By using this site, you agree to our Privacy Policy and our Terms of Use. Close

Forums - Microsoft Discussion - Teraflops are NOT a Measurement of Gaming Performance

Pemalite said:
JRPGfan said:


If theres no drastic bottlenecks to them, and their the same architecture in core design, you CAN absolutely use Teraflops as related to gameing performance.

No you can't.

Except...
a) We don't know if everything is the same.
b) Changes in resolution changes bottlenecks in the underlying hardware.

a) For Christ's sake, we have the dieshots of the PS4 and XOne. We know both consoles use the identical core designs. Those two consoles perform pretty much exactly as the TFlops predict.

b) No, bottlenecks are bottlenecks are bottlenecks. By changing to a higher resolution (a SOFTWARE change), an already existing HARDWARE bottleneck MIGHT emerge that was not noticable in the lower resolution.

And please stop babbling around the issue. You are constantly putting forth weird examples to make straw man arguments. As JRPGFan wrote, the same architecture in core design is the key. and he is absolutely right that under that condition, TFlops is a good way of relating gaming performance.



Around the Network

TF cannot be directly compared, just like how HP can't for a vehicles engine.

If the same company, or even two different companies, use an identical third party 400HP rated engine for their truck, both will be just as fast as each other, correct?

Not necessarily. Why? A few examples.

Either truck could have:
Added a turbocharger,
A more efficient drivetrain,
Tires with stronger grip,
Better tuned suspension,
Superior aerodynamics,
Less overall weight,
A better driver,
Etc.

The initial engine HP rating didn't take this into account, much like how the TF rating of a GPU leaves many things out.

If truck A has all these things, and truck B doesn't, truck A is going to leave truck B in the dust.

Truck B may even cost about the same as truck A, because truck B might come with a luxury interior, which won't make it faster, but offers a useful non HP related perk, which is part of the overall experience.



drkohler said:
Pemalite said:

No you can't.

Except...
a) We don't know if everything is the same.
b) Changes in resolution changes bottlenecks in the underlying hardware.

And please stop babbling around the issue. You are constantly putting forth weird examples to make straw man arguments. As JRPGFan wrote, the same architecture in core design is the key. and he is absolutely right that under that condition, TFlops is a good way of relating gaming performance.

I honestly don't think you know just how knowledgeable Pemalite and CGI are, compared to the likes of JRPG fan and everyone else on here. 

For a 2007 user, I would have thought you would have noticed this years ago, but apparently not. Also, for a "65 year old male gamer", you certainly type and talk like you're of a much younger age, just something I picked up on, when going over your posts in general.

He isn't "babbling", he's using actual logic, knowledge and sense within this discussion. All the others arguing against the two are using, are Tflops this, precision that, without making much sense, let alone not going very deep into the technical aspect, which is exactly what Pem was asking for, and hardly anyone is actually providing, because they do in fact know very little and are wholly out of their league. 

Do not try to think that you know more than those two, because I swear it will take you down a road you won't be winning, nor coming back from. I would honestly just quit, instead of insulting a mod who knows much more than you do.

Last edited by Chazore - on 09 March 2020

A hugely misinformed thread.

The performance of a computer is dependent on its architecture and the microarchitecture of its components. Its incredibly complex yes, but there are clear big factors: the CPU, GPU, RAM, data transfer rate, any bottlenecks, ROM.

Two computers with similar architectures (similar not exact microarchitectures) can be accurately compared and we know ROM is SSD for both. We know both will be using relatively the same CPU. The only contentious area is the GPU and the RAM (type and amount).

In the context of gaming GPUs are hugely important. They govern most of the matrix and vector math in games (vertex calculations, rotations and translations, ray tracing for lighting and sound, shading and texture mapping, vision and projection etc). 

Type of RAM is important for the speed of data transfer, which alongside and the amount will govern how much calculations can be executed per frame: it can increase graphics for fixed amount of frames or mean higher frames for fixed amount of graphics. 

Last edited by src - on 09 March 2020

Yet if TF was everything to MS than why would they release a rumour 4TF console that has less than the XB1X?..

It seems many took too serious the marketing appoach.



Around the Network
Chazore said:

I honestly don't think you know just how knowledgeable Pemalite and CGI are, compared to the likes of JRPG fan and everyone else on here. 

because I swear it will take you down a road you won't be winning, nor coming back from. I would honestly just quit, instead of insulting a mod who knows much more than you do.

You are right that I don't know how knowledgeable Permalite et al are. I only see the meandering around the issue in their posts.

As to call to authority ("He knows much more than you do"), it might surprise you that I actually designed and built computers and parts (Gepard, for example, that brand is certainly completely unknown nowadays) before Permalite was even born. By designing and building I actually mean start with a white sheet of paper (actually mm-grid-paper), draw the layout, transfer the layout to uv-transparent foils, transfer the foil layouts onto copper boards, develop the boards (might have cost a few brain cells, that part wasn't particularly healthy), edge the boards, populate the good boards with chips, (restart if it doesn't work), the full program. Those were the good old times of hand-made computers...



Mr Puggsly said:
DonFerrari said:

Nope, you said "MS boasted power given that seemed important and it worked" ... if it didn't addressed the gap in sales how did it worked? I swear many people in VGC say MS doesn't do what it do for console warriors but to improve its bottom line, so if you don't have evidence of it improving sales how did it work?

You got me, this was clearly a sales debate. Clearly anything MS does that doesn't close the sales gap is ultimately pointless.

Remember guys, if MS does anything good like making a fantastic piece of hardware like X1X. All you have to do is say, "b... b... but did it close the sales gap!?!?" You'll always win a debate that way. Even when it's completely irrelevant to the discussion.

Don't make a strawman to cover yourself.

You were the one that said "MS boasted power given that seemed important and it worked".

I didn't say it was useless/pointless or that it isn't good to have a strong HW. But for a company the measure of something working is it bringing revenue, and that didn't happen so you can't claim it worked.

You just know you made a bad point and can't admit.



duduspace11 "Well, since we are estimating costs, Pokemon Red/Blue did cost Nintendo about $50m to make back in 1996"

http://gamrconnect.vgchartz.com/post.php?id=8808363

Mr Puggsly: "Hehe, I said good profit. You said big profit. Frankly, not losing money is what I meant by good. Don't get hung up on semantics"

http://gamrconnect.vgchartz.com/post.php?id=9008994

Azzanation: "PS5 wouldn't sold out at launch without scalpers."

drkohler said:

a) For Christ's sake, we have the dieshots of the PS4 and XOne. We know both consoles use the identical core designs. Those two consoles perform pretty much exactly as the TFlops predict.

They are not using identical core designs. Far from it.

The Xbox One has the ESRAM which takes up a rather large chunk of the transistor budget.
The Playstation 4 has extra ACE units.
The Xbox One also features the texture mapping standards from the Xbox 360 baked into hardware.
The Playstation 4 includes extra functional CU units.

Identical core designs? Common.

drkohler said:

b) No, bottlenecks are bottlenecks are bottlenecks. By changing to a higher resolution (a SOFTWARE change), an already existing HARDWARE bottleneck MIGHT emerge that was not noticable in the lower resolution.

False.
A GPU might not be bottlenecked at say... 1080P because the fillrate it has is enough to keep up at that resolution, but increase that resolution to say... 1440P, suddenly the fillrate comes up shorter, yet the GPU might still have enough compute and geometry capabilities to handle the task... So reducing things such as texture resolution may have a linear increase in performance.

There are other instances where increasing resolution and visual settings will introduce additional cache thrashing, there is a reason why AMD and nVidia are constantly re-using, re-clocking, re-balancing GPU designs in order to remove bottlenecks.

For example... AMD took the Vega 64, cut back the GPU's pixel pipelines and texture mapping units... However AMD retained the number of Render Output Pipelines and increased clockrates with the Vega 7 GPU.
Which means that AMD increased ROP performance in total by about 4.3%, reduced compute and texturing by about 6.5% and increased bandwidth by 112% as AMD's modelling showed that Vega 7's bottleneck laid in the ROP/Memory section of the GPU rather than the compute or texturing.

https://www.anandtech.com/show/13923/the-amd-radeon-vii-review

drkohler said:

And please stop babbling around the issue. You are constantly putting forth weird examples to make straw man arguments. As JRPGFan wrote, the same architecture in core design is the key. and he is absolutely right that under that condition, TFlops is a good way of relating gaming performance.

My "babbling" about the issue is usually in response to other people "babbling" about the issue.

Either way... You are not my real mom, you can't tell me what to do.

src said:


Two computers with similar architectures (similar not exact microarchitectures) can be accurately compared and we know ROM is SSD for both. We know both will be using relatively the same CPU. The only contentious area is the GPU and the RAM (type and amount).

ROM is Read-Only Memory. That is not an SSD... SSD's are based on NAND, not ROM and thus are able to perform writes.

Nintendo will typically use ROM for it's game carts for that very reason rather than NAND... Mostly because ROM isn't prone to bit flipping like NAND and thus is more reliable for data retention.

drkohler said:

You are right that I don't know how knowledgeable Permalite et al are. I only see the meandering around the issue in their posts.

As to call to authority ("He knows much more than you do"), it might surprise you that I actually designed and built computers and parts (Gepard, for example, that brand is certainly completely unknown nowadays) before Permalite was even born. By designing and building I actually mean start with a white sheet of paper (actually mm-grid-paper), draw the layout, transfer the layout to uv-transparent foils, transfer the foil layouts onto copper boards, develop the boards (might have cost a few brain cells, that part wasn't particularly healthy), edge the boards, populate the good boards with chips, (restart if it doesn't work), the full program. Those were the good old times of hand-made computers...

Irrelevant.
You don't know what my qualifications or experience is, they may trump everything you have listed by several orders of magnitude, age has nothing to do with that.

I was using Atari 2600's when they were cool. I designed 8-bit Microcontrollers and controlled robots that I designed and built.
I was programming in Beginners All-Purpose Symbolic Instruction Code on the commodore 64 creating rudimentary "games" leveraging ASCII.. And this was all  before I was even in high-school.

I was there when the hardware accelerated 3D revolution took off, which got kickstarted by 3dfx. - I was there when TnL was the big buzzword in gaming graphics with the Geforce 256, I was using PC's before Windows 95 or NT even existed with the now-familiar interfaces.

Don't assume I am ignorant when it comes to technology.

Nor does it mean yours or mine past experience has any bearing on technology aspects of today, things are different now if you haven't been paying attention... Especially as we enter the Ray Tracing era.

At the end of the day, FLOPS is irrelevant, there are so many examples I have provided where even identical GPU's with the same flops will have half the performance... And if you have as much "experience" as you say you do... You should know this as a basic fundamental factoid.



--::{PC Gaming Master Race}::--

DonFerrari said:
Mr Puggsly said:

You got me, this was clearly a sales debate. Clearly anything MS does that doesn't close the sales gap is ultimately pointless.

Remember guys, if MS does anything good like making a fantastic piece of hardware like X1X. All you have to do is say, "b... b... but did it close the sales gap!?!?" You'll always win a debate that way. Even when it's completely irrelevant to the discussion.

Don't make a strawman to cover yourself.

You were the one that said "MS boasted power given that seemed important and it worked".

I didn't say it was useless/pointless or that it isn't good to have a strong HW. But for a company the measure of something working is it bringing revenue, and that didn't happen so you can't claim it worked.

You just know you made a bad point and can't admit.

You're projecting, you made the straw man.

The X1X did work at ending much of the boasting about power from Sony fans. I didn't imply it closed the sales gap, that was irrelevant.



Recently Completed
River City: Rival Showdown
for 3DS (3/5) - River City: Tokyo Rumble for 3DS (4/5) - Zelda: BotW for Wii U (5/5) - Zelda: BotW for Switch (5/5) - Zelda: Link's Awakening for Switch (4/5) - Rage 2 for X1X (4/5) - Rage for 360 (3/5) - Streets of Rage 4 for X1/PC (4/5) - Gears 5 for X1X (5/5) - Mortal Kombat 11 for X1X (5/5) - Doom 64 for N64 (emulator) (3/5) - Crackdown 3 for X1S/X1X (4/5) - Infinity Blade III - for iPad 4 (3/5) - Infinity Blade II - for iPad 4 (4/5) - Infinity Blade - for iPad 4 (4/5) - Wolfenstein: The Old Blood for X1 (3/5) - Assassin's Creed: Origins for X1 (3/5) - Uncharted: Lost Legacy for PS4 (4/5) - EA UFC 3 for X1 (4/5) - Doom for X1 (4/5) - Titanfall 2 for X1 (4/5) - Super Mario 3D World for Wii U (4/5) - South Park: The Stick of Truth for X1 BC (4/5) - Call of Duty: WWII for X1 (4/5) -Wolfenstein II for X1 - (4/5) - Dead or Alive: Dimensions for 3DS (4/5) - Marvel vs Capcom: Infinite for X1 (3/5) - Halo Wars 2 for X1/PC (4/5) - Halo Wars: DE for X1 (4/5) - Tekken 7 for X1 (4/5) - Injustice 2 for X1 (4/5) - Yakuza 5 for PS3 (3/5) - Battlefield 1 (Campaign) for X1 (3/5) - Assassin's Creed: Syndicate for X1 (4/5) - Call of Duty: Infinite Warfare for X1 (4/5) - Call of Duty: MW Remastered for X1 (4/5) - Donkey Kong Country Returns for 3DS (4/5) - Forza Horizon 3 for X1 (5/5)