AMD CPU speculation... and expert conjecture

[noob2222]

@juan

Your arguement about only looking at 720p resolutions kinda negates the other arguement that kaveri = high end gaming doesnt it?

As it turns out this whole time you were wrong in assuming that kaveri could = i5 2500k. Now that the reviews are out maybe you could admit something instead of complaining about how no one listens to you and putting everyone else down for not listening to you trying to convince the world that this APU direction is the way of the future because it is soo fast that AMD wont need any other products.

In the end going bulk screwed kaveri, and GF screwed AMD.

Are you kidding? Looking at 720p if one is doing a "CPU benchmark" is irrelevant to using Kaveri to game at 1080p or another resolution.

You must call "review" to what anandtech did. I don't.

I am still waiting for the review of Kaveri, when it was out I will compare with my predictions using the same software. At this moment I am neither optimistic nor pessimistic. However, I recall that some pages ago I wrote

http://www.tomshardware.co.uk/forum/352312-28-steamroller-speculation-expert-conjecture/page-213#12380437

Taking this review as base [...] Kaveri would score ~9.5, which is at the i5-2400 level and only ~8% behind FX-6300 and i5-2500k. It looks very close to my prediction ;-)

The same site has finally measured 9.08 (i.e. 9.1) for Kaveri.

This means that Kaveri is a ~11% behind instead a 8% behind. Yes, I see that this is a giant error of about 3% in my prediction about Kaveri in this concrete benchmark. LOL

AMD has a clear direction towards APU, that you don't like it doesn't change anything. And the issue about why bulk was the correct choice explained to you plenty of times, No need to repeat again.

margins of error don't work that way. If you put richland as the baseline for your claims, your margin of error is 47%.

8.26 to a predicted 9.5 with an actual of 9.08 is a 47% marginal error, not 3, since your baseline was not 0.

Take your Juanrga.com claim of kaveri going to hit 10.3 then the margin of error is 72%.

 

[noob2222]

Looking at hardware canucks gaming pages, kaveri manages to squeak by the fx 4300 at 720p just barely, but not once take out the slower clocked fx 6300. Not once is kaveri 30% faster than the 5800k at 720p.

Add price into the equation and APU are not for gaming with a discrete card. IGP is solid compared to richland, lower clock and higher performance.

Considering the lower clocks of both the cpu and gpu, power consumption is not impressive.

http://www.hardwarecanucks.com/forum/hardware-canucks-reviews/65031-amd-kaveri-a10-7850k-a8-7600-review-27.html

 

[Lessthannil]

Wow. Richland 2.0: overhyped then underdelivered also with it being overpriced ($175).

Now we know why AMD won't give FX Steamroller. The improvements (if any) just don't carry over in games. The A10 5800k still now remains the only A10 that has a competitive price. The per core performance is still lower than it needs to be at and the FP strength (or the lack thereof) is honestly sad.

At $175, most gamers will just go to Intel's stuff which (still) murders Kaveri in all areas relevant to them except IGP performance... but who doesn't get a dedicated GPU when you spend that much?

"bu-but muh Mantle and HSA". Mantle is late to it's own party and HSA naïvely relies solely on outside support (just like Mantle). Especially with this happening, don't count your chickens before they hatch.

I wanted AMD to have a comeback, but it still seems like business as usual.

 

[logainofhades]

You can get a 760k and an HD 7770 for a similar cost to that of the top Kaveri. APU's still just are not worth it to me.

 

[Reynod]

Juan if your referring to memory wall as in Patterson's three walls then you still don't understand palladin's point above ... he is talkign about the cache memory latency and efficiency on the die ... your referring to the size of the memory.

For other's here is the basic premise, courtesy of a very good article by Rich Pell, referenced below:

"Power Wall + Memory Wall + ILP Wall = Brick Wall"

- The Power Wall means faster computers get really hot.
- The Memory Wall means 1000 pins on a CPU package is way too many.
- ILP Wall means a deeper instruction pipeline really means digging a deeper power hole. (ILP stands for instruction level parallelism.)

http://www.edn.com/design/systems-design/4368705/The-future-of-computers--Part-1-Multicore-and-the-Memory-Wall

No. I am not referring to the size of memory. I was referring to what Palladin explained. Also by power wall I don't mean the above definition. I am using the terms as used in the exascale supercomputer research community.

These are general terms all engineers in the industry who design microprocessors are familiar with ...

 

[jdwii]

So on average 10-15% for the CPU in other words under the 8320(8350) level CPU and under the I5 2500K by at least 20%. GPU performance is around 20% faster for GPU tasks.

So i was wrong i stated 15% on average for CPU tasks and 25% on average for GPU base games, i'm guessing something is going on with either memory bandwidth or memory latency and possibly huma affects that in a negative way.

All the software benefits are huge if companies decide to use them all and i'm sure Intel will have support for it as well. I5 sandy is a better CPU hands down as well as a 8350fx compared to the top A10.

And if building a gaming platform picking a A10 over a 6300-6350fx would be nonsense unless you want to use it without a video card or your concerned about the AM3+ platform and future benefits

a10-7850k<I5 2500K with no GPU based acceleration
a10-7850k= I3 Ivy with no GPU based acceleration
a10-7850K> I7 with GPU acceleration
a10-7850k< Price/Performance 6350-6300fx

 

[griptwister]

Kaveri's fine...
The opening prices are not.

This. lol.

@jaunrga, yeah... you over shot the benchmarks a bit... just a little bit...

 

[8350rocks]

@juan

Your arguement about only looking at 720p resolutions kinda negates the other arguement that kaveri = high end gaming doesnt it?

As it turns out this whole time you were wrong in assuming that kaveri could = i5 2500k. Now that the reviews are out maybe you could admit something instead of complaining about how no one listens to you and putting everyone else down for not listening to you trying to convince the world that this APU direction is the way of the future because it is soo fast that AMD wont need any other products.

In the end going bulk screwed kaveri, and GF screwed AMD.

@ etoryus

You are correct. AMD doent like to see APUs benched against cpus, thats why its hard to even find it put against even a fx 4xxx cpu.

We FINALLY came into an agreement... after all the back and forth argues with you, and not only Juanrga was wrong, i was too... ill give you both (You and Gamerk) that.

@Yuka - I am on AM3 board that does not support AM3+ CPUs... that´s why i am so persistent with the 980, i got no upgrade path unless i buy a new MotherBoard either AM3+ or FM2+, and at this point i thought i would move to Kaveri if the performance was right, so i am clueless as to what Socket to upgrade to, Intel is out of the Question.

In any case I don´t think Kaver is a complete Fail, the CPU side kinda is... but the whole APU is very easy on the eye, i am thinking about recommending Kaveri for my Cousins who play games like Burnout Paradise, San Andreas, Left4Dead2 and such on a very very crappy Dual AMD CPU at 2.2 and a iGPU Radeon HD3200, i am not even sure which CPU model is it, they could easily upgrade to a new system without spending too much on the GPU and get massive amounts of Performance Increase and being able to play modern games on Medium the less.

*cough*

@juanrga: still waiting for the "you were right and I was wrong" post from you...

I will probably never see it...but I still know it, and so do you, even if you never admit it...

 

[-Fran-]

Wow. Richland 2.0: overhyped then underdelivered also with it being overpriced ($175).

Now we know why AMD won't give FX Steamroller. The improvements (if any) just don't carry over in games. The A10 5800k still now remains the only A10 that has a competitive price. The per core performance is still lower than it needs to be at and the FP strength (or the lack thereof) is honestly sad.

At $175, most gamers will just go to Intel's stuff which (still) murders Kaveri in all areas relevant to them except IGP performance... but who doesn't get a dedicated GPU when you spend that much?

"bu-but muh Mantle and HSA". Mantle is late to it's own party and HSA naïvely relies solely on outside support (just like Mantle). Especially with this happening, don't count your chickens before they hatch.

I wanted AMD to have a comeback, but it still seems like business as usual.

HSA, MANTLE and some other new stuff from AMD are as useful as QuickSync was when Sandy was introduced (I won't even talk about BCLK, lol). And you wouldn't say QS was useless, right? Well, to be honest, I think it is pretty useless, but that's me, haha.

Anyway, like I said, Kaveri is not a high performance part just because we all want it to be. It's just 53% (was it?) CPU and 47% GPU. Until the HSA promise and new software is delivered, it will be a bystander for high performance. Midstream if you OC the crap our of it (currently), but kind of defeating the purpose of HTPC or SFF.

I really don't understand why we even mix it with the i5 or the i7; they're more expensive CPUs and well, they're CPUs not APUs (palladin can give you all a deeper explanation for that).

This reminds me of something "funny" (ironic more than funny, but meh...). When a friend got his brand spankin' new Athlon64 3200, he had to keep his GeForce 2 MX200. We had a good laugh when he actually got like 2X the FPS's using Software Rendering in UT than OGL at the same resolution. My point here is that... Ok, Intel currently has that same effect in most programs because their CPU side is that good while they still have mediocre (upgrade from craptastic, lol) graphics solution to off load work in it and doing some black magic in their drivers. If you also take the software side of things, its even gloomier for Kaveri and Richland.

I really want to see Kaveri inside an HTPC roundup against Intel and even Tegra 4 or Snapdragon 800. Thanks to AMD, Intel has done a very good job polishing their drivers for video playback and other quirks they had for HW acceleration (included QS).

Credits due, where credits owed.

Cheers!

 

[wh3resmycar]

instead of banking on MANTLE, AMD shouldve just partnered with valve. i'm quite sure that with some modification in steamOS you can probably create an OS that fully utilize all these AMD cpu features.

 

[etayorius]

Ill just wait till DDR4 hits to what ever new AMD Socket will be on the high performance, i guess i waited 3 years since Bulldozer and i can wait little more than a year again i guess, i was tempted to go FX 8320 but nah... DDR4 will be supported in the Beggining of 2015 and AMD may just change Sockets soon enough.

Anyone knows if a 980BE OCed to 3.91GHz will Bottleneck a R9 270x?

 

[palladin9479]

Kaveri's fine...
The opening prices are not.

The 65W part is really cheap, $119 USD is perfect for gaming on a small box in the living room. The 95W is WAY overpriced, it should be $150 USD, $160 at most. It's going to require a bigger / louder cooling solution though.

Really don't see what people are getting worked up over, it's a fx4 chip without L3 cache and with a bolted on low ~ moderate power GPU. It's a big upgrade to the GPU side not the CPU one, which is what I was expecting honestly.

 

[palladin9479]

http://www.guru3d.com/articles_pages/amd_a8_7600_apu_review,1.html

Some interesting benchmarks, they are using the same build I would recommend, mini-itx A88X with 1866/2133 (I'd stay with 2133 honestly) memory and the 7600. Really strong gaming performance at 720p without a dGPU so good for set-top style PCs or a really cheap desktop. Absolutely not for gaming a high resolutions or bigger systems.

 

[juanrga]

Spoiler

@juan

Your arguement about only looking at 720p resolutions kinda negates the other arguement that kaveri = high end gaming doesnt it?

As it turns out this whole time you were wrong in assuming that kaveri could = i5 2500k. Now that the reviews are out maybe you could admit something instead of complaining about how no one listens to you and putting everyone else down for not listening to you trying to convince the world that this APU direction is the way of the future because it is soo fast that AMD wont need any other products.

In the end going bulk screwed kaveri, and GF screwed AMD.

Are you kidding? Looking at 720p if one is doing a "CPU benchmark" is irrelevant to using Kaveri to game at 1080p or another resolution.

You must call "review" to what anandtech did. I don't.

I am still waiting for the review of Kaveri, when it was out I will compare with my predictions using the same software. At this moment I am neither optimistic nor pessimistic. However, I recall that some pages ago I wrote

http://www.tomshardware.co.uk/forum/352312-28-steamroller-speculation-expert-conjecture/page-213#12380437

Taking this review as base [...] Kaveri would score ~9.5, which is at the i5-2400 level and only ~8% behind FX-6300 and i5-2500k. It looks very close to my prediction ;-)

The same site has finally measured 9.08 (i.e. 9.1) for Kaveri.

This means that Kaveri is a ~11% behind instead a 8% behind. Yes, I see that this is a giant error of about 3% in my prediction about Kaveri in this concrete benchmark. LOL

AMD has a clear direction towards APU, that you don't like it doesn't change anything. And the issue about why bulk was the correct choice explained to you plenty of times, No need to repeat again.

margins of error don't work that way. If you put richland as the baseline for your claims, your margin of error is 47%.

8.26 to a predicted 9.5 with an actual of 9.08 is a 47% marginal error, not 3, since your baseline was not 0.

Take your Juanrga.com claim of kaveri going to hit 10.3 then the margin of error is 72%.

I see you continue kidding...

In the link given above

http://www.tomshardware.co.uk/forum/352312-28-steamroller-speculation-expert-conjecture/page-213#12380437

I predicted that hardwarecanucks would measure ~9.5 for Kaveri in that benchmark and they measured 9.08 (which rounds to 9.1). The error is of about 3%, it is not 72% neither 47%.

 

[juanrga]

Juanrga facts are starting to fail, look he has even resorted to stating things such as " that main goal of AMD behind APUs is not cheap gaming" note that this is used to overcome his information previously about steamroller being 20% more powerful on average(which he used to say 30% but will deny it) and 30% more IGPU performance.

I said 20--30% IPC, which translates to ~10--20% faster CPU than Richland. S|A measured 8% average over Richland. Thus, I was wrong about this one by ~2--12%.

iGPU is a 28% better than Richland in average. I was wrong by ~2%.

http://semiaccurate.com/assets/uploads/2014/01/Kaveri-benchmarks-complete.png

 

[Lessthannil]

You can get a 760k and an HD 7770 for a similar cost to that of the top Kaveri. APU's still just are not worth it to me.

Agreed. This setup will do the same song and dance as the 7850k for the same or less for better GPU performance. You don't have to worry about pampering it by buying high speed RAM which can carry a noticable markup over standard 1333 or 1600 or having system RAM being siphoned for the APU to use.

It does seem like the A8-7600 is a lot less absurd in pricing as its bigger brother. Its 45w TDP is wonderful and it being priced at around $119 is where an APU should be IMO.

Once you give it a $175 price tag, it doesn't belong. I think at this CPU price point that most people will get a dGPU right off the bat anyway. That renders the IGP useless outside of acceleration. Then, the FX 6300+ and Intel i3 3220+ outspeeds the CPU portion of the APU often for less money. And then, the final nail in the coffin is that the 760k + 7750/7770 GDDR5 is better for the same $ without interfering with your system RAM.

 

[juanrga]

@juanrga: still waiting for the "you were right and I was wrong" post from you...

I will probably never see it...but I still know it, and so do you, even if you never admit it...

Regarding what? SOI vs. BULK? Your post is too cryptic.

 

[ohyouknow]

Well, I was considering adopting the APU for my itx/OC the crap out it build. Looks like I will have to wait until prices come down and updated bios out of the box arrive. Now if only my microcenter can stock some ITX mobos.

In the meantime. Team Vulcan 2133 RAM. Anybody know of them? Quality?

 

[palladin9479]

It does seem like the A8-7600 is a lot less absurd in pricing as its bigger brother. Its 45w TDP is wonderful and it being priced at around $119 is where an APU should be IMO.

Well the CPU can do either 45W or 65W configuration depending on your platforms. Those numbers aren't random, they are picked because they are the most common thermal thresholds offered by low profile / SFF mini-ITX solutions. Most of your low profile quiet CPU coolers have a TDP limit of 65W with the ultra low ones being 45W which is damn near fanless. I use / built a lot o fmini-ITX systems and Kaveri looks really good for them. It's for the $499.99 HP / Dell / ect.. box's at wallmart or the 45W part inside thin notebooks like what HP offers.

 

[esrever]

Just accept you were wrong juanrga and move on. Your predictions were based on 1 sentence from AMD about 30% better throughput and you took it as a 30% increase in IPC. Your predictions were way off base. You never said 10-20%. You said 20-30%.

 

[jacobian]

It seems like the iGPU on the APUs just can't reach its full potential. Slow memory is the issue perhaps?

To add insult to injury, the 3GHz Haswell Pentium G3220 is $69 on newegg right now. It's effectively a slightly slower clocked i3-4130 without hyperthreading. Add a $100 Radeon card (like 7730), or better the 7750 for $120, and I think it will have better frame rate than the $189 Kaveri A10-7850K APU.

But on the bright side, the A10-6800k dropped to $130-140 online, so same price as i3-4150, but the A10-6800K is slightly better APU. You get two extra integer cores as well as Richland iGPU, which is still better than HD 4600 by a noticeable margin.

 

[juanrga]

Esrever, I already admitted my error about the CPU. I can say you it again: I was expecting more from Steamroller.

Yes, you are right on that I took the 30% increase as IPC. Fortunately, my article was archived by a third party:

http://www.brightsideofnews.com/news/2013/11/4/what-to-expect-from-kaveri-a-detailed-predictive-analysis.aspx

During the Hot Chips 2012 conference, AMD said that the new Steamroller modules will provide a 30% gain in IPC over the initial Bulldozer modular design. Piledriver introduced about a 8% gain over Bulldozer at the same clocks. This means that Steamroller will introduce about a 20% IPC gain over Piledriver, because 1.30 = 1.08 × 1.20 rounded to two decimal digits.

And here goes my prediction, then I assumed Kaveri CPU was clocked @4GHz:

Combining all this data, I predict that the CPU of the top Kaveri APU will be about 26% faster than top Trinity APU and about 17% faster than top Richland APU.

The S|A review gives Kaveri CPU (finally clocked @3.7GHz) an 8% performance over Richland CPU. Therefore, I was wrong about the CPU performance by about 8%.

 

[juanrga]

It seems like the iGPU on the APUs just can't reach its full potential. Slow memory is the issue perhaps?

Someone said me that some reviews are using 1600MHz memory. I could not believe him, at least not before I found this

http://www.hardwarecanucks.com/forum/hardware-canucks-reviews/65031-amd-kaveri-a10-7850k-a8-7600-review-8.html

2400MHz kit underclocked at 1600MHz? Am I reading it correctly?

 

[palladin9479]

Umm juanrga that's not her percentages work.

If you state X is 30% of Y, and it ends up being 25% of Y, then you are not off by 5%. Instead you are about 17% over estimation. Your working with percentages here, it's actual/predicted (25/30).

 

[jdwii]

@juan

Your arguement about only looking at 720p resolutions kinda negates the other arguement that kaveri = high end gaming doesnt it?

As it turns out this whole time you were wrong in assuming that kaveri could = i5 2500k. Now that the reviews are out maybe you could admit something instead of complaining about how no one listens to you and putting everyone else down for not listening to you trying to convince the world that this APU direction is the way of the future because it is soo fast that AMD wont need any other products.

In the end going bulk screwed kaveri, and GF screwed AMD.

@ etoryus

You are correct. AMD doent like to see APUs benched against cpus, thats why its hard to even find it put against even a fx 4xxx cpu.

We FINALLY came into an agreement... after all the back and forth argues with you, and not only Juanrga was wrong, i was too... ill give you both (You and Gamerk) that.

@Yuka - I am on AM3 board that does not support AM3+ CPUs... that´s why i am so persistent with the 980, i got no upgrade path unless i buy a new MotherBoard either AM3+ or FM2+, and at this point i thought i would move to Kaveri if the performance was right, so i am clueless as to what Socket to upgrade to, Intel is out of the Question.

In any case I don´t think Kaver is a complete Fail, the CPU side kinda is... but the whole APU is very easy on the eye, i am thinking about recommending Kaveri for my Cousins who play games like Burnout Paradise, San Andreas, Left4Dead2 and such on a very very crappy Dual AMD CPU at 2.2 and a iGPU Radeon HD3200, i am not even sure which CPU model is it, they could easily upgrade to a new system without spending too much on the GPU and get massive amounts of Performance Increase and being able to play modern games on Medium the less.

*cough*

@juanrga: still waiting for the "you were right and I was wrong" post from you...

I will probably never see it...but I still know it, and so do you, even if you never admit it...

I'm guessing in some delusional way he feels like he was right and will probably find some benchmarks here and their that support his claim.