AMD CPU speculation... and expert conjecture

Page 385 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Status
Not open for further replies.
And what can I do with integer performance inside my daily life, Juan?

Cheers!

EDIT: This question reads a bit trollish, but trust me it ain't that the case. I really want to know what we use on a daily basis that is biased towards integer performance.
 


Richland is not 4.2Ghz and if you compare turbo frequencies instead base clocks the reduction in clock speed is less.

We don't know the IMC used in Kaveri and its iGPU is GCN not VLIW4, but If you believe that the iGPU in Kaveri is "massively bottlenecked" then what happen with a HD7750 DDR3? Is it ultramegamassively bottlenecked?

Kaveri iGPU is about a 33% bigger than Richland iGPU, and offers about a 30% more performance. That doesn't look as a massive bottleneck.
 




What do you do game(more FP heavy) or other things?
 


I can mostly agree with this but I think SR was 30% over BD so only ~ 20% IPC over pd. That leaves the average cpu to around 5-10% with the clock reduction.

GPU side, 25-30% would be pretty nice.
 


For one never compare turbo speeds since its not consistent
" then what happen with a HD7750 DDR3? Is it ultramegamassively bottlenecked?"
Yes it is bottleneck and quite frankly a horrible product
"Kaveri iGPU is about a 33% bigger than Richland iGPU, and offers about a 30% more performance. That doesn't look as a massive bottleneck."

Die size alone is not what matters Amd can add 43% more transistors(290x vs 280x) but yet it still only provide 25% more performance(based on techpowerup benchmarks)

So again with memory still being DDR3 its hard to say if 33% more die space will add 30% more performance without throwing "up to" statements around instead of averages(which matter way more)
 
^^ not the 7750, but even lower 7730 and still showing the DDR3 bottleneck.

Skyrim.png


http://www.tomshardware.com/reviews/radeon-hd-7730-cape-verde-review,3575-7.html
 


My compiler... my test web server... my astrophotography image processing software... compression utilities... media codec converters... plenty of stuff.
 


Isn't that just as much of a claim as saying the roadmap can't change eventhough AMD says it can change today, tomorrow or next week? Do you have any specific authorized AMD evidence to support your claim?`I would acknowledge your claim, if it was a press statement from AMD saying that this would be the way the future goes. Then it would also go from claim to fact... But as long as it is only suggested by a roadmap that has a disclaimer saying that they can change it any time they see fit, I don't really buy it, although it is indeed plausible.
 






I have no idea what you and JDWII are trying to state, Steamroller is not Piledriver because is mostly a NEW ARCH and whether Kaveri is clocked 5% slower to Richland is irrelevant, we already know Kaveri is at least 20% faster than Richland even with Kaveri having 5% less speed, as Far as i know Kaveri 7850K will have a Max Speed of 4Ghz while Kaveri 7700K will have 3.8Ghz Max Speed:

http://www.chw.net/2013/12/especificaciones-de-los-apu-amd-a10-7850k-y-a10-7700k-kaveri-dt/

And please don`t start with this crap that CHW.net is not reliable just because is a Chilenian site, that site has prove some of you orthodox and mainstream TechSites followers to have been wrong all along,

It does not matter if Kaveri has slightly slower clocks compared to Richland because is mostly a new Arch and there are some CPU only benchmarks out there that claims Kaveri is At least 20% faster to Richland, the performance is there even with Slower clocks, we already knew that going bulk would limit Max Clock Speeds but we also knew that Kaveri had promised a 30% IPC Performance increase compared to Piledriver, this may sound repetitive but let´s make it clear, Steamroller is not Piledriver and having 5% less speed will not affect final performance.

So just stop with this crap that Because it`s clocked at 4.0 Max it will lose performance, i noticed you guys try your best to make Kaveri Seem like a piece of shit, But it is not and for anyone who is stuck to a PhenomII like myself would get a 40-50 Performance increase in IPC, that very much guarantees a huge Boost and a worthy Upgrade.

I am no fanboy, i been btching about AMD for not going 6 Cores and GDDR5 as most of you may remember i was claiming to move to Intel if AMD did not pulled this right, some asked me to wait and so i did, but right now all i see is pure win from Kaveri plus HSA and MANTLE and this will be my next Gaming Rig with a R9 280X, so i don`t see why you are all so turn down with Kaveri.

You can`t Blame AMD, if it were for them they would release a 8 Core APU with 768 GCN and GDDR5 if they had a smaller node, but GlobalFoundries is AMD worst nightmare/enemy, they are stuck to this mediocre Foundry (At best) with little choice to choose someone else for the job while they just keep leeching as much as they can from AMD.

Intel lead is purely Node Advantage... i bet if AMD had the FinFet 20-14nm like Intel it would release a monster of a APU that would leave Intel biting it`s own ass out of Rage, sadly it is not the case and they are stuck to 28nm (at best) with TSMC and GloFo, yet releasing kick ass APUs at no higher than $150 USD, i don`t see Intel doing the same or even getting remotely close on that regard, they are way far behind in the performance/cost/GPU, and since most of us are Gamers i bet most of you would take an strong APU instead of a super strong CPU bundled with a slow ass GPU.

So just end this argument about Kaveri being only 15% faster on Max compared to Richland, the 20% IPC boost is there even at 4.0Ghz Max.


[Edit]

I think Juanrga claims has proven reliable most of the time, i don`t see any reason to doubt his claims, but i do believe AMD will somehow release a new FX (not Steamroller but Piledriver again) Refresh, but we may not see this new FX until late 2014... i think AMD desktop roadmap it is very clear they will stick with Piledriver til mid-late 2015, if they will stick to FX-9590 as their top CPU or with a new Piledriver FX is debatable, but seriously irrelevant... they are clearly Going APU all the way and that is future for AMD, but looks very very promising.
 
“Steamroller is not Piledriver because is mostly a NEW ARCH”
A new ARCH is Bulldozer to phenom steamroller is piledriver with improvements
“and whether Kaveri is clocked 5% slower”
Around 11% slower not 5% and surly this will lower the overall performance like any other processor when your lower you clock speed.
“we already know Kaveri is at least 20% faster than Richland even with Kaveri having 5% less speed”
Based on what benchmarks or what actual proof?
“Kaveri had promised a 30% IPC Performance increase compared to Piledriver” What instruction sets? Is this in average or upto statement?
“Steamroller is not Piledriver and having 5% less speed will not affect final performance.”
Lets be clear and say Steamroller is piledriver with improvements.
“But it is not and for anyone who is stuck to a PhenomII like myself would get a 40-50 Performance increase in IPC”
WTF extremely doubtful Phenom II is already 5-10% better in per clock performance on average compared to PD
“Intel lead is purely Node Advantage”
So an Intel I7 920 still stacking up to a 8350fx is a node advantage or how about sandy-bridge being 32nm as well?
“So just end this argument about Kaveri being only 15% faster on Max compared to Richland, the 20% IPC boost is there even at 4.0Ghz Max.”
On max? I said on average not max, max speeds don’t matter when the majority over through them?

 

This "20% faster than richland" is an open-ended statement. What metric are they talking about, flagship to flagship, clock to clock, multithreaded only (since half of SR improvements is in the shared module), raw IPC (ie single threaded).

Richland ranges from 2.1 ghz (laptop) up to 4.2 ghz, thats a large range that they can claim Kaveri is 20% faster than _________ ... wich one?

We probably won't know until Jan 14th

http://hothardware.com/News/AMD-A10-Kaveri-APU-Details-Leaked/

The problem with people always stating the MAXIMUM possibility, when it comes out and people see the minimum gains, they complain because it wasn't what "AMD SAID" even though it came from someone guessing or exaggerating what AMD actually said.

Expect low and be pleasantly suprised when its higher, or be pissed off when it under-performs expectations. Marketing did the latter with BD, thats why you don't see the same antics they pulled with that launch.
 
I seen some "Leaked" Benchmarks since a few months back, it may not be Official or Final performanche but it gives a good idea on what performance Kaveri has:

http://wccftech.com/amd-kaveri-steamroller-performance-revealed-integor-point-results-show-viable-competition-haswell/

You`re right, it seems Richland Max Speed is 4.4, i had the impression that it max Boost of 4.2 Ghz.

Vishera FX 8350 is about 10% IPC faster than the PhenomII 980, which is the Fastest Phenom... and we know Richland is 10% faster than FX-8350 and if we take into account that Kaveri is 20% faster than Richland it is a good 40-45% (or possibly more) compared to my 980.

Here is a good Benchmark test against PhenomII-FX-APUs , 955, 965BE, 980BE, Trinity, Richland, FX-4100, FX-4350, and FX-8350 against my very own 980:

https://www.dropbox.com/s/e7luobxhgidmuos/PTst.jpg

My 980 seems to have slightly lower score than default 980s but i had Steam, Fraps, PerformanceTest and Skype running in the background so that may explain why i got few dozen points less, in any case my CPU may still be within the margin of error.

I don`t know about you but that Richland APU is Almost 25% ahead of my 980... AMD already Claimed 20% Kaveri lead over Richland and thats a good 45% in IPC compared to my 980, why would i doubt AMD while they also claim they will lose against latest i5? If that`s not clear enough then... WTF... are you lacking simple Math Skills?

I could Bench a Richland APU against a 2500k in "PerformanceTest", but the Benchmarks tend to favor Intel by a good 10-15% with comparable CPUs of both AMD and Intel, PerformanceTest is known to be Biased toward intel, heck PerformanceTest claims that a Q9650 is ahead by 10-20% of my 980 in Single Threaded Benchs... we all know its BS, but Benching AMD CPUs without bringing any Intel CPUs may be accurate.

http://cdn2.wccftech.com/wp-content/uploads/2013/11/AMD-Kaveri-APU-Platform-Details.png
 


Do note that the video you referenced is running the game at 1024x768.
 



Who gives a heck! Kaveri will be a Good 40-45% ahead of my current CPU, Plus FM2+ has a good upgrade path for Excavator APU, and Kaveri will have a price of about 150$ which is heck of a deal! you will also get a somewhat good GPU bundled in the APU that blows the heck out of any Integrated GPU from Intel in any of their CPUs.
 


I don't think anyone was denying ANY of that.

40-50% is likely a bit of an overestimation; assuming your CPU is a Phenom II.
 
Well, there were benchmarks that showed the 4300/4350 sitting side by side (ergo, clock per clock) with a Phenom II. I don't recall if it was a Thuban or not, but at least showed it was on par with the Phenom line when newer instructions were used. As you can imagine, it was a game.

I'm not sure about 40-45%, but at least 20% is perfectly feasible in stock form (no L3 could amount to a 3-5% penalty for games mostly). Maybe widen the gap with some OC (bulk can't be THAT bad, right) in favor of Kaveri and you might have a winner there.

Still, I DO expect Kaveri to beat handily any Phenom II X4, including the beloved 980BE and 965BE-C3. Hope my wishes come true, lol.

Cheers!
 


It'll likely beat it. But 150% of the performance is... Really pushing the envelope.

 
A10 kaveri will be around 25-30% faster then a 980 at best but will offer a bit more features HSA,Mantle, Amd True audio, PCI-E 3.0. Well worth the upgrade from a 980 but with games being more multithreaded won't it lose when a person owns a 8350fx and a mid-range video card? Should be around 15% faster in applications that use 4 cores or less when compared to a 8350fx and the lack of L3 cache might affect the lower FPS measure.

http://hothardware.com/News/AMD-A10-Kaveri-APU-Details-Leaked/

agree 100% with his realistic expectations
 
Also notice how his slide had the words "upto" which is the same as saying this car can go 170MPH down a hill with heavy wind pushing behind me.
Anyone here that remembers anything about Amd's past know's a lot about their up to statements
 


Who knows for sure as of now, but I highly doubt the FX-4350 will even compete with Kaveri... the FX-4350 is slightly behind Richland 6800K by about 3-5% as seen on my ^PerformanceTest Benchmark, while the PhenomII 980 is almost 25% behind Richland in Single Threaded Applications, if Kaveri is in fact at least 20% faster than Richland a 40-45% performance gain sounds about right, in any case Kaveri sounds like a worthy upgrade to my 980 and i can´t wait to upgrade to FM2+ Kaveri.
 


You´re right, Clock for Clock Phenom has Higher IPC compared to Piledriver, a FX 4350 downclocked to 3.7Ghz is about 4% slower compared to the Phenom II 980 at 3.7Ghz, but a FX-4350 using it´s Max Speed of 4.3Ghz does pulls Ahead the Phenom II by at least 7-10% and Richland is about 5% faster than the FX-4350.



Noob2222, I tested several FX-4350 against my own 980 and every single one of them 4350 beat my 980 sometimes by 5 sometimes by up to 10% in all areas of Raw Compute Power, there is no single benchmark area that the 980 beats the 4350... 7-10% may not be that much, but Richland even pulls slightly further than the 4350, i really think Kaveri will be a good bunch ahead of Richland (20%). as stated by AMD.
 


Well Trinity is a bit slower compared to a FX-4300 (not the 4350):

http://www.anandtech.com/bench/product/675?vs=700

Check the Single Threaded, i also believe that AnandTech is not using the Boost features of any of the CPUs shown in their site, if a Piledriver based CPU runs at it´s default Speed there is no way it can compete against a default clocked Phenom like the 980, but there is a reason why the Visheras FX are clock higher 10-15% than the Phenom II.

http://media.bestofmicro.com/Y/Q/357650/original/skyrim-1680.png

As you can see a FX-8350 is a bit faster than the 980 and the FX-4350 is a tad faster than the FX-8350 in Single Threaded Games such as Skyrim, if Richland 6800K was present in that Benchmark it would had beaten both the FX-8350 and the FX-4350 in that particular Game.

http://cpuboss.com/cpus/AMD-FX-4350-vs-AMD-A10-6800K
 
Status
Not open for further replies.