AMD CPU speculation... and expert conjecture

[GOM3RPLY3R]

In fact Intel is introducing Haswell ULV without PCIe support eliminating the possibility to add a dGPU.

May I see the source you read that from?

Here goes the theory

http://semiaccurate.com/2012/12/21/intel-cuts-out-gpus-technical-tidbits/

And here the reality (not PCIe 3 support for dGPU)

http://ark.intel.com/products/75459/Intel-Core-i5-4200U-Processor-3M-Cache-up-to-2_60-GHz

Well if it still runs PCIe 2 then there should be no problem considering that PCIe 3 can run on 2 without problems. Even some of the 3000 series CPUs and the Extreme series CPUs don't do that.

 

[GOM3RPLY3R]

And with that, you mean that what you're saying is not solid evidence?

Solid evidence of what ?

What i said is that we know that every measure has an error associated(edit), but we measure everything in any case, and there are ways to determine (number theory) the significance of that measured numbers, that can have enormous error margins to incredibly diminute ones (very high precision), that is, there are ways to determine the significance of those numbers... from very high precision to very low... like some of your posts (sorry).

one of the first things that i learned in first year of University, is that there aren't measures that don't include an error margin, yet is not because of that we don't measure, only we must be aware of significance.

we don't measure

don't measure

A measurement is a solid fact (a.k.a solid evidence).

In other words you are saying, "We don't have solid evidence."

Does that clear it up for you?

 

[-Fran-]

A measurement is a solid fact (a.k.a solid evidence).

In other words you are saying, "We don't have solid evidence."

Does that clear it up for you?

So you've been arguing over 2 pages over a semantics issue? Really? lol.

Anyway... If Intel gets rid of PCIe altogether, they'll still have to use *something* to connect to USB, sATA and all other stuff around. I think they'll use something proprietary to connect around stuff and charge a pretty penny for it, making enthusiast market the way it was back in the 50s: "for the 1%" expensive.

So, as it stands now and according to what Charlie came up with, it's about timing from AMD and nVidia (and VIA, of course) to put their complaint through the FTC. Hell, I think since AMD has the APU, they won't care much and probably ally with Intel on this, putting nVidia in a VERY bad/uncomfortable position and figuring out a way to either license or create their own or creating a new standard/spec (like HSA will with hUMA and stuff).

Scary thought indeed.

Cheers!

 

[montosaurous]

I know this is a little bit off topic, but is anyone else having issues with staying logged on?

 

[Cazalan]

Well if it still runs PCIe 2 then there should be no problem considering that PCIe 3 can run on 2 without problems. Even some of the 3000 series CPUs and the Extreme series CPUs don't do that.

Either way the Haswell 1 chip platform are only dual core chips. What purpose would there be really to add a dGPU?

http://www.anandtech.com/show/7047/the-haswell-ultrabook-review-core-i74500u-tested/2

I guess there is a x4 connection but that's going to limit your throughput.

 

[Cazalan]

Anyway... If Intel gets rid of PCIe altogether, they'll still have to use *something* to connect to USB, sATA and all other stuff around. I think they'll use something proprietary to connect around stuff and charge a pretty penny for it, making enthusiast market the way it was back in the 50s: "for the 1%" expensive.

So, as it stands now and according to what Charlie came up with, it's about timing from AMD and nVidia (and VIA, of course) to put their complaint through the FTC. Hell, I think since AMD has the APU, they won't care much and probably ally with Intel on this, putting nVidia in a VERY bad/uncomfortable position and figuring out a way to either license or create their own or creating a new standard/spec (like HSA will with hUMA and stuff).

Scary thought indeed.

Cheers!

They could force us to use Thunderbolt *shiver*.

The FTC can't do anything about what a SoC removes for ports. If they could Apple users would have done something a while ago for omitting a microSD slot in ALL of their products.

 

[de5_Roy]

I was looking forward to a Steamroller FX as well (6C actually not 8), but if you read between the lines of the latest roadmaps that is unlikely to happen. AMD has switched gears almost entirely to APU/SoC.
The server roadmap shows no new high end cores all through 2014. If there were a Steamroller FX there would at least be a 1P Server chip with 8 Steamroller cores. This is not listed at all.

Instead there is a Steamroller APU (4 core) for the 1P server market. This is basically the same as Kaveri but with ECC support (key for servers). And we glean that Kaveri will have 8 CU (512 sp).

i was looking forward to the 6c kaveri as well. thought that combined with onboard (on package or on mobo) gddr5, it might offer substantial performance improvement. i think, if that apu worked, amd woulda made a server/workstation sku out of it. i still think kaveri can have a top end 6c variant, may be unsuitable to servers, but fine for desktops...hopefully...

both intel and amd have been aiming to become soc makers. it was pretty clear from brazos/llano and sandy bridge launches. although i didn't expect amd to ignore their performance cpu lineup like this. still, the year hasn't ended, may be something cpu-related will come out in the upcoming months.

Note that the only reason the Bulldozer/Piledriver FX exists in the first place is because it's the same die used for Operton 16 core CPUs. Bulldozer was architected from the ground up to be a server chip.

i thought it was the 4m/8c die... which one was called orochi again?

yeah.. server... that bled amd's 20% share to single digits... that was the biggest surprise for me (with bd) even though i know very little about servers. if multi-core and multithreading was so 'future' then servers would be the ones getting the most bd cpus. the server chips seemed a lot cheaper than intel ones as well. i also assumed that server people were more aware of technology and software (contrary to dt where most things circle around games and movies and stuff), unlike public who wants to buy a box that runs windows and browses internet and facebook.

You're hilarious? Do you understand the gamer world? Who, even with an APU, in the right mind would use the integrated graphics. If your point were absolutely true, then GPUs would not exist.

And the Intel Peasants rejoice. Good Jab. But AMD will still allow a discrete GPU.

that remains to be seen. if i've read this slide right, kabini has pcie 2.0 x4 lanes for discreet gfx (total x4+x4), not much unlike the u-series intels:
http://media.bestofmicro.com/Q/Y/385882/original/kabini-system-platform.jpg
since both amd and intel are going soc route (since 2011 or before), if they manage to put 'gaming' class (1080p@30-60fps, relatively speaking) gpu on soc package or on the die itself, they both will have the ability to cut out discreet gfx from their respective mainstream lineups. and then they would jack up their high end desktop (hedt) platform prices and offer discreet gfx for those platforms only. intel has been doing that already(the overpricing of their hedt lineup). amd is behind... but amd is already cannibalizing entry level/low end discreet gfx for a couple of years since llano launch.

intel is citing pcie 3.0's power inefficiency. so if pci-sig can offer a power efficient pcie 3.0/4.0 spec, i think they'd be forced to back off. iirc intel shares some kind of business agreement with nvidia... so i don't think nvidia will sit still and let intel kick them off discreet market (unless they get absorbed into intel or intel rents their gpu i.p.). amd won't have to back off, btw, they will likely use hypertransport. basically, in the future, both intel and amd will have the ability to cut mainstream users off discreet gfx and offer their own closed platforms. mainstream laptops(not thin-ultrabooks) will be the platform most likely to lose discreet gfx capability. afaik, even though low end and ultra-thin laptop cpu still have discreet gfx support for the most parts, both intel and amd are slowly slipping socs there, albeit intel is brazenly doing it with hype. both intel amd amd will continue to offer discreet gfx capability for their respective hedt lineups (high revenue).

 

[griptwister]

More than Likely (since AMD is catering to the gamers) there will be discreet Gfx support in the GPU section. You're also forgetting AMD has more/better experience in teh GPU department. Intel has some in the server department, but that won't get them much ground as HD 4000 still sucks. So, In the near future, AMD will probably be a better option as they control the console market and soon the CPU market if intel goes ahead with plans.

 

[de5_Roy]

More than Likely (since AMD is catering to the gamers) there will be discreet Gfx support in the GPU section. You're also forgetting AMD has more/better experience in teh GPU department. Intel has some in the server department, but that won't get them much ground as HD 4000 still sucks. So, In the near future, AMD will probably be a better option as they control the console market and soon the CPU market if intel goes ahead with plans.

no i didn't forget. and i agree to some extent that amd will be in better position if/when intel and/or amd tries to snip discreet gfx support. i think, from an overall technological and technical pov, amd already has the capability but they're under pressure from intel and they still have small brand presence. amd achilles hill is software support, though.
imo this uncool move to cut off discreet supprt is being done purely from a business standpoint using technology as an excuse. i disagree about controlling the console and possibly controlling the cpu market, though. imo console will be largely decided how customers as well as sony and microsoft acts and if most people turn away from tablets. cpu- intel owns the x86 isa, unless amd can successfully popularize arm (or arm-x86 hybrid), that won't happen.
i don't see current landscape changing for the next couple years since broadwell seems to have been pushed back, silvermont is being more central and ivb-e and haswell-e coming up in the next years. after that, and when excavator comes out, i dunno what will happen then.

edit: amd doesn't cater to gamers, neither does intel. it's just p.r. it's as simple as sticking a 'gamer-oriented' tag on a high perf product.

 

[esrever]

http://www.techpowerup.com/186056/amd-super-pi-history-to-be-rewritten-courtesy-the-stilt.html

We all know AMD is pretty terrible at the Super PI benchmark and it is not likely to change. The benchmark is very old and the instruction set it requires is no longer relevant to real world workloads. It turns out however, that AMD is worse at Super PI calculations than what the architecture is actually capable of. The Stilt from Finland figured out how to significantly speed up the benchmark by going through the BIOS developers guides. The same guides are available to all the BIOS R&D teams of motherboard vendors, but seemingly no one figured out this performance issue. It is quite funny and perhaps sad at the same time that one man is able to beat an entire industry.

 

[noob2222]

And with that, you mean that what you're saying is not solid evidence?

Solid evidence of what ?

What i said is that we know that every measure has an error associated(edit), but we measure everything in any case, and there are ways to determine (number theory) the significance of that measured numbers, that can have enormous error margins to incredibly diminute ones (very high precision), that is, there are ways to determine the significance of those numbers... from very high precision to very low... like some of your posts (sorry).

one of the first things that i learned in first year of University, is that there aren't measures that don't include an error margin, yet is not because of that we don't measure, only we must be aware of significance.

we don't measure

don't measure

A measurement is a solid fact (a.k.a solid evidence).

In other words you are saying, "We don't have solid evidence."

Does that clear it up for you?

so your arguing about the margin of error?

What sample size to you want to use to determine the margin of error? 10, 100, 1,000 ... how long would it take to benchmark 1,000 times to figure the margin of error. Or is it simpler to be aware that there is a resonable "margin of error"

Maybe thats why its not measured all the time when it comes to computing. <-- It referring to the "margin of error". I think what HCL is ultimately getting at is how important is the EXACT margin of error. Answer : not very in most cases.

p.s. Woohoo, i can edit my posts now, forums have been changed some it seems.

 

[hcl123]

And with that, you mean that what you're saying is not solid evidence?

Solid evidence of what ?

What i said is that we know that every measure has an error associated(edit), but we measure everything in any case, and there are ways to determine (number theory) the significance of that measured numbers, that can have enormous error margins to incredibly diminute ones (very high precision), that is, there are ways to determine the significance of those numbers... from very high precision to very low... like some of your posts (sorry).

one of the first things that i learned in first year of University, is that there aren't measures that don't include an error margin, yet is not because of that we don't measure, only we must be aware of significance.

we don't measure

don't measure

A measurement is a solid fact (a.k.a solid evidence).

In other words you are saying, "We don't have solid evidence."

Does that clear it up for you?

You are a troubled guy... at least try to read all... and why was one of my post deleted ?

yet is not because of that we don't measure, only we must be aware of significance.

meaning we only don't measure when the the measuring instruments are missing, not because every measure has an associated error. got it ?

Benchmarks, actually see it like "ELASTIC" measuring tapes, the precision is very very very low most of cases, yet we measure any way... yet ppl rip each other eyes out because of elastic effects... only time to time comes someone like me, telling to calm down, to watch out about significance...

 

[hcl123]

And with that, you mean that what you're saying is not solid evidence?

Solid evidence of what ?

What i said is that we know that every measure has an error associated(edit), but we measure everything in any case, and there are ways to determine (number theory) the significance of that measured numbers, that can have enormous error margins to incredibly diminute ones (very high precision), that is, there are ways to determine the significance of those numbers... from very high precision to very low... like some of your posts (sorry).

You forgot to add something there, "very high precision to very low failure rate".

Since you claim that there are, give me these equations.

Yes i give you the first semester of *numeric analyses* of a university "science" course... perhaps i could give the all course if TH is willing to pay lol

Sometimes i wonder if you have a clew!.. or only hot wind...

 

[sarinaide]

Bulldozer Conditioner Fix

[/URL][/img]

Without fix

[/URL][/img]

 

[Cazalan]

More than Likely (since AMD is catering to the gamers) there will be discreet Gfx support in the GPU section. You're also forgetting AMD has more/better experience in teh GPU department. Intel has some in the server department, but that won't get them much ground as HD 4000 still sucks. So, In the near future, AMD will probably be a better option as they control the console market and soon the CPU market if intel goes ahead with plans.

Intel has passed the "some experience" mark in the server department. Their Phi kit now powers the fastest supercomputer in the world, Tianhe-2. It has 3 Phi per node (16,000 x 3 = 48,000 Phi).

Experience doesn't always equal success. Most would say NVidia has great GPU experience, yet when they put those in the phone/tablet space Apple/Qualcomm easily surpass them on graphics performance. Now on their 4th generation Tegra part it looks like Qualcomm is going to blow past them yet again with the Snapdragon 800.

AMD must be kicking themselves for selling the ATI Imageon divison to Qualcomm instead of licensing the technology. It got renamed to Adreno (anagram of Radeon) later. Qualcomm is now a 100 billion dollar company while AMD still struggles.

As a further twist now AMD will be competing with their own graphics technology as they're fighting to get into the ARM space.

 

[Cazalan]

http://www.techpowerup.com/186056/amd-super-pi-history-to-be-rewritten-courtesy-the-stilt.html

We all know AMD is pretty terrible at the Super PI benchmark and it is not likely to change. The benchmark is very old and the instruction set it requires is no longer relevant to real world workloads. It turns out however, that AMD is worse at Super PI calculations than what the architecture is actually capable of. The Stilt from Finland figured out how to significantly speed up the benchmark by going through the BIOS developers guides. The same guides are available to all the BIOS R&D teams of motherboard vendors, but seemingly no one figured out this performance issue. It is quite funny and perhaps sad at the same time that one man is able to beat an entire industry.

Some interesting stuff there. I hope to see AMD follow up on this at some point. Most people wouldn't give a rats arse about SuperPi benches but with 7bil people in the world it's bound to pique someones curiosity.

Unfortunately for us, AMD/Intel and others like to keep these guides under NDA for far too long. Which puts AMD at a distinct disadvantage because they have much fewer staff that actually works with these guides. If AMD were more transparent they could get free help from the enthusiast community more often.

 

[GOM3RPLY3R]

hcl, I am reading everything that you are saying. It seems that we have fallen into a pit of unknown and we are not on the right page. I'm done with our conversation because of this. I have not failed, do not think that, you just simply don't understand. These "classes" you took must be from the 1800s.

 

[griptwister]

hcl, I am reading everything that you are saying. It seems that we have fallen into a pit of unknown and we are not on the right page. I'm done with our conversation because of this. I have not failed, do not think that, you just simply don't understand. These "classes" you took must be from the 1800s.

Can't argue with that kind of logic. Lol.

 

[juanrga]

Performance boost will be introduced by HSA?

If you ask some of the developers enabling HSA in their applications, they are seeing five times the performance – that's 500%.

http://www.expertreviews.co.uk/processors/1299913/the-big-interview-apus-hsa-and-where-next-for-amd

I wait that to be the best possible case, but gives a good idea of the potential of the new technology.

 

[juanrga]

Performance boost will be introduced by HSA?

If you ask some of the developers enabling HSA in their applications, they are seeing five times the performance – that's 500%.

http://www.expertreviews.co.uk/processors/1299913/the-big-interview-apus-hsa-and-where-next-for-amd

I wait that to be the best possible case, but gives a good idea of the potential of the new technology.

Interesting so AMD hopes are all on HSA. I like how he says haswell is like a 10% improvement only with HSA you can improve by 5x lol. AMD are pinning there hopes on gpu acceleration as they know there cpu architecture is rubbish.

No. In the first place, AMD Piledriver CPU architecture is good enough to win the new Haswell CPU. There are benchmarks where the 'old' FX-8350 is faster than the new i7-4770k. Moreover, the new Centurion FX 9000 series will be about a 15% faster than 8000 series.

In the second place, AMD continues its plans with Steamroller and Excavator architectures. Steamroller CPUs will be about a 20% faster than Piledriver 8000, clock per clock, probably a 30% with a clock boost. If a 8000 or 9000 series can outperform a top-of-line 4770k, imagine what Steamroller series can do.

Finally, I think you miss what HSA is. HSA is a brilliant attempt to optimize the resources in your computer. CPUs are good for serial workloads, GPUs are good at parallel workloads. HSA proposes a 'abstract' programming model where the best unit (CPU or GPU) is used at each time for a given workload in function of its nature: serial or parallel. If you can boost existent application by 5x by using HSA what is wrong with that?

 

[mayankleoboy1]

Finally, I think you miss what HSA is. HSA is a brilliant attempt to optimize the resources in your computer. CPUs are good for serial workloads, GPUs are good at parallel workloads. HSA proposes a 'abstract' programming model where the best unit (CPU or GPU) is used at each time for a given workload in function of its nature: serial or parallel. If you can boost existent application by 5x by using HSA what is wrong with that?

This is already being offered by OpenCL.
Where HSA is different is that it aims to remove/diminish the difference between a CPU and GPU at the hardware level.

Also, i am much less impressed by developer claims that HSA increased some tasks perf to 5x in the PS4. It is easy to develop a specialized software if you know all the hardware internals, and are coding at the machine level. The real test of HSA is on a desktop, with general purpose software running on multitudes of HSA devices, each with a different configuration.

The only HSA we have yet seen on desktops is Richland, which is mostly a CPU+GPU on a single die. What is the timeframe for AMD to release a true HSA device for personal computers ? What is the chance that PC's will get a unified RAM+VRAM ? Will MS windows work properly in the HSA world ?

From the PC perspective, HSA is still only a vision shown by AMD.

 

[amdfangirl]

The only HSA we have yet seen on desktops is Richland, which is mostly a CPU+GPU on a single die. What is the timeframe for AMD to release a true HSA device for personal computers ? What is the chance that PC's will get a unified RAM+VRAM ? Will MS windows work properly in the HSA world ?

How long did it take for Microsoft to fully support x86_64?

How long did it take for 64bit to be the only Windows option? (oh wait)

 

[sarinaide]

Finally, I think you miss what HSA is. HSA is a brilliant attempt to optimize the resources in your computer. CPUs are good for serial workloads, GPUs are good at parallel workloads. HSA proposes a 'abstract' programming model where the best unit (CPU or GPU) is used at each time for a given workload in function of its nature: serial or parallel. If you can boost existent application by 5x by using HSA what is wrong with that?

This is already being offered by OpenCL.
Where HSA is different is that it aims to remove/diminish the difference between a CPU and GPU at the hardware level.

Also, i am much less impressed by developer claims that HSA increased some tasks perf to 5x in the PS4. It is easy to develop a specialized software if you know all the hardware internals, and are coding at the machine level. The real test of HSA is on a desktop, with general purpose software running on multitudes of HSA devices, each with a different configuration.

The only HSA we have yet seen on desktops is Richland, which is mostly a CPU+GPU on a single die. What is the timeframe for AMD to release a true HSA device for personal computers ? What is the chance that PC's will get a unified RAM+VRAM ? Will MS windows work properly in the HSA world ?

From the PC perspective, HSA is still only a vision shown by AMD.

Its been fine that developers have been limited forcefully or by necessity to archaic legacy because Intel says so, but why not optimized coding for AMD systems? When will there ever be a change for something new and with potential except the same old excuse is thrown around. Coders for EA and other developer companies have said we are stuck in a limbo and HSA provides a way of expanding beyond that, Some have said we are still in the 80's all that has changed is things look prettier yet there is the same old "you cannot do it" which is funny when you see those who can do it, it ultimately comes down to who gives a damn and has the skills.

Performance boost will be introduced by HSA?

If you ask some of the developers enabling HSA in their applications, they are seeing five times the performance – that's 500%.

http://www.expertreviews.co.uk/processors/1299913/the-big-interview-apus-hsa-and-where-next-for-amd

I wait that to be the best possible case, but gives a good idea of the potential of the new technology.

Interesting so AMD hopes are all on HSA. I like how he says haswell is like a 10% improvement only with HSA you can improve by 5x lol. AMD are pinning there hopes on gpu acceleration as they know there cpu architecture is rubbish.

No. In the first place, AMD Piledriver CPU architecture is good enough to win the new Haswell CPU. There are benchmarks where the 'old' FX-8350 is faster than the new i7-4770k. Moreover, the new Centurion FX 9000 series will be about a 15% faster than 8000 series.

In the second place, AMD continues its plans with Steamroller and Excavator architectures. Steamroller CPUs will be about a 20% faster than Piledriver 8000, clock per clock, probably a 30% with a clock boost. If a 8000 or 9000 series can outperform a top-of-line 4770k, imagine what Steamroller series can do.

Finally, I think you miss what HSA is. HSA is a brilliant attempt to optimize the resources in your computer. CPUs are good for serial workloads, GPUs are good at parallel workloads. HSA proposes a 'abstract' programming model where the best unit (CPU or GPU) is used at each time for a given workload in function of its nature: serial or parallel. If you can boost existent application by 5x by using HSA what is wrong with that?

The i7 980x is faster then the fx8350 still and that was an early 2010 cpu. Also the 4770k wins on around over 80% of the benchmarks against an fx8350. 15% nothing really, oc the 4770k to 4.2ghz and you get that 15% easily. Intel barely have any competition hence why they hold most of the market as they are that far ahead. Intel architecture is so far ahead of amd now I reckon 2016 will be the year they get to sandy bridge levels and that will take a gigantic leap from where amd are now. I think steamroller will get close to sandy bridge level though.

trolololololololol

 

[hcl123]

Performance boost will be introduced by HSA?

If you ask some of the developers enabling HSA in their applications, they are seeing five times the performance – that's 500%.

http://www.expertreviews.co.uk/processors/1299913/the-big-interview-apus-hsa-and-where-next-for-amd

I wait that to be the best possible case, but gives a good idea of the potential of the new technology.

I don't think so... i mean those are not extraordinary results, but representative in a conservative POV for the *REAL* HSA software. I happen to have stumble against a lot of patents, and the JIT compiler (Just-in-time) and the "runtime" they are building are pretty advanced... proactive loop unroll, branch elimination, extensive prefetch, HTM (hardware transactional memory), and with that(HTM), even some forms of spMT (speculative multi-threading), and this in either Latency processors (CPU - integer) or Parallel Processors (APD- accelerated processing devices -> the heterogeneous parts).

And their GPGPUs will be "processors" for all concern, they will have a MMU(memory management unit) with TLB, extensive context save states with preemption, exception handling, centralized advanced interrupt control and even adaptive caches like in CPUs. Yes AMD is building GPUs exactly as if they were advanced CPUs, meaning those GCN CU(compute units) will be prepared to fit "inside" a BD *module* without much trouble, the next FlexFPU can very well be a GCN CU of sorts... all served by stacked (3D, 2.5D) High Bandwidth Memory, with advanced interconnects.

4 modules 4 CUs, running at 4Ghz, 4x the speed of GPU card(typical 1ghz), could be equivalent or more to 16 CUs of actual GPUs, that is 16x64 =1024sp, above HD7770 is perfectly at hand

That is why i think this might not be too off base http://diybbs.zol.com.cn/11/11_106489.html

Good thread http://www.xtremesystems.org/forums/showthread.php?282723-AMD-quot-Steamroller-Excavator-quot-info-speculations-and-experience/page7

Don't know... it could be a 1th April joke... but who ever invented it, has good knowledge of the matters, all that JIT oriented software, fits well in a "dataflow" paradigma... the GPU will be equal to CPUs, that is, the FlexFPU can be of GCN CU kind... and spMT( speculative multithreading) based on HTM ( thread level data speculation) are on AMD patents...

If all well build, 500% for HSA software ( it must be build to take full advantage of all that) might even be low for "average"... but it must be HSA software...

EDIT:

This is already being offered by OpenCL.

No OpenCL is NOT HSA... OpenCL is a kronos standard for compute jobs. HSA can do C/C++, fortran, C#, Java , Python, Ada , Ruby, etc, and OpenCL. All HSA software could run in non-HSA platforms, only needs the proper runtime( which is software), but full speed only on HSA hardware( minimal hardware requirements). Likewise, OpenCL based HSA software can run in all platforms for which there is a runtime, it will not run without the runtime, similarly, all OpenCL that is starting to show now, *that is non-HSA*, will run on HSA platforms, only it will not show the full potentialities... not even close.

 

[8350rocks]

trolololololololol

He opens his mouth just often enough to change feet it seems...