RIP <=8 GB Vram

[caliberlife]

Weird question guys.

On my z97 motherboard with my 5700xt the total vram my system tells me is 12gb (8gb dedicated and 4gb shared) is this a good thing?

 

[Grim5]

Weird question guys.

On my z97 motherboard with my 5700xt the total vram my system tells me is 12gb (8gb dedicated and 4gb shared) is this a good thing?

Sort of same thing here - I’ve 11gb dedicated on the 2080ti and 32gb system ram - but task manager reports like 20gb for the 2080ti with 9 of it shared. I’ve never seen memory go over 10gb on the 2080to so I don’t think it means much

 

[TNA]

Need to compare the texturing quality on AMD vs Nvidia, because on geforce cards they often look less sharp and a bit washed out. Which could be related to performance and vram requirements.

Surprised that no green shade and leather jacket wearing people took issue with this comment

 

[ANDREW GREEN]

https://gamegpu.com/rpg/ролевые/greedfall-test-gpu-cpu

Would be interesting to see 1% lows

 

[andicole0]

Weird question guys.

On my z97 motherboard with my 5700xt the total vram my system tells me is 12gb (8gb dedicated and 4gb shared) is this a good thing?

Yes it means you're not using a console
Andi.

 

[Poneros]

Yeah take a look at charts, GTX 1060 3GB can handle Ultra texture pack just fine with 5GB less which matched RX 580 8GB performance at 1080p and 1440p, at 4K it able to matched RX 590 8GB performance. GTX 1060 6GB, GTX 1660 6GB, GTX 1660 Ti 6GB and RTX 2060 6GB can handle it just fine with 2GB less than 8GB cards.

VRAM did not mattered.

Those benchmarks are not real. They're as if someone recorded fps looking at the sky. Until they do more than boot up the first 10 minutes and record a 30s segment of an empty room, they're simply not valid (or in RUs case the beginning outdoor area, perhaps the least taxing scene possible). Like I said, city loops are the benchmark standard for the game and that's when you'll see how the GPUs get used.

Edit: Should also add, camera rotation needs to be part of the benchmark run as well.

 

[Grim5]

I think Gamegpu just uses data extrapolation - aka it's estimates, that's how they can "test" 5 years of graphics cards in a day.

I don't know their exact method, but maybe they just bench a handful of cards, like 3 or 4 and use the known relative performance to calculate all the other cards.

 

[Kaapstad]

I've addressed this on the first page. If you don't have enough vram the frametimes plummet. So clearly using. Hence why I got a stable line with HBCC but not without.

Without HBCC, it looks like this (at best) whenever you move camera, and gets worse with time (also note the extra 3GB RAM usage vs the OP!):

Why do you keep on about HBM2 it is not superior to GDDR6 when gaming.

Titan V uses HBM2 and is very fast when gaming but up against one of my RTX Titans it is the Turing card with GDDR6 that has the better minimums.

 

[Nasher]

HBM is the better memory. The only reason you don't see it on more gaming cards is because it's expensive. Nvidia squeezing more profit is the most likely the reason for it not being on the latest Titan.

 

[Kaapstad]

HBM is the better memory. The only reason you don't see it on more gaming cards is because it's expensive. Nvidia squeezing more profit is the most likely the reason for it not being on the latest Titan.

You are missing the point.

RTX Titan (GDDR6) is far better at maintaining good fps and avoiding lows than Titan V (HBM2).

HBM is good memory but not for gaming.

 

[Nasher]

The Titan V is 2 years old and based on an old architecture though?

 

[Kaapstad]

The Titan V is 2 years old and based on an old architecture though?

The memory it uses is HBM2 which is not old and cards performance is the same as a 2080 Ti. NVidia still sell the Titan V as there is still nothing on the market better for what it is used for.

 

[Poneros]

Why do you keep on about HBM2 it is not superior to GDDR6 when gaming.

Titan V uses HBM2 and is very fast when gaming but up against one of my RTX Titans it is the Turing card with GDDR6 that has the better minimums.

I didn't say that and I have no clue how you got to thinking that. I'm talking about HBCC which relates to memory availability and not framerate (except insofar as not having enough vram will obviously lead to framerate instability).

But since you've broached this subject let me assure you: HBM2 is absolutely the better of the two for just about anything except saving money. Discussing memory through a comparison of two cards with wholly different architectures is misguided because other variables are going to interfere with performance.

 

[shankly1985]

Until high bandwidth cache controller is used on another type of Ram then HBM is the better solution.
What is not good is the price.

Am with OP here I have seen first hand the benefits that HBCC bring.

 

[Stanners]

Surprised that no green shade and leather jacket wearing people took issue with this comment

Your comments seem intent on goading people. Please refrain

 

[robfosters]

Just posted on another thread that I think 8GB is not enough to run Gears 5 at top settings. I was getting out of memory crashes on my V64.

 

[Nasher]

You shouldn't get out of memory crashes unless something is wrong with the game. Probably a memory leak.

 

[Gregster]

Just posted on another thread that I think 8GB is not enough to run Gears 5 at top settings. I was getting out of memory crashes on my V64.

Possible bug or memory leak in truth. 8 GB might well not be enough in some games though at 4K but would need more delving and stringent testing.

 

[robfosters]

What is a memory leak please?

 

[Gregster]

What is a memory leak please?

Where the VRAM just keeps increasing in useage. Happens in games on rare occasions and could be a game/driver/system bug.