AMD CPU speculation... and expert conjecture

[Master-flaw]

A lot of people don't want a 360...they want to play Dota or Starcraft after checking their Email.

yes, but i'm willing to bet an arm and a leg that those people are not looking forward to p@wn people while keeping visual settings low-medium. i've been playing dota 2 for a long time now and playing it @ 720p or lower at fluctuating FPS really hurts the experience. and that 384 GCN cores will not be enough to satisfy any competitive multiplayer gamer out there. my 7750 ddr3 which is twice as fast didn't well do well enough.

I don't think the word P@wN is in the demographics vocabulary...lol
They're good for light gaming and having the GPU horsepower is very convenient.
I'd go A8/A10 long before I5/I3 if i didn't game as much as I did. Having the GPU horsepower is much more useful than the CPU power as I don't need an I5 for MS office or web-browsing, but the ability to run HD movies smoother or the option for light gaming is enticing. The GPU/CPU ratio makes more sense. Especially for a laptop.

 

[gamerk316]

CPU test of headline has some bugs to work out. HTT is mucking with the CPU results, like I thought it was.

Look at it logically: If I have 4 cores, and I do the same amount of work in 4 threads, you would expect linear absolute performance gains up to 4 threads, correct?

Taking a look at two 2600(k) runs:

http://www.headline-benchmark.com/results/c309c1f4-6f00-4def-8127-b8d679fe5989
http://www.headline-benchmark.com/results/93a871e9-4a76-42aa-8f91-ef08fedef7dc

Notice the various dropoffs in performance as threads increase? That shouldn't be happening, unless multiple threads are getting stuck on the same core, OR a HTT core is getting used before a "full" core, and the shared frontend is causing massive performance drops.

Compare that to this run:

http://www.headline-benchmark.com/results/359025ad-a71c-4df1-a21e-2dcac265a956/87080cac-179d-4a2c-a2a9-cb1e90667a3f

Where I disabled the HTT cores, and compared against my last benchmarks. Note how the performance dropoffs vanished, and in some cases, performance improved by DISABLING cores?

Now, we still have an issue where performance stalls. What I *think* is going on, is multiple threads are going to the same core, so performance flatlines, then jumps by double the next time a thread gets assigned, maxing at about 5 threads (remember: HTT cores disabled). So in the case of the Integer test, I think the treads are being assigned like this:

Thread 1: Core 0
Thread 2: Core 1
Thread 3: Core 0/1
Thread 4: Core 2
Thread 5: Core 3

Hence why performance stalls on three threads, then increases linearly up to 5 (which doesn't make sense for a quad). Same thing likely happening in the FP test:

Thread 1: Core 0
Thread 2: Core 0
Thread 3: Core 1
Thread 4: Core 1
Thread 5: Core 2

Based on the huge jump going from 4 to 5, I'd suspect one of the other threads gets reassigned to core 3 at some point, but I'd need to break out my low level debug tools to prove this.

Ags1, I'd recommend that you manually assign threads to cores for this test. Start with the Even (Non HTT/CMT) cores, then proceed to the Odd ones. Once you have more threads then cores, don't bother anymore (really doesn't matter at that point). That should remove the Windows scheduler from affecting performance...Right now, due to HTT having a much higher performance cost if used wrong, Intel's numbers are probably grossly understated in your benchmark. [Can't wait to see the response to THAT one.]

[Can't wait to see the response to THAT one.] +1

It's a limitation of Java that I can't assign threads to specific cores. I have to rely on the default OS scheduling. Trust me, I've tried. Still, it's interesting that Microsoft gets it pretty wrong. And relevant to the 90% of users stuck with the Windows scheduler.

The issue does not affect any scores however, as I only count the single-threaded and maximum performance data points. I used to include intermediate threading results, but stopped that and recalculated everything over a month ago after feedback from Yuka and gamerk.

You are however seeing what I think is a real effect, where the assymetric complexity of hyperthreading is difficult for the OS to schedule optimally, whereas the FX chip presents 8 roughly equivalent cores, which is easier to schedule. I.e. This may be a real advantage of the FX architecture that you can't see simply by looking at max theoeretical performance.

Regarding the number of threads, I go to logical processors + 4 because on Windows this squeezes a bit more performance from the CPU. I think what happens is the extra threads squeeze the CPU time of background processes. On linux, performance tends to peak at threads = cores.

Jed, Regarding any bias in my postings, the results come in at different times and not all results are available at the same time. But then, I have an actual website that shows to the whole world that Intel is on top, so what are you complaining about? Also, all CPUs on the site are shown compared to the top (i.e. Intel) CPU. How is that biased to AMD? Also, this an AMD thread, so I only tend to post AMD-relevant results.

The issue with scheduling comes down to what threads are running at the EXACT INSTANT the nth threads are kicked off. If the core is being used by a OS level task, then your app won't bump it, and a HTT core could get used. And for anything math heavy, that's not a good thing. AMD's CMT has a similar problem, but its not nearly as bad (only ~20% performance loss, versus possible negative gains with HTT)

But yeah, that's the downside to HTT, and why for any tasks that are very math heavy, I tend to mask the thread to only non-HTT cores. Using C/C++ of course And bench many, many times to make sure I'm not mucking anything up along the way.

Worst case, you can invoke the OS API's manually to get the job done, but you'd have to preempt the Java threading model. Might want to see if anyones come out with a good wrapper class for this...

Also FYI, I get a 10.94 CPU score with only one core active. Tried to submit, but I'm getting a 503 "Over Quota" message.

 

[elemein]

Nevermind this post.

 

[de5_Roy]

i read the first page of that chinese link showing a10 7800 benches. okay i just looked at the images, didn't understand a single letter of text. the spec table looks incorrectly constructed, seems based on richland specs. even if it's real, some of the gaming benches just don't make sense. in the cpu benches, the cores look like good upgrades over pd but in gaming benches the whole apu is barely better than 6800k. what's up with that? it screams bottleneck. this is the first time i've seen a 65w kaveri bench with a proper-looking model number attached to it.

 

[juanrga]

Cazalan The Haswell i5 only gives about a 10% better scores than the 2500k in those benchmarks.

de5_Roy Nice to see that you continue ignoring what is said. Well done!

 

[Master-flaw]

http://www.dailyfinance.com/2014/01/03/can-amd-win-the-high-end/

This is what I was assuming they where doing...
Bleeding the market with familiarity with AMD. It spreads to developer optimization as well as popularity.

Problem with this article is is they say Mantle is basically for consoles and carries over to PC's...anyone who's looked into Mantle for the slightest can say that this isn't the case at all. it's solely for P.C.'s.

 

[szatkus]

Difference between A10-7800 and A10+R7750 is similar to gap between Radeon 7750 DDR3 vs GDDR5.

http://ht4u.net/reviews/2012/msi_radeon_hd_7750_2gb_ddr3_test/index36.php

That GPU is starving with DDR3.

 

[juggernautxtr]

http://gpuboss.com/gpus/Radeon-HD-7750-vs-GeForce-GTS-450

yeah that 450 is faster........definately not with a celeron behind it.
a SB celeron won"t even play dcuo online.

TO ADD INSULT TO INJURY
http://cpuboss.com/cpus/Intel-Celeron-G540-vs-AMD-A10-6800K

an A6 4400m dual core will play dcuo, i had a celeron dual core laptop it won't even load dcuo.

 

[blackkstar]

Some folks @ S|A forums are raising some points (I'm assuming they can read Mandarin).

1. The Kaveri is an ES
2. The reviewer has problems with thermal throttling.
3. Comparing 100w CPU to 65w CPU.

Seems like there's room for optimism but I'm still reserved.

 

[de5_Roy]

Some folks @ S|A forums are raising some points (I'm assuming they can read Mandarin).

1. The Kaveri is an ES
2. The reviewer has problems with thermal throttling.
3. Comparing 100w CPU to 65w CPU.

Seems like there's room for optimism but I'm still reserved.

the first one is probably correct. the second and 3rd ones - only if stock cooler is used and the whole apu get's loaded. xbitlabs showed that 65w apus throttle when loaded. i assumed it may be the case, but i had a stronger suspicion on how the tester got a 65w apu when only 2 kaveri models were announced for launch - 7850k and 7700k. and there were the consistent inconsistency with the presented data.
oh, another thing, that chinese site (and a couple of others) showed win 7 64bit being used. pclabs.pl showed that for cpu-bound gaming, win 8.1 64bit will be significantly better (in their bf4 mp testing) than win 7. i don't know if it affects kaveri as well but i started to ignore gaming benches (among other benches) with win 7 from then onwards (for amd cpus).

 

[juanrga]

Some folks @ S|A forums are raising some points (I'm assuming they can read Mandarin).

1. The Kaveri is an ES
2. The reviewer has problems with thermal throttling.
3. Comparing 100w CPU to 65w CPU.

Seems like there's room for optimism but I'm still reserved.

2 and 3 were mentioned in this thread before... in this same page.

Backstar, you missed point number 4: they didn't use final drivers...

 

[juanrga]

Some folks @ S|A forums are raising some points (I'm assuming they can read Mandarin).

1. The Kaveri is an ES
2. The reviewer has problems with thermal throttling.
3. Comparing 100w CPU to 65w CPU.

Seems like there's room for optimism but I'm still reserved.

the first one is probably correct. the second and 3rd ones - only if stock cooler is used and the whole apu get's loaded. xbitlabs showed that 65w apus throttle when loaded. i assumed it may be the case, but i had a stronger suspicion on how the tester got a 65w apu when only 2 kaveri models were announced for launch - 7850k and 7700k. and there were the consistent inconsistency with the presented data.
oh, another thing, that chinese site (and a couple of others) showed win 7 64bit being used. pclabs.pl showed that for cpu-bound gaming, win 8.1 64bit will be significantly better (in their bf4 mp testing) than win 7. i don't know if it affects kaveri as well but i started to ignore gaming benches (among other benches) with win 7 from then onwards (for amd cpus).

You don't need to guess about the first point. The reviewers stated that they used an ES and that will repeat the tests with the final silicon was available. The million dollar question is what memory settings used for the tests. Up to now all the leaked benchmarks used ultralow memory to hide real performance of Kaveri.

Throttling is observed in the benchmarks.

That W8.1 increases performance of APUs by 5-15% compared to W7 was mentioned in this thread many times. Kaveri is an APU.

 

[griptwister]

poor amd. they still believe that casual gamers are people who play triple A titles at crap-to-low settings... casual gamers play with their tablets and their smarthphones and these silly apus are completely useless.

my celeron SB ($30) + gts 450 ($60) , beats any AMD apu setup out there.

ROFL!!!

As someone who has owned a GTS 450 and owns a A8 system. You're WAY wrong. Even with PD core... Celerons aren't impressive... Except for the little Quad core thing they got... that either is a baller, or will be one. Lol.

I've recently decided to build a mATX rig so FM2+ sounds like the best option for me

 

[jdwii]

poor amd. they still believe that casual gamers are people who play triple A titles at crap-to-low settings... casual gamers play with their tablets and their smarthphones and these silly apus are completely useless.

my celeron SB ($30) + gts 450 ($60) , beats any AMD apu setup out there.

Keep telling yourself that,

then prove me wrong. a 7750 ddr3 (i had one which i got dirt cheap but i decided to sell it for the same price as this gts450 brand new at a discount, is a tad faster) yields faster gaming performance than the top of the line 6800k chip.

Huge market for people who want a cheap OEM with the ability to play a game or two.

"huge" yet pc sales are declining.

between playing games at low to medium on PC, i'd stick with an xbox360 instead (case in point: skyrim). see, the problem with these APUS is that it's fixed per segment, you get the lowest CPU+GPU for the cheapest, fastest CPU+gpu for the most expensive one.

give me a 4 to 5 ghz single module / 384 (gpu)core APU chip at half the price of the 6800k and you probably have a winner. design a mother board that doesnt sport a pcie slot make it BGA but sell the board at a massive discount make it micro ITX and you have winner. put it all together sell it at $120 (call this a5-6xxxk)), and that makes sense. this way you can have a disposable steambox that'll last 2 years to play previous gen steam games. bank on mantle and it'll probably last 3.

another cool idea is to design a single module / 768 gpu core chip APU chip. it doesn't have to be top of the line x86 core + top of the line gpu core.

i just don't get it with AMD. i just don't.

For 1 its a wrong way to look at it never use sales use retail prices when comparing i could of gotten a A10 6800K for 100$ one day which makes your prices seem like a not so good of a deal.

Also i'll take a 4 core processor over a Cereron anything, to me Celerons are complete rubbish

Second http://appleinsider.com/articles/13/12/29/npd-chromebook-sales-outperform-macbooks-in-commercial-sector-as-ipad-loses-ground

Funny Has anyone every thought of people not needing another laptop or desktop? Since a PC from 2006 is still good enough for Youtube+Web browsing and a PC from 3 years ago can play 1080P easy. Yet Desktops and laptops still count for 61.9% market share.

 

[jdwii]

A lot of people don't want a 360...they want to play Dota or Starcraft after checking their Email.

yes, but i'm willing to bet an arm and a leg that those people are not looking forward to p@wn people while keeping visual settings low-medium. i've been playing dota 2 for a long time now and playing it @ 720p or lower at fluctuating FPS really hurts the experience. and that 384 GCN cores will not be enough to satisfy any competitive multiplayer gamer out there. my 7750 ddr3 which is twice as fast didn't well do well enough.

Low-medium settings still beat a 360 anyways

 

[jdwii]

http://gpuboss.com/gpus/Radeon-HD-7750-vs-GeForce-GTS-450

yeah that 450 is faster........definately not with a celeron behind it.
a SB celeron won"t even play dcuo online.

TO ADD INSULT TO INJURY
http://cpuboss.com/cpus/Intel-Celeron-G540-vs-AMD-A10-6800K

an A6 4400m dual core will play dcuo, i had a celeron dual core laptop it won't even load dcuo.

http://www.xbitlabs.com/articles/cpu/display/celeron-g540-g440_6.html#sect0

Lol are you kidding me it can barley beat a old athlon II a A10 6800K would laugh

 

[juggernautxtr]

http://gpuboss.com/gpus/Radeon-HD-7750-vs-GeForce-GTS-450

yeah that 450 is faster........definately not with a celeron behind it.
a SB celeron won"t even play dcuo online.

TO ADD INSULT TO INJURY
http://cpuboss.com/cpus/Intel-Celeron-G540-vs-AMD-A10-6800K

an A6 4400m dual core will play dcuo, i had a celeron dual core laptop it won't even load dcuo.

http://www.xbitlabs.com/articles/cpu/display/celeron-g540-g440_6.html#sect0

Lol are you kidding me it can barley beat a old athlon II a A10 6800K would laugh

thats even worse than my sarcasm

 

[logainofhades]

poor amd. they still believe that casual gamers are people who play triple A titles at crap-to-low settings... casual gamers play with their tablets and their smarthphones and these silly apus are completely useless.

my celeron SB ($30) + gts 450 ($60) , beats any AMD apu setup out there.

ROFL!!!

As someone who has owned a GTS 450 and owns a A8 system. You're WAY wrong. Even with PD core... Celerons aren't impressive... Except for the little Quad core thing they got... that either is a baller, or will be one. Lol.

I've recently decided to build a mATX rig so FM2+ sounds like the best option for me

Yes, it is a shame that there aren't any 970 or 990fx M-ATX boards out there. I have the latest revision Gigabyte GA-78LMT-USB3 with my FX 8320 in my spare AMD rig. It is all I could afford when I went to Microcenter for their $100 FX 8320 deal. I I multiplier bumped it to stock FX 8350 speeds, since the board supports the 8350, but I have no intentions of going any higher.

 

[juggernautxtr]

my youngest son want to build a gaming machine, to get him started putting in an A10 on fm2+ with an h50 water cooler, gonna cross fire it with a decent card thinking 7750 since other have had success.
waiting to see if kaveri will be evolutionary or revolutionary.
even if only evolutionary with good improvement will spring for that. other wise dropping a 6800k

 

[The Q6660 Inside]

A 450 is mediocre by today's standards. A 7750 beats it (and the 460/550ti) fair and square and thus the 7850K will perform like a 460/550ti.

 

[jdwii]

http://gpuboss.com/gpus/Radeon-HD-7750-vs-GeForce-GTS-450

yeah that 450 is faster........definately not with a celeron behind it.
a SB celeron won"t even play dcuo online.

TO ADD INSULT TO INJURY
http://cpuboss.com/cpus/Intel-Celeron-G540-vs-AMD-A10-6800K

an A6 4400m dual core will play dcuo, i had a celeron dual core laptop it won't even load dcuo.

http://www.xbitlabs.com/articles/cpu/display/celeron-g540-g440_6.html#sect0

Lol are you kidding me it can barley beat a old athlon II a A10 6800K would laugh

thats even worse than my sarcasm

Yeah all you have to do is look up facts

 

[juanrga]

Stop the off-topic war folks, there is plenty forums at toms where you can discuss your celeron vs amd thing.

This is HSA in action for skeptics

This is a financial compute test. The review also mentions that JPEG decompression is accelerated by about 90% thanks to HSA. And Steamroller CPU decompress JPEG about 10% faster than Piledriver. Therefore Kaveri is close to 2x faster for decompression.

Here Richland CPU is 43% slower than Kaveri CPU

In footnote 2 of my article I wrote: "Steamroller could be up to 40% faster (approx. 35% from module plus 4% from memory subsystem) than Piledriver." :-D

Another 23% IPC gain

The Cinebench R11.5 and Excel 2010 scores are odd. It is interesting that both are tests optimized for Intel chips.

 

[Reynod]

Guys jdwii is correct ... stop trolling with the Celeron rubbish or I'll send you on a holiday from toms.

A discussion on the benchmarks from the other sites is more of interest.

 

[szatkus]

This looks interesting.

Faster even than K10.5@3.5GHz. Although SuperPi is really poor benchmark that result is quite surprising, AMD doesn't care about x87 instructions. Maybe it's side-effect of some fundamental improvements.

 

[juanrga]

szatkus It seems to be a side effect from the L1 cache optimizations in Steamroller.

tourist I discussed ARM when people asked me about it, and also to understand the server and desktop roadmaps (aka I used AMD plans to explain why there is no Steamroller Opteron and, as consequence, why there is no Steamroller FX). Also Kaveri includes an ARM core inside that makes security stuff: TrustZone. Therefore it is more on-topic than your ridiculous Celeron vs APU and the previous i7 vs 8350.

The hypocrisy here is that people who complained against me because ARM was off-topic, has been posting in those off-topic topics without complaining.

Or postings here are on-topic for everyone or for nobody, not just when some of you want it to be.