Ryzen Allegedly Matching Intel in 3DMark FS Physics

[leadeater]

4 minutes ago, That Norwegian Guy said:

Wrong, Nvidia would love for AMD to vanish as they wouldn't come under anti-trust because FTC would consider Intel competition for Nvidia as long as Intel keeps putting iGPUs on their CPUs.

 

Intel is the only one that'd get their butts plugged if AMD disappeared. They'd at the very least hurry to stop making iGPUs to take Nvidia down with them.

I doubt FTC would see it this way. They don't purely look at GPU market adoption they also look at production and competing companies. Intel is not a competing add-in/dedicated GPU manufacture so Nvidia would be subject to the monopoly clause.

 

There are also other companies that source GPUs off AMD to make their own products, Matrox for example, who would not be happy about AMD leaving the market and would file with the FTC way before the FTC would need to step in as they would do it instantly.

[That Norwegian Guy]

1 minute ago, leadeater said:

I doubt FTC would see it this way. They don't purely look at GPU market adoption they also look at production and competing companies.

And they'd quickly see that Intel has 73% market share in graphics

 

2 minutes ago, leadeater said:

Intel is not a competing add-in/dedicated GPU manufacture so Nvidia would be subject to the monopoly clause.

FTC is not competent enough to distinguish between discrete and integrated graphics, sorry.

[leadeater]

4 minutes ago, That Norwegian Guy said:

And they'd quickly see that Intel has 73% market share in graphics

 

FTC is not competent enough to distinguish between discrete and integrated graphics, sorry.

Yes they are. Not only that they don't need to be, complainant lawyers will do that job.

 

Even an idiot can pick up an AMD GPU and go "this goes in the same hole as that Nvidia card, but that Intel CPU doesn't".

 

Edit:

Don't confuse companies ability to hire very good lawyers to find legal arguments with the competence of regulatory authorities. They can only do what is within the law and no more.

[LAwLz]

Just two more weeks and then all this speculation will be over...

Just two more weeks...

 

Then we will see how close Patrick's "25% behind the 6900K/X in Cinebench R15" holds up.

 

Edit: He was banned!? Noooooo

[leadeater]

58 minutes ago, LAwLz said:

Edit: He was banned!? Noooooo

Yep, took you this long to notice? 

[Humbug]

huh? Patrick was banned? I though he must just be busy. What did he do?

he was entertaining.

[Curufinwe_wins]

FYI, 3dmark Firestrike actually shows pretty poor scaling in core count above 6 cores 12 threads...

 

Also the 5960x boost clock in this situation is super low compared to the sorts of things that the rest of the stack has.

[Hawx]

The 3Dmark physics benchmark is pretty terrible for guessing the overall performance of a CPU. It's very linear and more cores tends to automatically equal a higher score.

 

For example, a FX8350 gets a higher score than a i5 6600k, and for some odd reason a FX4300 gets higher than a i5 6500, yet we all know which CPU is actually faster.

[cj09beira]

3 hours ago, PianoPlayer88Key said:

I partly agree, but would really like to see larger improvements per unit of time.

 

As I said in my edited post above, my dad went from a 286-10 + mobo + 1MB RAM for $940 in 1989, to a 486 DX4-120 + mobo + 4MB RAM for $322 in 1995.  (The DX4-120 was $120, but the invoice from 1989 doesn't give the price of the 286-10 CPU by itself.  Per Wikipedia's MIPS page, a 286-12 was 1.28 MIPS, and a 486DX4-100 was 70 MIPS.

 

At any rate, that's basically a 78.75x (that's TIMES, not percent!; with estimated extrapolation for the actual clocks) increase in single-threaded performance, and a reduction in cost to about 1/3 or so, over about 6 to 7 years.

 

Yes, I know there's also RAM, GPU, storage & other speeds, but I still would like CPUs to "catch up" to where we would be if that rate had not slowed down.

That type of gain is over. Never to be seen again. (On silicon) from now on we will have imcremental improvements of 5-15%. The only way to increase perf a bunch would trow efficiency right out of the window.

[cj09beira]

this would mean there's a new cheap cpu to use to make Hwbot records :-)

[leadeater]

5 minutes ago, cj09beira said:

this would mean there's a new cheap cpu to use to make Hwbot records :-)

Just when I hide this double post you edit it .

[cj09beira]

2 minutes ago, leadeater said:

Just when I hide this double post you edit it .

thanks, and sorry

using the phone isn't ideal :-(

[You_are_a_cunt]

WCCF

 

My grains of salt are prepared.

 

On topic, I hope this is actually true, or at least mostly true. AMD really has to step it up a notch with Ryzen, for the sake of all of us.

[PianoPlayer88Key]

15 minutes ago, cj09beira said:

That type of gain is over. Never to be seen again. (On silicon) from now on we will have imcremental improvements of 5-15%. The only way to increase perf a bunch would trow efficiency right out of the window.

If that's the case, I guess it's multi-core efficiency instead, like the trend has been going toward?

 

Also, what about having software be more efficient so it doesn't need as much resources to run? so it feels like as big of an improvement as I was talking about in the previous post.  Bonus points for some significant improvement being able to somehow apply to already-existing software without it having to be rewritten / patched.  (This would primarily be for legacy programs that people still like to use that are no longer updated, or things like that.)

 

Oh, and speaking of legacy compatibility ... I've been hearing rumors that the next major architecture from Intel (post-Tigerlake) might do away with some of the legacy x86 instructions.  If people ran programs that still needed them, would software emulation be used, with a likely performance penalty?  Or, what about an optional PCI-Express x4 or x8 card with a low-power older x86 CPU/FPU on it, and a GPU cooler slapped onto it?

[Space Reptile]

3 hours ago, LAwLz said:

He was banned!? Noooooo

[Bananasplit_00]

isnt the 1800 pro the top model? i remember it beeing two teirs over the 1700X, so if its as good then this is going to be a crazy launch, less then $500 for 6900K preformance. if this flops AMD is dead. there is no coming back if they fail now

[PianoPlayer88Key]

18 minutes ago, cj09beira said:

thanks, and sorry

using the phone isn't ideal :-(

Okay, so maybe we won't be there with this generation, but, I'd love to have phones that have desktop-grade performance and UI.  (Better yet, workstation/server performance.)  For example, if we WERE there now, you'd have a phone with identical performance to 2x Xeon E5-2687W v4, BCLK @ 160 MHz, 2x GTX 1080 (Ti?) Classified, 4x Samsung 960 EVO 1TB in Raid 5, etc.   (Price adjusted to be where those parts SHOULD be priced, not the sky high prices they now are.)  Or, if on a budget, maybe an R7-1600X, GTX 1060 SC and 512GB 600p or something like that.  Also of course, be able to use the full desktop interface.

[cj09beira]

34 minutes ago, PianoPlayer88Key said:

If that's the case, I guess it's multi-core efficiency instead, like the trend has been going toward?

 

Also, what about having software be more efficient so it doesn't need as much resources to run? so it feels like as big of an improvement as I was talking about in the previous post.  Bonus points for some significant improvement being able to somehow apply to already-existing software without it having to be rewritten / patched.  (This would primarily be for legacy programs that people still like to use that are no longer updated, or things like that.)

 

Oh, and speaking of legacy compatibility ... I've been hearing rumors that the next major architecture from Intel (post-Tigerlake) might do away with some of the legacy x86 instructions.  If people ran programs that still needed them, would software emulation be used, with a likely performance penalty?  Or, what about an optional PCI-Express x4 or x8 card with a low-power older x86 CPU/FPU on it, and a GPU cooler slapped onto it?

well one of the problems of x86 is that its architecture is slowed down by legacy instructions that prevent some types of changes, moving away from those instructions is great, and they will probably be emulated by cpu, just taking more cycles, the apps that use them are probably light enough that it doesn't matter much, (thou it will probably matter in the server space as there are a lot of super old software still running).

The software side is complicated, making software use multiple cores well is hard, really hard, because of how the information depends many times on data on other cores, making chore you are using the right data is hard, libraries and APIs help but they don't do magic.

Cpus are probably going to keep increasing in core count, them they will probably split in two to be able to grow even more, although we will end up ditching x86 sometime in the future.

Same goes for gpus, although gpus can still increase in frequency and improve architecture, the next jump is in multiple dies (amd has it on their road-map) with will allow bigger effective die area and increased dies per wafer (lower costs)

[cj09beira]

40 minutes ago, PianoPlayer88Key said:

Okay, so maybe we won't be there with this generation, but, I'd love to have phones that have desktop-grade performance and UI.  (Better yet, workstation/server performance.)  For example, if we WERE there now, you'd have a phone with identical performance to 2x Xeon E5-2687W v4, BCLK @ 160 MHz, 2x GTX 1080 (Ti?) Classified, 4x Samsung 960 EVO 1TB in Raid 5, etc.   (Price adjusted to be where those parts SHOULD be priced, not the sky high prices they now are.)  Or, if on a budget, maybe an R7-1600X, GTX 1060 SC and 512GB 600p or something like that.  Also of course, be able to use the full desktop interface.

to get there we need a superconducter at room temperature, with would permit most of the cpu to not waste energy, would be the holy Grail of cpu power consumption,

phones are limited by the batteries, and stupid brands that refuse to make fat phones with huge batteries :-(

my best experience with a phone was when i doubled the battery of my S4, i could bo anything and not be worried, it was amazing! :-)

but then i lost the phone in 2 weeks :-(

[Tieox]

3 hours ago, LAwLz said:

Just two more weeks and then all this speculation will be over...

Just two more weeks...

 

Then we will see how close Patrick's "25% behind the 6900K/X in Cinebench R15" holds up.

 

Edit: He was banned!? Noooooo

 

I take it as proof the LMG photocopier has become sentient and is plotting world domination...

 

 

[VerticalDiscussions]

 

This is how AMD gets them scores. Pure, unadulterated raw speed no cooling applied.

 

#PlzDontKill-It. #RedTeamBetter or is it?

[leadeater]

2 minutes ago, VerticalDiscussions said:

-snip-

Damn, could you add some commentary to that post so I or another mod doesn't have to hide it. This is hilarious and needs to say lol.

[Bouzoo]

3 hours ago, Humbug said:

huh? Patrick was banned? I though he must just be busy. What did he do?

he was entertaining.

What didn't he do?

Time for @MageTank to make a thread to reverse that decision. Batman can't work without his Joker. 

[Humbug]

4 minutes ago, Bouzoo said:

What didn't he do?

Time for Mage to make a thread to reverse that decision. Batman can't work without his Joker. 

Especially Ryzen launch will be boring without him

[Bouzoo]

4 hours ago, LAwLz said:

Just two more weeks and then all this speculation will be over...

Just two more weeks...

 

Then we will see how close Patrick's "25% behind the 6900K/X in Cinebench R15" holds up.

 

Edit: He was banned!? Noooooo

If you ever miss him, you can often find him arguing in WCCF comment section, and I stumbled on him on Anandtech comments (and Overclockers forum, not sure if he's active though).