AMD CPU speculation... and expert conjecture

[juanrga]

@palladin9479: Your "professional qualified expert opinion" is a collection of thoughts which are unrelated to what I wrote... specially the first "barrier".

@lilcinw: No I am not referring to project Denver.

@blackkstar: But do you know what kind of performance I am talking about? Besides that, there are many other issues with your post. For instance, HSA is not only about raw performance it is also about power. HSA can reduce power consumption by 2x---8x depending of the workload. Doubling the number of modules in a CPU to get a 4m/8c SR dCPU increases the power consumption.

@gamerk316: Nvidia and Intel openly disagree with you. And I suspect that AMD also.

 

[griptwister]

Steamroller is dead on arrival come january 14th.

Unless they have plans for releasing more powerful products in the future.

How is it DOA if Intels Broadwell has been delayed to Q3-2014?

Idk, about broadwell, but I heard from a insider that they're on path for 14nm coming 2014 (they're starting production).

 

[de5_Roy]

I feel that this thread is barely worth commenting on... It used to exciting and factual. Now it's just drama and insults. And worst of all, it's against other fans. We're arguing over speculation here... Not even actual benchmarks... I have no idea what is going on with you guys, but I'm not taking part in it. At least Hafijur picked a team and sided with it. lol.

easy, don't reply to troll/bait posts. start ignoring them and stop feeding them. always works for me. 😀

 

[elemein]

I have a question. If AMD is gonna stop making dCPUs because the desktop market is dying, does that mean they're gonna stop making dGPUs as well?

What is the Intel equiavlent of the dGPU in the a10 7850k ?

There is none at this time.

 

[noob2222]

If AMD can find a source for SOI nodes, I guarantee their roadmaps will change.

Here's a good white paper on the challenges at 20nm.

http://www.cadence.com/rl/Resources/white_papers/custom_20nm_wp.pdf

They're estimating a 20nm node to cost 7B-10B. Who's got that kind of spare cash besides Apple?

GF has some deep pockets with their oil rich backers but they're not crazy. I've heard IBM is even considering dumping their fabs because the costs are just getting too high. The only good news is at the 14nm node the double patterning can be reused and the nodes won't be as expensive as the 28nm>20nm shift.

I was posting something like this when I was pondering why GF would cancel FD-SOI, one theory was that GF didn't feel like buying the equipment since ST-Ericsson broke up and were the ones driving it.

The forums ate my post (error 404 - thanks)

My question was why AMD went for 28nm designs. Were they going 28nm (a massive 4nm shrink from 32nm SOI that is known to work) because at the time GF was promising 28nm FD-SOI then said fk it when ST-Ericsson pulled out ...

Would AMD have stuck with 32nm SOI had they known GF would fail to produce another SOI node?

Sure, 28nm sounds like a good thing, but think about it this way. SOI offers lower heat and lower power. so does 28nm, but apparently not at the same rate. This leads to increasing ipc by x% and dropping the clock speed by y%. Is 32nm soi better than 28nm bulk?

Doesn't look like it to me considering the clock drop going from richland to kaveri, power numbers during a "cpu stress testing" will shed some light on this. Temperature results should be interesting as well.

GF needs 28nm customers, did they offer AMD some sort of deal to make the switch?

These questions we will likely never know the answer.

With GF only offering a low power 20nm node, things don't look good on performance parts. Should be another interesting development should Samsung's Exynos 6 actually be a 14nm finfet product in 2014. Intel and Samsung on 14nm, GF (possible) and TSMC on 20nm.

 

[noob2222]

What is the Intel equiavlent of the dGPU in the a10 7850k ?

First its an iGPU in the 7850k. Basically its lacking the dedicated memory.

The closest thing Intel has would be the Iris pro found in a few select laptop chips and can't be purchased as a DT option. In that sense, Intel has no DT product to compete with the GPU side.

 

[gamerk316]

@gamerk316: Nvidia and Intel openly disagree with you. And I suspect that AMD also.

Everything that's been added to the DX API going back to DX10 has been either API cleanups, or post-processing effects (programmable shaders, new AA modes, Tesselation, etc). Nothing really new in terms of new functionality. Hence the sudden interest in physics: We can't do much more graphically, so we'll push physics instead. Or new AA modes. Or tesselation. And so on.

We're not getting any groundbreaking jumps in graphical quality anymore; those days are long gone. For this generation, I expect stagnation. After this generation, I expect we move toward Ray Casting.

 

[blackkstar]

@blackkstar: But do you know what kind of performance I am talking about? Besides that, there are many other issues with your post. For instance, HSA is not only about raw performance it is also about power. HSA can reduce power consumption by 2x---8x depending of the workload. Doubling the number of modules in a CPU to get a 4m/8c SR dCPU increases the power consumption.

And there are workloads where you want performance over power consumption.

Look at what cryptocurrency is doing to AMD GPUs. You can not find them anywhere anymore and used 7950s are commonly going for $350+ on ebay.

Cryptocurrency mining is a real world example of a very successful consumer level GPGPU application and AMD is completely dominating.

Meanwhile an APU is completely uncompetitive for mining. The same will happen when more things spring up that are GPGPU based.

I think that you are speaking because you do not have experience with these sorts of workloads. If you are doing anything compute-heavy you're not going to magically have every piece of software you use flip a switch and go "we HSA NAO!"

But my JPEG decoding example was spot on why AMD"s APU focus right now is out of necessity because they can't make a big chip instead of by choice. HSA is good and it has potential but some of the ways AMD are using it and showing it off are a complete waste where a standard dCPU would be better at it anyways.

And, as I've said before, if AMD did drop dGPU and went APU only and didn't have a platform where you could have the power of a dCPU and a dGPU in a single product, everyone is just going to switch to Xeon Phi + Intel. And then you're also forgetting that to get something running on Xeon Phi it's a simple recompile while HSA needs retooling.

What you're suggesting is absolute suicide for HSA and AMD. You need to back off of the whole "Dell and HP are not shipping as many desktops, the desktop is DYIING!" train and the "ARM will take over everything because what the desktop needs right now is a slower chip that uses less power and is not compatible with the entire x86 windows software library"

 

[tracker45]

What is the Intel equiavlent of the dGPU in the a10 7850k ?

First its an iGPU in the 7850k. Basically its lacking the dedicated memory.

The closest thing Intel has would be the Iris pro found in a few select laptop chips and can't be purchased as a DT option. In that sense, Intel has no DT product to compete with the GPU side.

woops, I meant to ask the NVdia equivalent of the igpu in the 7850k

 

[designasaurus]

woops, I meant to ask the NVdia equivalent of the igpu in the 7850k

More powerful than a GT 640, but less powerful than a GTX 650.*

As has been mentioned before, the 7850k's igpu has the same number of GCN cores as the discrete Radeon 7750, but it will run slower than the Radeon 7750 due to memory bottlenecking and lower clockspeed.


*Judging by techpowerup website's charts.

 

[palladin9479]

After this generation, I expect we move toward Ray Casting.

That's prohibitively expensive but I can see us heading that way in a few generations. We'll need an order of magnitude increase in Vector processing power at the minimum before we can get real time rendering at anything resembling decent resolutions. But when it happens it's going to be beautiful.

The DT formfactor isnt' going away anytime soon, there is nothing capable of replacing it. APU's can't replace dedicated processors due to economics slamming into physics. dGPUs exist and won't suddenly die off dinosaur style come next year. If you have a dGPU then an APU is useless to you in the DT form factor. HSA might be able to salvage some performance but only in dedicated applications and not in general processing power, aka scalar processing. AMD not releasing another high performance desktop CPU, at a time when it's target market is growing, sounds more like a manufacturing problem then AMD just deciding to axe a revenue source

 

[tracker45]

Didn't they say there were going to make some normal CPUs, and call them Phenom ?

 

[Ags1]

Palladin, I have created a cut-down version of the benchmark that will run without OpenGL or OS-specific code. Download it from:

https://bitbucket.org/aclarke/headline-distribution/downloads/hb.zip

Be aware the CLI tool autosubmits to the web site as it relies on the web site to display the results. If necessary, you can copy the URL to a system with internet connectivity to submit it.

 

[etayorius]

Didn't they say there were going to make some normal CPUs, and call them Phenom ?

Really? you got a link we can check on?

 

[juanrga]

@blackkstar: There is again many issues with your post. E.g. Phi requires much more than a simple recompile, unless you want your code to run slower than in a CPU. But I will repeat my question, which you avoided: "But do you know what kind of performance I am talking about?"

@palladin9479: There are at least two companies, Intel and Nvidia, that disagree with your "APU's can't replace dedicated processors due to economics slamming into physics" mantra.

 

[juanrga]

woops, I meant to ask the NVdia equivalent of the igpu in the 7850k

More powerful than a GT 640, but less powerful than a GTX 650.*

As has been mentioned before, the 7850k's igpu has the same number of GCN cores as the discrete Radeon 7750, but it will run slower than the Radeon 7750 due to memory bottlenecking and lower clockspeed.

*Judging by techpowerup website's charts.

There are two versions of the Radeon 7750 and, graphically, Kaveri iGPU will be faster than the DDR3 version but slower than the GDDR5 version if the game fits inside the VRAM of the 7750 GDDR5.

From a compute perspective, the Kaveri iGPU could be faster because avoids the PCIe bottleneck and has hUMA.

 

[bmacsys]

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!

I think every enthusiast loves Phenom II's. I have a 1075T, a 965 BE, a unlocked X6 960T. All of them still kick ass. Unfortunately the big OEM's didn't share the same feeling.

 

[bmacsys]

Steamroller doesn't look very powerful at all.

Very disappointing.

Please, stop trolling. Haven't you read any of the posts? Looked at any of the benchmarks? AMD is making headway. they have limited resources and can't do everything at once.

 

[palladin9479]

PD already beats out Phenom II in absolute single threaded performance. For those asking for a "die shrunk" Phenom II, they already made them, it's called Llano. And before anyone screams about L3, AMD equipped Llano with double the L2 cache of the Phenom II to compensate for no L3. Thuban / Zosma both had 512KB worth of L2 with 6MB of L3 shared while Llano had 1MB of L2 with no shared L3. Llano was 32nm and the K series had unlocked multipliers for decent overclocking.

People are just using nostalgia when remembering the Phenom II's. My 970BE is still running in my backup system and it's far inferior to my FX8350.

 

[bmacsys]

At this point their is no way i could recommend anything from Amd at this point from the AM3+ side and quite honestly i'm happy i did not buy the 8350fx and instead got some new headphones.

The 63xx series is still amazing for the money, the 8350 is also pretty good provided you actually have a task that wouldn't be done with the 63xx.

The problem I still see people using is they look at absolutes and not realities. Intel's highest performing CPU is better then AMD's higher performing CPU, nobody doubts that. Money wise it's incredibly hard to not recommend the $119 USD FX6300 for lots of builds. The FX8350 itself $170 USD with the FX8320 being $160. The only Intel CPU in that range is the i3 with the occasional low power i5 dipping into the 83xx range. So many people want to do a "well the fx8 can do 8 threads and so can the i7, lets compare them", or "the fx4 is the low end so lets compare it to the i3" and end up with slanted results. When comparing CPU's of the same price range the conclusion changes dramatically.

Just stop and think, I have a $1600 USD graphics solution coupled with a $300 USD case and a $400~500 USD cooling solution. A $170 USD CPU is running all that without any issues and I foresee zero reasons to change it anytime soon.

I run a 6300 in my main rig. I don't do any benchmarking etc.... But, I have observed this..... When I use any of the three other rigs I have. a 965 BE, a 960T unlocked to X6, and a 1075T. All running in excess of 4 GHz. All with very similar hardware otherwise. The 6300 just performs all I throw at it noticeably smoother and faster. Its a shame AMD can't charge enough for their products to actually thrive.

 

[tracker45]

Didn't they say there were going to make some normal CPUs, and call them Phenom ?

Really? you got a link we can check on?

http://wccftech.com/rumor-amd-phenom-iv-x12-170-baeca-25nm-cpu-leaked-features-12-cores-6-ghz-core-clock-am4-socket-compatbility/#ixzz2bGSLHWbh

Maybe they are bringing back good old fashioned, normal CPUs. We can hope.

 

[palladin9479]

Didn't they say there were going to make some normal CPUs, and call them Phenom ?

Really? you got a link we can check on?

http://wccftech.com/rumor-amd-phenom-iv-x12-170-baeca-25nm-cpu-leaked-features-12-cores-6-ghz-core-clock-am4-socket-compatbility/#ixzz2bGSLHWbh

Maybe they are bringing back good old fashioned, normal CPUs. We can hope.

Hahahaha
That is such a troll article.

 

[bmacsys]

Didn't they say there were going to make some normal CPUs, and call them Phenom ?

Really? you got a link we can check on?

http://wccftech.com/rumor-amd-phenom-iv-x12-170-baeca-25nm-cpu-leaked-features-12-cores-6-ghz-core-clock-am4-socket-compatbility/#ixzz2bGSLHWbh

Maybe they are bringing back good old fashioned, normal CPUs. We can hope.

I live in NYC. Close by in Brooklyn there is a bridge I will sell you cheap.....

 

[etayorius]

Didn't they say there were going to make some normal CPUs, and call them Phenom ?

Really? you got a link we can check on?

http://wccftech.com/rumor-amd-phenom-iv-x12-170-baeca-25nm-cpu-leaked-features-12-cores-6-ghz-core-clock-am4-socket-compatbility/#ixzz2bGSLHWbh

Maybe they are bringing back good old fashioned, normal CPUs. We can hope.

Oh... that... we already discussed that article was a very bad rumor.

 

[griptwister]

OH MAI GAWD!!! I'm litterally laughing my butt off! I thought this whole GTX 780 Ti GHz thing was a joke from The Q6660 Inside... turns out, they're doing a GHz edition...

http://www.eteknix.com/gigabyte-gtx-780-ti-ghz-edition-specs-released/