The top-end RTX GPU struggles to hit 60fps 1008p(Update)People areMADabout NVIDIA RTX... jayz 2 cents think this is why

@GiveMeSomething: The 2070 will be slightly faster than a 1080. A 2080 will be slightly faster than a 1080ti. And the 2080ti will be easily the strongest GPU on the market. I fail to see how its a downgrade. You can say you expected a larger jump in performance if you'd like, but calling it a downgrade is completely false.

Lol, people going off the rails over a new GPU with completely new tech and under developed drivers, running a game that isn't finished.

That's what I been saying in another thread. This isn't me about defending Nvidia, but this is of course a new GPU we're all talking about. We still gotta wait for actual Benchmarks before writing it off. If the benchmarks proves to be great and surpassing the 1080Ti, I'll consider upgrading next year of course when the hype is dying down and hopefully the prices will be reasonable for a purchase.

I'm just remaining optimistic on the whole RTX.

I'm certainly not going to rush out an preorder one of those cards, that's for sure, but I'm not going to dismiss them off hand because of some random guys tweet at a reveal conference.

Ray tracing is new and taxes the gpu, it takes common sense to know this. Not buying a 2070 or 2080 for ray tracing. I’m buying it so I can run 4K 60 on a single card.

I’m not gonna buy at launch probably wait a few months and just upgrade cpu and mobo for now, then around Xmas I’ll try and pick one up with my bonus.

Can anybody agree with me it's a good time for AMD to come in and fill the void?

It always is but seems AMD can't be winning both the CPU and GPU fronts. If it competes in one, it's losing the other. Kinda understandable because Intel and NVIDIA operate in just one of these markets and are both richer than AMD. Once upon a time it was AMD/Intel. ATI/NVIDIA.

The benchmark numbers look really good. I have no clue what people are bitching about TBH.

The RTX series does 50% better in 4k using current rendering methods and does 75% better using DLSS.

Seems well worth it in my book. Not sure what planet people are living on where that is bad.

I agree the prices are too high, but people are willing to pay it, so what are you going to do?

What's the problem?

GPUs are made with functional blocks e.g.

1. Geometry. Rasterization mass floating point to integer pixel conversion units

2. Compute Shader units, The TFLOPS marketing, which are useless without read/write units.

3. Texture read/write units, texture blending units. Collectively, these are TMU (texture management units), Compute Shader usually uses this read/write path.

4. Graphics read/write units (ROPS), graphics blending units (ROPS). Collectively, these are ROPS (raster operation units), Pixel Shader usually uses this read/write path. This is the classic GPU hardware.

Workloads like crypto currency uses point 2 path. Vega 64 is competitive with this path against GP102 (GeForce GTX 1080 Ti). AMD GPU optimized games are bias towards this path.

Traditional gaming workloads are bias towards point 1 and point 4 paths. This is where Vega 64 is inferior to GTX 1080 Ti i.e. Vega 64 has GP104 (GTX 1080) level hardware with GP102 level compute hardware.

False, Your Crytek link doesn't mention the actual method for ray trace approximation reflection.

https://www.techpowerup.com/gpudb/3268/radeon-pro-vega-20

AMD Radeon Pro Vega 20 has 2Ghz boost mode clock speed which is 30 percent higher than the old Vega 64. It's compute units and classic GPU hardware performance is 30 percent higher due to clock speed overclock. Without factoring 1TB/s memory bandwidth, it's performance estimation is somewhere GTX 1080 Ti to near Titan V levels.

NAVI has GDDR6 for mainstream gaming markets. AMD CPU engineers are helping RTG engineers for higher clock speed.

NVIDIA has superior delta color compression with it's ROPS.

Pascal's low hanging fruit is L2 cache bandwidth. Hawaii's 1MB L2 cache has 1 TB/s bandwidth(not connected to ROPS) while GTX 1080 Ti's 2.88MB (2883584 bytes) L2 cache is just above 1 TB/s (connected to ROPS). NVIDIA could improve this section.

TU104 has 4 MB L2 cache(?) while TU102 has 6 MB L2 cache. GP104 has 2MB L2 cache. TU's tile cache rendering function could been improved.

Dude stop it, you're a fraud.

You're the real fraud. I'm just running CUDA memory benchmarks fool.

Hint: GTX 980 Ti has inferior (~2/3 from R9-290X) L2 cache bandwidth when compared to R9-290X at 1Ghz, but AMD GPU can't use it for ROPS read/write. My GTX 1080 Ti has overclock version results with L2 cache bandwidth. DDC is not use for crypto currency.

You're the fuking fraud.

https://wccftech.com/nvidia-geforce-rtx-2080-turing-tu104-gpu-pcb-exposed/

WCCftech claims TU104 has 4MB L2 cache which is twice of GP104's 2MB L2 cache.

https://gist.github.com/nelson-liu/623eb54d977c98db005eaf2fbc449238 has data on GTX 1080 Ti's L2 cache size which is 2883584 bytes or 2.88 MB. DCC can yield up to twice effective storage, but it's not use for crypto compute.

You f*koff.

@ronvalencia: You already quoted me lol. Is it because you were lying the first time and got caught with your pants down? Stop lying and pretending you know what you’re talking about.

@Juub1990: I don't get any of what he says and I don't think any of it matters in raytracing or something.

JayzTwoCents isn't wrong. Just wait for the reviews, benchmarks, and don't pre-order at all!

@jasonofa36: Because he’s saying nonsense as usual.

I will most likely be skipping this generation of cards. The ray tracing techniques are in their infancy and the jump in rasterization is not enough from my 1080TI is just not there. The prices are just insane as well. I hope the 3080TI will be better. Hell, at this point I am waiting to see what Radeon brings to the table. Most likely not enough to make me jump. But hey, one could hope.

That's whats needed anyway...at least we will have Intel also soon ...triforce! :P

Ummm....what? They explicitly mentioned 'approximate ray-traced HDR reflections'. LMAO, the DC. Fact is RLR is universally calculated like that, GG did nothing new, the only difference is Crytek didn't really hype it as ray-tracing. Sony did to woo their brain dead fanbase in thinking they are doing something new. That exactly what happened, cows begin bragging about it as 'ray-tracing' without knowing who did it first and that it existed for years.

Here's an article that caught into Sony's hype machine but later corrected it. Actually Crysis 2/3 implementation was more advanced because it featured actual area and volumetric lights unlike KZ:SF that said they will use area lights but the actual game uses little to no area lights.

Now coming back to the word 'ray-tracing', neither Crysis 2/3 nor KZ:SF's implementation is 'ray-tracing' in a traditional sense, you're using rays in screen space not in world space. Rays have been in use in games for decades examples includes shooting ray to find out the line of sight for NPC's, occlusion, distance between NPC's and Player Character.

I loved the AMD/nvidia era

Even though the amd equal to the 6800 ultra was solid.

PS5 it is!

https://videocardz.com/newz/nvidia-upgrades-l1-and-l2-caches-for-turing

NVIDIA improved L1 and L2 cache for Turing GPUs e.g.

L1 cache bandwidth improved by 2X,

L2 cache storage improved by 2X.

I'm right and you're wrong. I'm right with NVIDIA going after Pascal's low hanging fruit i.e. cache.

https://wccftech.com/amd-confirms-new-7nm-radeon-graphics-cards-launching-in-2018/

Radeon Vega 20 gaming card with 20 TFLOPS FP32. Without CU count increase, this Vega 20 would need about 2.45 Ghz clock speed which also improves the classic GPU hardware by about 56 percent.

Took you 10 days to come up with a response. Pathetic lol.

wccftech links should be banned.

I dont understand this new breed of PC gamers whatsoever. Crysis was a slideshow on the 8000 series but everyone was excited for DX10 at the time despite that. Cards got better and games ran better. This isn't a new thing, but seeing gamers revolt against new tech sure af is.

I think this is more about the price of cards than it is about the tech. High end cards are priced to the moon these days, so gamers make impossible demands on them. It's a bad situation that could stunt progress.

Everyone who says "I just want high framerates" is going to make technological progress slow to a crawl.

Agreed.

If it needs to be done it needs to be done.

But, given the current climate where only now GPU prices are where they were TWO YEARS AGO, and RAM is still extortionate, I would hope gaming companies would of at least thrown a "catch up" bone to all those people that simple can't put that kinda money down on a graphics card.

I think it was the wrong time to go "Oh hey, know that GPU drought? well we are going to make it last longer by stagnating old card prices and selling new £1000+ cards with new features and low performance gains per £.

I honestly shocked if PC Gaming Growth hasn't stalled in the last 1-2 years based on everything going on.

I'l just stick to the 1080 TI until these Ray Tracing cards mature.

@ronvalencia: Everything you just said is meaningless.

Everything you just said is useless.

https://www.techpowerup.com/gpudb/3233/radeon-instinct-vega

Radeon Instinct Vega 20 already has 2025 Mhz clock speed at ~225 watts. That's ~32 percent clock speed increase from Vega 64's 1536 Mhz at ~295 watts.

Classic GPU hardware performance increased by 32 percent along with CU's TFLOPS via clock speed increase.

AMD switched from GoFlo's 14 nm LPP Vega 10 into TSMC's 7 nm LPP Vega 20.

225 watts --> 295 watts = 31 percent increase

225 watts --> 400 watts = 77 percent increase

2025 Mhz --> 2445 mhz = 20 percent increase.

RX Vega 64's TFLOPS increase should have been tied to actual memory bandwidth increase instead of relying on tile cache rendering and delta color compression improvements with Fury X's memory bandwidth.

Haven't you been absolutely owned multiple times and left the forums in small increments in time?

You post graphs, numbers, stats and nothing every comes to fruition.

I'll just wait until you exile yourself again.

BS, Xbox One X has near Vega style ROPS design i.e. 2MB tile cache rendering and there are game results rivaling GTX 1070 and higher than RX-580/R9-390X OC.

Vega 56/64 has unified 4MB L2 cache for CU and ROPS

X1X's GPU has 2MB L2 cache for CU and 2MB render cache for ROPS (different design from Polaris uArch). Tile render from CU to ROPS capable. Polaris can only tile render via CU not ROPS.

LOL you're still going on that.

You continued "Haven't you been absolutely owned multiple times" as your "you're still going on that." Hypocrite.

GS forum is not mandatory.