RDNA4

hkultala said:
With new manufacturing technologies, SRAM is becoming more expensive. Making big on-die cache is expensive and is gettin more expensive. And also, making the L2 bigger also makes it slower.

The only way to have relatively cheap big cache is to put that cache on a separate die which is made on older manufacturing process(what AMD did in RDNA3). But that increases latency and power.

One reasonable direction might be keeping the outer level cache on a separate die, but trying to minimize the overheads of the die-to-die traffic, for example by integrating the dies vertically, cache die below or above the logic die. Something similar than what AMD did on the Zen-3D/v-cache.


Apple has fast access to their DRAM because the memory controllers on the same die (which is costly on new mfg processes) and also because they use LPDDR line of memory which is more latency-optimized, less bandwidth-optimized than GDDR line of memory.

And Apple can afford to pay for the cost of having the memory controllers on-die because they are selling their products at very high price, they have good margins anyway. Consumer GPUs have much less margins than Apple products and AMD has to try to save more on mfg costs.
Click to expand...

On the flip side resolutions aren't really getting higher. So do you need bigger Sram caches for infinity cache ? Wouldn't the original 128megs and under still be large enough for their purposes ?
 
eastmen said:
On the flip side resolutions aren't really getting higher. So do you need bigger Sram caches for infinity cache ? Wouldn't the original 128megs and under still be large enough for their purposes ?
Click to expand...

Big enough to hold all render targets and the BVH at the same time would be pretty helpful, that's as much as 512mb @4k right there, if the BVH is enormous

Also, leak? Really easy to fake leak

https://twitter.com/x/status/1796254005389312451
 
Last edited:
Kaotik said:
Like said it's fake, but it's 0.9x faster, not 0.9x as fast, so that would still mean 90% faster
Click to expand...
It's not 0.9x faster. The 1.9x is 'faster' by 1.9x, but we are dealing with a starting point of 1 to note the level of base performance, not 0. So 0.9x would be 0.9x as fast.

Because if the base comparison point was 0, then clearly all these data points would make no sense at all, and just be 0.
 
Seanspeed said:
It's not 0.9x faster. The 1.9x is 'faster' by 1.9x, but we are dealing with a starting point of 1 to note the level of base performance, not 0. So 0.9x would be 0.9x as fast.

Because if the base comparison point was 0, then clearly all these data points would make no sense at all, and just be 0.
Click to expand...
Mathematics would like to have a word with you. Something x faster is literally something times faster than something. Not something times as fast. And it's fake, it doesn't have to make sense.
 
Maths and marketing are very, very different branches of human developmental thinking.

The word "faster" needs to die in a fire, started by an asteroid strike, just after a zombie invasion prompted by a plague of fleas.

The phrase should be "the speed of". ie. "1.9x the speed of the shit you already own". So 0.9x is "90% the speed of the thing you already own" so yeah, slower.

But it's marketing. What can you do? I still haven't got over them lying to me about the butter substitute. Yes, I can f**king believe it's not butter, it tastes like wallpaper paste.
 
Kaotik said:
Like said it's fake, but it's 0.9x faster, not 0.9x as fast, so that would still mean 90% faster
Click to expand...
The chart, fake or not, clearly shows the bar for MW3 going below the baseline, so it is indeed saying that MW3 is 10% slower. And yes, whoever wrote this worded it poorly. They should have said "up to 1.9x as fast".
 
Kaotik said:
Yes, but then you don't say it's "faster", you say "as fast". If it's faster, it's baseline + whatever times is the number, if it's as fast, it's baseline times whatever
Click to expand...
It's a bad chart. The 'faster' part refers only to the 1.9x part. Because yes, that's a positive increase, aka faster. But 0.9x would be slower. Baseline x 0.9 is less than the baseline.
 
Seanspeed said:
It's a bad chart. The 'faster' part refers only to the 1.9x part. Because yes, that's a positive increase, aka faster. But 0.9x would be slower. Baseline x 0.9 is less than the baseline.
Click to expand...

Yes the chart is bad. However “faster” isn’t a multiplier in the way you’ve described it. The corresponding language for a multiplier is “as fast”. Marketing slides have been butchering the use of “faster” for ages though so much that it’s now become accepted language.

90% faster isn’t the same as 1.9x faster.
 
trinibwoy said:
Yes the chart is bad. However “faster” isn’t a multiplier in the way you’ve described it. The corresponding language for a multiplier is “as fast”. Marketing slides have been butchering the use of “faster” for ages though so much that it’s now become accepted language.

90% faster isn’t the same as 1.9x faster.
Click to expand...
1.9x faster is fine.

It just needs to be understood that anything >1 is faster, and anything <1 would be slower.

It's not the best way to word it obviously, but it also shouldn't be too hard to understand what it's getting across.
 
You must log in or register to reply here.

Similar threads

F
RDNA5 Speculation
Replies
2
Views
1K
Erinyes
E
A
Holy Wall of Text: My impressions post Playstation 4 (PS4) unveil
2 3 4
Replies
70
Views
20K
(((interference)))
I
A
NVIDIA Q2 Results
2 3
Replies
47
Views
10K
Razor1
R
A
  • Locked
NVIDIA Q3 Earnings + CC
2
Replies
32
Views
9K
Arun
A
A
NVIDIA Analyst Day: Webcast Report (+PSP2 Hints)
2
Replies
36
Views
9K
Geo
G
Back
Top