silicon for those depending on yield. I can't see them launching a 3070Ti with just 8GB more memory as has been rumored, there has to be more to it than that at 3070 performance levels. You only really need the memory for going above 1440p realistically.

There's always multiple reasons: yields, marketing strategy, competitive analysis, revenue potential. They know what they are doing, and they know there's a segment that will overspend on the 3090 to have THE BEST, doesn't matter if it's more than 2x the cost of the 3080. Pricing on the 3070/3080 seems reasonable and probably driven by how much they know about what AMD is going to do.

EDIT: Oh, I forgot they are also leaving room for a 3080Ti/Super.. obviously..

 

I should have said the 3070 likely has the "best value" for folks who don't want to break the bank--supposedly beating 2080ti performance for less than half the price, and having a much more reasonable energy cost compared to the 3080 and 3090. Still, it's possible the 3080 will have a better ratio. Either way, I'd wait for the benchmarks to make sure, since everything is mostly guesswork until they come out.

 

With 3090 at $1500 they have made room for 3080 Ti at $1000/1200 depending on how “Big Navi” Radeon graphics do it. Same also between 3070 and 3080. Perfect strategy to keep average prices up.

 

Really curious since I'm planning on getting the 3090 but I don't want to have scenario if when the 3080TI comes out, Nvidia would advertise it as better than the 1500$ card in gaming. Correct me if I'm wrong that's what happened when 1080ti arrived, it was beating the Titan in-terms of performance on majority of the games tested.

 

I don’t think they will kill 3090 as the big boy. Only if AMD is able to compete above 3080. And I doubt that will happen.

 

I highly doubt that as well. Especially in RT. They are still a generation behind there. RT+DLSS is truly a stake in the heart right now. They might catch up at some point. But TBH they really only need to compete at the 3050 to 3070 or MAYBE 3080 level as that's where the bulk of the sales volume is.

 

Yeah. But Nvidia need to either provide better performance or reduce the price point. And we know Nvidia rather prefer put in the Ti tag than have to reduce the prices on current cards

AMD Big Navi Radeon 6000 With 16GB Allegedly Priced To Aggressively Undercut RTX 3080 hothardware.com

War is usually hell, except when we are talking about the GPU wars, of which a new skirmish is about to get underway. Hot on the heels of NVIDIA formally introducing its GeForce RTX 30 series based on Ampere, it is rumored AMD will look to undercut its rivals pricing with the release of its Radeon RX 6000 series, with...

 

We shall see. Speculation is fun, but I prefer to just let shipping products speak and tend to discount rumor and prediction. Really the only sane play by AMD is to try to undercut them for similar performance. nVidia has the mind share and frankly they have more stable products from a software standpoint and more mature features overall. Everything has bugs, building software and hardware is immensely complex, but Nvidia wins there right now. I hope to see AMD take it to them though, we need the competition.

 

Spot on bro Paps. I bet my left nut Nvidia will release a 3080ti. I'd go so far to even bet they'll even eventually release it even if AMD don't have anything competitive anyway. They can't help themselves.Milked to perfection

 

They are a company trying to make money, makes sense. They've obviously left a hole in the product offering there for a 2080Ti/Super. They'd be idiots not to. They'll give it a little time until they can create another splash, and as well get the yields in the proper place for it to make sense to generate revenue from it. Hey, all that has to happen to get nVidia to drop their prices is either the products don't sell or someone competes with them aggressively enough to force their hand.

I'm going to take a guess that neither one of those things is happening soon. People will buy them and AMD will still be behind far enough that it doesn't happen.

 

It will most likely happen IMO. There is not a big difference between 3080 and 3090 in terms of performance. A 3080ti with bigger memory pool and slightly improved specs will surely be within the 3090 performance at a lower price.

But it won't really "beat it" just be close to it. Similar to 2070 super to the 2080.

 

Sounds about right. There aren't really any gamers that need a 3090, but at EDIT: $1500 "need" has nothing to do with it unless you require the 24GB for creative or professional work. The 2080Ti was definitely a "gamer" card, but it seemed more about recouping R&D costs for all the work that went into getting to 1st Gen RT/DLSS since there's enough people that would pay handsomely for that level of performance either out of wanting the best or just for benchmark bragging rights.. whatever the motivation. And like the 2080Ti, it seems like no one is going to compete with them at that flagship level. I'm sure there's enough people that will get a 3090 for that reason as well, but I'm bargaining not as many since the 3080 looks so good. The 2080 wasn't nearly as attractive.

 

Maybe fake, high overclocked RTX 2080 Ti, not optimized drivers or the 100% more CUDA cores is more as real marketing bluff. Pick you'r own poison.

NVIDIA GeForce RTX 3080 4K AotS benchmark leaks out videocardz.com

 

That seems like it lines up somewhat correctly, given that's not really a great way to do this test.

EDIT: I just took a look at the site, they used a 9900K, it could be that AotS is stacked with a bunch of Ryzen results.. Ryzen performs better in that bench than Intel.

I'd like to see head-to-head instead of a single results vs median.

 

So one point I saw raised earlier in this thread was the RTX 3080 has a 320bit memory bus - something which (I believe) we've never seen in laptops before mainly due to PCB layout constraints whether it be MXM or the smaller form factors required by laptop motherboards (compared to desktops). I'm not convinced a 320bit memory bus can be worked into laptop designs (as they currently stand). By comparison, look at NVLINK - never made it to laptops simply because there wasn't enough PCB real estate to accommodate that bus.

I'm drawing two conclusions. Either:
1) MXM is completely dead

or

2) Laptops won't get anything better than the RTX 3070.

Thoughts?

 

Design changes. Compact PCB package is possible also for MXM formfactor.

 

I think MXM stopped being standard a while ago, since every manufacturer started going with the shape they wanted. Manufacturers maybe will still use a highly modified MXM port or they will migrate to their own proprietary form factor like Dell did. But I do believe that we are going to see better solutions than a desktop 3070 on laptops

 

NVIDIA GeForce RTX 3080 Graphics Card Benchmarks & Game Performance Tests Leak Out, Allegedly 30% Faster Than The RTX 2080 Ti wccftech.com | Today

The full breakdown of the tests is provided below:

• 3DMark Fire Strike Performance: 31919 (+25% 2080Ti, +43% 2080S )
• 3DMark Fire Strike Extreme: 20101 (+24% 2080Ti, +45% 2080S )
• 3DMark Fire Strike Ultra: 11049 (+36% 2080Ti, +64% 2080S )
• 3DMark Fire Strike Time Spy: 17428 (+28% 2080Ti, +49% 2080S )
• 3DMark Fire Strike Time Spy Extreme: 8548 (+38% 2080Ti, +59% 2080S )
• 3DMark Fire Strike Port Royal: 11455 (+45% 2080Ti, +64% 2080S )

 

The RTX3080 performs pretty good. These numbers are better than a watercooled and heavily overclocked 2080Ti even at a frequency locked at 2,130Mhz in games and benchmarks. It looks like the RTX3080 is running stock and on air though haha.

I remember when my 2080Ti broke 40K in Firestrike graphics, and 17,200 in timespy graphics. I thought it was the most amazing thing ever. And now we’ve got a $699 RTX3080 that runs right past even the fastest 2080Ti by around 5%, doing it all stock and on air cooling. It’s gonna be nice to see how far the RTX3080 will go. I am certain it’ll be a monster once on water cooling with a good overclock.

I am certainly getting one. I bet we will get those short waterblocks that look really cool. Kinda like the RX480 did on watercooling.

 

If you mean 30% better performance with 100% more CUDA cores and 25% higher TGP is pretty good ... I would say nope.

 

That's only half of the equation though. It's also half the cost so i'd say its a pretty good deal.

 

Well, my 2080Ti getting about 5% less performance than a stock RTX3080 was pulling around 450 watts playing games. So, it was certainly power hungry too.

Maybe this RTX3080 is running at a pretty tame clock speed. I’m sure if we cool it down and get that clock speed up some and that power limit, we will start to see the performance easily scale up.

 

I don't know about you guys but i'm excited for the next gen ports. it looks like a 1060 will handle the new next gen games at 1080p and you will need a 3070 or greater for 4k. at 299 the xbox series s will be dominating the market and lucky for us pc gamers we wont need to upgrade for the ports and geesh did I not see that coming