^ isn't hsa depends on using same memory pool for cpu and gpu and to accelerate task with gpu if not then will i be able to use my hd6770 to get scores like i saw few post ago (a10 vs i5)
inbetween, microsoft trolled console gamers :lol: can't play games on 1080p on x1, shame
http://wccftech.com/thief-nextgen-resolutions-reportedly-leaked-xbox-suffer-1080p-problem/
and here is a quote from that article
The ESRAM and GPU that powers Xbox One are not capable enough to provide similar video game resolutions like PlayStation 4.
and i thought that it is due to far less bandwidth of ddr3 (bottleneck)
Cazalan :
It is not an equal comparison and that claim is just one person based on browse benchmarks. It would be the same as comparing one game, say Max Payne 3, on XP on a FX 8350 with DX9 to a Intel i7 4770K on 7 using DX11. Of course the XP/AMD system would look faster as DX9 doesn't use the GPU as much as DX11 does..
So, if you like WCCFTech or not, we have some (I hate to admit this) 2nd hand bench marks for HSA software. Assuming this information has not been previously posted, and that these are 100% valid, I am impressed. First set of benchies are x86 code, second set of benchies are HSA enabled.
7850k with full on hsa whopping a 4670k intel
thats a pounding
Sure, when the software supports it...
What I find far more interesting is that when the software does not support it...the 7850k loses to the 6800k in all but one benchmark and is borderline better than the A8-3870 Llano in a few.
Not promising for HEDT.
juggernautxtr :
8350rocks :
juggernautxtr :
Sigmanick :
So, if you like WCCFTech or not, we have some (I hate to admit this) 2nd hand bench marks for HSA software. Assuming this information has not been previously posted, and that these are 100% valid, I am impressed. First set of benchies are x86 code, second set of benchies are HSA enabled.
What I find far more interesting is that when the software does not support it...the 7850k loses to the 6800k in all but one benchmark and is borderline better than the A8-3870 Llano in a few.
Not promising for HEDT.
more are coming, the benefits are astounding.and they will get stronger.as time progresses, i may not know all the programming aspects. but when a major tech company tells me it can be done, I assume they know what they are doing and try not to second guess it.
adobe,java and 3-4 other companies are already programming in this area.
even music/sound is starting to be processed by ppu/gpu.(parallel processing unit)
But again, which is more attractive to developers:
1: Use OpenCL, which can take advantage of ANY GPU (Intel iGPU, AMD APU, or NVIDIA/AMD dGPU)
OR
2: Use HSA, which only takes advantage of AMD APU's, and CUDA, which only takes advantage of NVIDIA GPU's
The fail on this argument was mentioned before. HSA and OCL are not disjoint approaches. As mentioned in the HSA FAQ: "HSA is not an alternative to OpenCL."
http://hsafoundation.com/f-a-q/
In fact, OCL is central to HSA up to the point that new version of OCL has been designed with HSA in mind. I already gave in this thread the map of OCL 2.0 key features to HSA features.
You continue repeating forever the same mistakes about HSA that you make about 'mantel', despite you are corrected continuously. Why?
HSA brings additional benefits to what API's like OpenCL offers, but doesn't add significant performance benefits over what OpenCL already offers.
Yes, you can do OpenCL + HSA, but we're right back to benefit/work+cost equation.
Hence, even though HSA leverages OpenCL, it "competes" with it as a result of the fact Devs have to decide rather to use it or not; HSA does not replace OpenCL by itself, due to not being vendor independent (Intel/NVIDIA).
^ I think HSA its more liek AMD and Nvidia GPU both can use openGL just that on AMD hardware there's extra hardware tuning and openGL will be coeded to take advantage and have better performance... so its not really vendor independent.
Graphics Chip Shipments Up In Q4 2013, Intel and Nvidia Gain Market Sharehttp://www.techpowerup.com/198029/graphics-chip-shipments-up-in-q4-2013-intel-and-nvidia-gain-market-share.html
So, if you like WCCFTech or not, we have some (I hate to admit this) 2nd hand bench marks for HSA software. Assuming this information has not been previously posted, and that these are 100% valid, I am impressed. First set of benchies are x86 code, second set of benchies are HSA enabled.
7850k with full on hsa whopping a 4670k intel
thats a pounding
Sure, when the software supports it...
What I find far more interesting is that when the software does not support it...the 7850k loses to the 6800k in all but one benchmark and is borderline better than the A8-3870 Llano in a few.
Not promising for HEDT.
juggernautxtr :
8350rocks :
juggernautxtr :
Sigmanick :
So, if you like WCCFTech or not, we have some (I hate to admit this) 2nd hand bench marks for HSA software. Assuming this information has not been previously posted, and that these are 100% valid, I am impressed. First set of benchies are x86 code, second set of benchies are HSA enabled.
What I find far more interesting is that when the software does not support it...the 7850k loses to the 6800k in all but one benchmark and is borderline better than the A8-3870 Llano in a few.
Not promising for HEDT.
more are coming, the benefits are astounding.and they will get stronger.as time progresses, i may not know all the programming aspects. but when a major tech company tells me it can be done, I assume they know what they are doing and try not to second guess it.
adobe,java and 3-4 other companies are already programming in this area.
even music/sound is starting to be processed by ppu/gpu.(parallel processing unit)
But again, which is more attractive to developers:
1: Use OpenCL, which can take advantage of ANY GPU (Intel iGPU, AMD APU, or NVIDIA/AMD dGPU)
OR
2: Use HSA, which only takes advantage of AMD APU's, and CUDA, which only takes advantage of NVIDIA GPU's
I expect we'll see the market move toward OpenCL over time, though the legacy CUDA market is going to remain a problem for AMD.
I'm near sure the on original source of those benchmarks it states all other cpus than kaveri are using a 780ti, gpu acceleration doesnt work as well over pcie, on kaveri they are on the same die and share the ram.
Edit: wrong benches but this is true of the ones me and juan posted a few pages ago.
So, if you like WCCFTech or not, we have some (I hate to admit this) 2nd hand bench marks for HSA software. Assuming this information has not been previously posted, and that these are 100% valid, I am impressed. First set of benchies are x86 code, second set of benchies are HSA enabled.
7850k with full on hsa whopping a 4670k intel
thats a pounding
Sure, when the software supports it...
What I find far more interesting is that when the software does not support it...the 7850k loses to the 6800k in all but one benchmark and is borderline better than the A8-3870 Llano in a few.
Not promising for HEDT.
juggernautxtr :
8350rocks :
juggernautxtr :
Sigmanick :
So, if you like WCCFTech or not, we have some (I hate to admit this) 2nd hand bench marks for HSA software. Assuming this information has not been previously posted, and that these are 100% valid, I am impressed. First set of benchies are x86 code, second set of benchies are HSA enabled.
What I find far more interesting is that when the software does not support it...the 7850k loses to the 6800k in all but one benchmark and is borderline better than the A8-3870 Llano in a few.
Not promising for HEDT.
more are coming, the benefits are astounding.and they will get stronger.as time progresses, i may not know all the programming aspects. but when a major tech company tells me it can be done, I assume they know what they are doing and try not to second guess it.
adobe,java and 3-4 other companies are already programming in this area.
even music/sound is starting to be processed by ppu/gpu.(parallel processing unit)
But again, which is more attractive to developers:
1: Use OpenCL, which can take advantage of ANY GPU (Intel iGPU, AMD APU, or NVIDIA/AMD dGPU)
OR
2: Use HSA, which only takes advantage of AMD APU's, and CUDA, which only takes advantage of NVIDIA GPU's
I expect we'll see the market move toward OpenCL over time, though the legacy CUDA market is going to remain a problem for AMD.
I'm near sure the on original source of those benchmarks it states all other cpus than kaveri are using a 780ti, gpu acceleration doesnt work as well over pcie, on kaveri they are on the same die and share the ram.
Edit: wrong benches but this is true of the ones me and juan posted a few pages ago.
Yes, when using OpenCL and a dGPU, you lose some performance due to having to copy everything over PCI-E. That is not an issue using OpenCL with an APU. Or an Intel iGPU.
HSA makes sense for embedded systems. It does not make sense for HEDT.
So, if you like WCCFTech or not, we have some (I hate to admit this) 2nd hand bench marks for HSA software. Assuming this information has not been previously posted, and that these are 100% valid, I am impressed. First set of benchies are x86 code, second set of benchies are HSA enabled.
7850k with full on hsa whopping a 4670k intel
thats a pounding
Sure, when the software supports it...
What I find far more interesting is that when the software does not support it...the 7850k loses to the 6800k in all but one benchmark and is borderline better than the A8-3870 Llano in a few.
Not promising for HEDT.
juggernautxtr :
8350rocks :
juggernautxtr :
Sigmanick :
So, if you like WCCFTech or not, we have some (I hate to admit this) 2nd hand bench marks for HSA software. Assuming this information has not been previously posted, and that these are 100% valid, I am impressed. First set of benchies are x86 code, second set of benchies are HSA enabled.
What I find far more interesting is that when the software does not support it...the 7850k loses to the 6800k in all but one benchmark and is borderline better than the A8-3870 Llano in a few.
Not promising for HEDT.
more are coming, the benefits are astounding.and they will get stronger.as time progresses, i may not know all the programming aspects. but when a major tech company tells me it can be done, I assume they know what they are doing and try not to second guess it.
adobe,java and 3-4 other companies are already programming in this area.
even music/sound is starting to be processed by ppu/gpu.(parallel processing unit)
But again, which is more attractive to developers:
1: Use OpenCL, which can take advantage of ANY GPU (Intel iGPU, AMD APU, or NVIDIA/AMD dGPU)
OR
2: Use HSA, which only takes advantage of AMD APU's, and CUDA, which only takes advantage of NVIDIA GPU's
I expect we'll see the market move toward OpenCL over time, though the legacy CUDA market is going to remain a problem for AMD.
I'm near sure the on original source of those benchmarks it states all other cpus than kaveri are using a 780ti, gpu acceleration doesnt work as well over pcie, on kaveri they are on the same die and share the ram.
Edit: wrong benches but this is true of the ones me and juan posted a few pages ago.
Yes, when using OpenCL and a dGPU, you lose some performance due to having to copy everything over PCI-E. That is not an issue using OpenCL with an APU. Or an Intel iGPU.
HSA makes sense for embedded systems. It does not make sense for HEDT.
Until PCI-E bandwidth achieves sufficient data transfer rates...like freedom fabric and other technologies would provide if offered in the public sector instead of only for commercial solutions. IBM has some interesting tech in this regard.
Until PCI-E bandwidth achieves sufficient data transfer rates...like freedom fabric and other technologies would provide if offered in the public sector instead of only for commercial solutions. IBM has some interesting tech in this regard.
why if you can achieve the same thing on a die,and achieve better efficiency doing it. for higher end yes the pci slot or a gpu socket may be implemented. but it's also going to effect the price of board.
Has PCI-E 3.0 effected the price of motherboards dramatically? No...did PCI-E 2.0? No...neither did the advent of HTX. Now, there are some interesting concepts for an internal system bus that could potentially scale well beyond what we have.
You might be able to achieve the same thing on a single die...but you cannot achieve efficiency and all out performance at the same time...(see: Kaveri, Ivy Bridge, Haswell, etc.).
Therefore, a combination of 2 dedicated discrete options would offer better performance all out, at the expense of efficiency. Most power users care absolutely zero what happens to the electric bill one way or the other, but they care how much they can get out of a pair of R9-290X GPUs paired with a 6 or 8 core CPU of their choice.
Kaveri = 856 GFLOPS
FX8350 + R9-290X CF = ~10 TFLOPS
Now, if you absolutely needed the raw horsepower, what makes more sense?
Has PCI-E 3.0 effected the price of motherboards dramatically? No...did PCI-E 2.0? No...neither did the advent of HTX. Now, there are some interesting concepts for an internal system bus that could potentially scale well beyond what we have.
You might be able to achieve the same thing on a single die...but you cannot achieve efficiency and all out performance at the same time...(see: Kaveri, Ivy Bridge, Haswell, etc.).
Therefore, a combination of 2 dedicated discrete options would offer better performance all out, at the expense of efficiency. Most power users care absolutely zero what happens to the electric bill one way or the other, but they care how much they can get out of a pair of R9-290X GPUs paired with a 6 or 8 core CPU of their choice.
Kaveri = 856 GFLOPS
FX8350 + R9-290X CF = ~10 TFLOPS
Now, if you absolutely needed the raw horsepower, what makes more sense?
did i not say that for high end yes.........?
yes it did affect board prices not by much but the effect was there, i have stated that dedicated parts will not and cannot go away as some systems need dedicated parts.but they will also most likely be in low production and extremely high priced.
the effect of price on implementation is up to those that have to make the parts capable of communication.
HSA brings additional benefits to what API's like OpenCL offers, but doesn't add significant performance benefits over what OpenCL already offers.
Wrong.
gamerk316 :
HSA does not replace OpenCL by itself, due to not being vendor independent (Intel/NVIDIA).
HSA is vendor independent and developed by the HSA foundation. Intel and Nvidia are developing their own proprietary alternatives to HSA. Intel response to HSA is its proprietary "neo-heterogeneous" approach. Nvidia has CUDA and openpower consortium.
So, if you like WCCFTech or not, we have some (I hate to admit this) 2nd hand bench marks for HSA software. Assuming this information has not been previously posted, and that these are 100% valid, I am impressed. First set of benchies are x86 code, second set of benchies are HSA enabled.
7850k with full on hsa whopping a 4670k intel
thats a pounding
Sure, when the software supports it...
What I find far more interesting is that when the software does not support it...the 7850k loses to the 6800k in all but one benchmark and is borderline better than the A8-3870 Llano in a few.
Not promising for HEDT.
juggernautxtr :
8350rocks :
juggernautxtr :
Sigmanick :
So, if you like WCCFTech or not, we have some (I hate to admit this) 2nd hand bench marks for HSA software. Assuming this information has not been previously posted, and that these are 100% valid, I am impressed. First set of benchies are x86 code, second set of benchies are HSA enabled.
What I find far more interesting is that when the software does not support it...the 7850k loses to the 6800k in all but one benchmark and is borderline better than the A8-3870 Llano in a few.
Not promising for HEDT.
more are coming, the benefits are astounding.and they will get stronger.as time progresses, i may not know all the programming aspects. but when a major tech company tells me it can be done, I assume they know what they are doing and try not to second guess it.
adobe,java and 3-4 other companies are already programming in this area.
even music/sound is starting to be processed by ppu/gpu.(parallel processing unit)
But again, which is more attractive to developers:
1: Use OpenCL, which can take advantage of ANY GPU (Intel iGPU, AMD APU, or NVIDIA/AMD dGPU)
OR
2: Use HSA, which only takes advantage of AMD APU's, and CUDA, which only takes advantage of NVIDIA GPU's
I expect we'll see the market move toward OpenCL over time, though the legacy CUDA market is going to remain a problem for AMD.
I'm near sure the on original source of those benchmarks it states all other cpus than kaveri are using a 780ti, gpu acceleration doesnt work as well over pcie, on kaveri they are on the same die and share the ram.
Edit: wrong benches but this is true of the ones me and juan posted a few pages ago.
Yes, when using OpenCL and a dGPU, you lose some performance due to having to copy everything over PCI-E. That is not an issue using OpenCL with an APU. Or an Intel iGPU.
HSA makes sense for embedded systems. It does not make sense for HEDT.
Until PCI-E bandwidth achieves sufficient data transfer rates...like freedom fabric and other technologies would provide if offered in the public sector instead of only for commercial solutions. IBM has some interesting tech in this regard.
8350rocks :
Has PCI-E 3.0 effected the price of motherboards dramatically? No...did PCI-E 2.0? No...neither did the advent of HTX. Now, there are some interesting concepts for an internal system bus that could potentially scale well beyond what we have.
You might be able to achieve the same thing on a single die...but you cannot achieve efficiency and all out performance at the same time...(see: Kaveri, Ivy Bridge, Haswell, etc.).
Therefore, a combination of 2 dedicated discrete options would offer better performance all out, at the expense of efficiency. Most power users care absolutely zero what happens to the electric bill one way or the other, but they care how much they can get out of a pair of R9-290X GPUs paired with a 6 or 8 core CPU of their choice.
Kaveri = 856 GFLOPS
FX8350 + R9-290X CF = ~10 TFLOPS
Now, if you absolutely needed the raw horsepower, what makes more sense?
A 300hp car is not 3x faster than a 100hp motorbike. A 4TFLOP PC is not 2x faster than a 2TFLOP console.
I already mentioned that AMD, Nvidia, and Intel will abandon discrete cards. I already explained why. Intel is the first that will abandon its own discrete cards
No hypothetical super-fast PCIe 6.0 will change this. The laws of physics are very clear at this point; the overhead is of ~10x. This implies you would need a hypothetical ~3000W dGPU to compete with a 300W APU in raw performance. This is why Nvidia, Intel, and AMD are developing APUs for exascale. AMD plans for using a 10TFLOP APU are here
your 10TFLOP (FX8350 + R9-290X CF) cannot be used for exascale supercomputers. Only the APU can be used. I already explained why and I did lots of times...
I'm pretty sure Amd needs to improve their design now
Its amazing its just as fast as my GPU and uses 1/3 the power...Also i spent like 320$ on my card i think back in the day.
The 750 is a 14% slower than the 265, because (100-116)/116 = -0.1379 ---> -13.79%
Nvidia new design has increased efficiency by about 2x compared to former Kepler cards.
This implies new cards can offer the double of performance at same power consumption
This is a first step towards their exascale plans for supercomputers. Efficiency has to be improved a lot of to achieve exascale performance. AMD has similar plans as well, therefore we will see an efficient new GCN architecture soon.
AMD’s Radeon Dual Graphics: Looking at the Desktop
Dual Graphics is still the redheaded step child of the Radeon family...
http://semiaccurate.com/2014/02/20/amds-radeon-dual-graphics-desktop/
Now before you say "look at the percentage gains", AMD is going to see bigger gains via HSA/OpenCL due to a stronger GPU. I'd like to see Intel and AMD OpenCL numbers with, say, a GTX 780 and R9 290x thrown in, so you don't have cases where the dedicated GPU is making AMD look better by comparison.
I'm still amazed you still attempt to cherry pick after how badly that JUST bit you.
HSA contains OCL, there's no denying in that. What you're ignoring is that HSA is actually MORE than just OCL programming. That's what the LibreOffice graph is telling you. You have CPU, then OCL (only) and then HSA improved software. If a program only contains the OCL part, it might nor MIGHT NOT be HSA compatible (or approved, etc) by what the standard defines as compliant.
This is a work in progress from AMD, so we all have to sit and wait until more software comes with the "HSA approved" sticker on it to draw more conclusions.
What about testing with applications used every day. That is what people will be looking for. Does it open my excel spreadsheet faster or handle a 240mb pst file. How does it handle compiling c++ or java. Is auto cad even a option ?
So, if you like WCCFTech or not, we have some (I hate to admit this) 2nd hand bench marks for HSA software. Assuming this information has not been previously posted, and that these are 100% valid, I am impressed. First set of benchies are x86 code, second set of benchies are HSA enabled.
7850k with full on hsa whopping a 4670k intel
thats a pounding
Sure, when the software supports it...
What I find far more interesting is that when the software does not support it...the 7850k loses to the 6800k in all but one benchmark and is borderline better than the A8-3870 Llano in a few.
Not promising for HEDT.
juggernautxtr :
8350rocks :
juggernautxtr :
Sigmanick :
So, if you like WCCFTech or not, we have some (I hate to admit this) 2nd hand bench marks for HSA software. Assuming this information has not been previously posted, and that these are 100% valid, I am impressed. First set of benchies are x86 code, second set of benchies are HSA enabled.
What I find far more interesting is that when the software does not support it...the 7850k loses to the 6800k in all but one benchmark and is borderline better than the A8-3870 Llano in a few.
Not promising for HEDT.
more are coming, the benefits are astounding.and they will get stronger.as time progresses, i may not know all the programming aspects. but when a major tech company tells me it can be done, I assume they know what they are doing and try not to second guess it.
adobe,java and 3-4 other companies are already programming in this area.
even music/sound is starting to be processed by ppu/gpu.(parallel processing unit)
But again, which is more attractive to developers:
1: Use OpenCL, which can take advantage of ANY GPU (Intel iGPU, AMD APU, or NVIDIA/AMD dGPU)
OR
2: Use HSA, which only takes advantage of AMD APU's, and CUDA, which only takes advantage of NVIDIA GPU's
I expect we'll see the market move toward OpenCL over time, though the legacy CUDA market is going to remain a problem for AMD.
I'm near sure the on original source of those benchmarks it states all other cpus than kaveri are using a 780ti, gpu acceleration doesnt work as well over pcie, on kaveri they are on the same die and share the ram.
Edit: wrong benches but this is true of the ones me and juan posted a few pages ago.
Yes, when using OpenCL and a dGPU, you lose some performance due to having to copy everything over PCI-E. That is not an issue using OpenCL with an APU. Or an Intel iGPU.
HSA makes sense for embedded systems. It does not make sense for HEDT.
Uh, it does though. Having to shuffle everything over the PCIe bus shouldn't make it slower than doing everything on the CPU.
Yeah, an APU with something like FX 8350 CPU with 280x GPU would be a lot faster than FX 8350 dCPU and 280x dGPU, but it's still going to be faster than FX 8350 without anything on the CPU.
You're also assuming that the only bus to connect GPU and CPU that will ever exist for HEDT market is PCIe 3.0.
PCIe 4.0 is going to offer almost 32GB/s bandwidth in a 16x slot. DDR3 2133 is around 17GB/s.
AMD’s Radeon Dual Graphics: Looking at the Desktop
Dual Graphics is still the redheaded step child of the Radeon family...
http://semiaccurate.com/2014/02/20/amds-radeon-dual-graphics-desktop/
These are the mining machines monopolizing high-end Radeons
http://techreport.com/news/26063/these-are-the-mining-machines-monopolizing-high-end-radeons
You're also assuming that the only bus to connect GPU and CPU that will ever exist for HEDT market is PCIe 3.0.
PCIe 4.0 is going to offer almost 32GB/s bandwidth in a 16x slot. DDR3 2133 is around 17GB/s.
For Intel they're talking about using QPI or dragonfly. But yeah at some point chips still have to talk to each other for any kind of exascale computing. Die can only get so big before the yields tank. Cray dragonfly interconnect is 500GB/s which Intel licensed.
The biggest leap really in the next few years is the 3D memories. HMC is 160-320GB/s for local memory bandwidth.
"As for pricing and competitive positioning, AMD will be launching the 290 at what we consider to be a very aggressive price of $399".
$399 is a map.(manufactures advertised price) $550 is MSRP (manufactures suggested retail price)
gotta be careful of those kinda numbers written in stories, most people don't realize there are 2 prices for products.
And on THG but I am too lazy to look up the review. The MSRP for the R9 290 is $400 for the stock configuration, the MSRP for the R9 290X is $550 for the stock configuration and the MSRP for the R9 280X (HD 7970) is $300.
$600 for a R9 290 is $200 more than MSRP and $150 more than what XFX, Asus or Sapphire are putting on MSRP.
juanrga :
jimmysmitty :
juanrga :
jimmysmitty :
juanrga :
jimmysmitty :
juanrga :
What is the problem with the links? AMD has just given more info and technical details about the Steamroller modules used in Kaveri, whereas Intel has given more info about Haswell.
With both, none are really news. Everything listed about Haswell has been known since last year, same with the Steamroller modules in Kaveri.
We would rather get news on a non APU version of Steamroller and possibly Intels next CPU, Broadwell/the Haswell refresh.
Except that during the 2014 IEEE talks, both AMD and Intel have given details that they didn't before. If you can link to an old article with the info in the IEEE slides, please do it.
How could AMD give news about a non-existent "non APU version of Steamroller"?
The original rumor was that Steamroller was to be the next after PD. And we all knew AMD was going to integrate its new design into APUs but never knew they were going to kill off the idea of a CPU only.
Of course things change over time and that's why rumors are rumors.
As for the fabrication, that is what I said. I said issues with yields which could be many things unless specifically stated. It could even be due to the arch itself, but more than likely is the process tech.
As for Apple/NVidia they are using ARM as the core design and adding, as I said ARM as a basis and adding on to it. Neither are fully their own designs like say Haswell or Pile Driver are for Intel and AMD.
And still in terms of complexity, Apples A7 does not yet have OoOE. Still it is ARM who is making the claim to 10nm, not Apple and it means ARMs designs are what they see on 10nm, not Apples or nVidias.
I would like to know who is fabbing the 10nm though.
Since AMD presented its server roadmap and plans at early 2013, we knew that AMD was not releasing Steamroller Opteron CPUs. I was one of the first who predicted that the server roadmap implied that AMD was killing the idea of a FX Steamroller CPU. I received lots of replies: from a hard "you have no idea, what you say is impossible" to a weak "wait for the desktop roadmap". All the roadmaps for 2014 and 2015 are now well-known and confirm what I said.
I assume that you have no link that give to my request.
I already gave you a link from Intel acknowledging that the problem is in the fabrication process. I also said you that Altera is considering to leave Intel by TMSC. The problem cannot be due to Broadwell arch.
Neither Apple A7 nor Nvidia Denver are using ARM core designs as base. Both companies are designing their own cores from scratch. Specially Nvidia, whose design is derived from their initial x86 core. I already mentioned this to you.
Apple predecessor A6 SoC (32 bit) already had OoO. Even standard Cortex cores from ARM have been OoO designs since 32 bits. I think you would also compare performance before making claims about complexity
Apple A7 SoC (a dual core @ 1.3GHz) is able to compete with "the best AMD and Intel have to offer in this space" aka quad-cores at higher frequencies: 1.46GHz for Intel (Turbo of 2.39GHz) and 1.5GHz for AMD. You need to double the number of x86 cores and maintain higher frequencies to offer the performance of ARM cores from Apple.
The ultra-high-performance APU from Nvidia that I mentioned in the past is designed for 10nm.
8350rocks :
Samsung is the only company I know of that has actually prototyped anything on 10nm, even then it wasn't a SoC...it was NAND memory, which is a completely different technology from CPU/APU/SoC of any kind. I have not heard anything from anyone about tape outs on 10nm, even Intel isn't getting what they want from 14nm.
ARM has already fabbed 14nm chips
And I already mentioned the next slide to you. Check that part about tapping out 10nm chip
That benchmark you posted doesn't show anything but BayTrail, a A4 5000 and the Apple A& all in a bunch of web browser benchmarks. How is that equaling performance? Those benchmarks show nothing really and besides CPU they also are dependent on what browser is used and since each CPU was tested on different OSes ( 8 for the AMD A4, 8.1 for the Atom and iOS for the Apple A7) meaning different browsers for each (IE10, IE11 and Safari) it is a null and void comparison.
The only way to get a true equal comparison is to compare the chips on an equal OC with the same software therefore eliminating any personal OS optimizations (iOS is optimized for Apples hardware, much like games for a 360 are optimized for that specific hardware). Without that equal playing field, you can make a weaker CPU like the A7 look as powerful as a Core i5 through software optimizations.
And when I see the 14nm chips out (Intel has fabbed them, they just haven't gotten the yields they desire) from Samsung or whoever is putting up the billions to setup the fab I will believe it. ARM is not really doing any fabbing or process manufacturing. They develop archs and sell them to others.
The benchmarks were good enough for Anand to make his claim, which I quoted. Evidently you cannot compare the chips up to the ms unit due to the different software used, but you can compare orders of magnitude. Changing the version of the browser your x86 hardware will not run 7x faster, which means that Anand quote is accurate: Apple A7 SoC (a dual core @ 1.3GHz) is able to compete with "the best AMD and Intel have to offer in this space" aka quad-cores at higher frequencies: 1.46GHz for Intel (Turbo of 2.39GHz) and 1.5GHz for AMD.
Moreover, we have recent data from AMD confirming that the new A57 core is faster than jaguar core using the same benchmarks. This is why AMD is replacing Opteron-X servers by Seattle servers. The ARM cores are faster than the x86 cores that are replacing. And we know that Apple custom core is much faster than standard Cortex A57...
ARM develops arch, cores, interconnects, and more things. And they work closely in collaboration with foundries (TMSC, GF, Samsung, IBM). The names of the foundries appear in the slides that I gave above.
ARM and Samsung already showed a working 14nm chip time ago
Again, I wont accept what Anandtech is stating mainly because it is across different OSes and different browsers. Even the Windows 8 vs Windows 8.1 alone is not acceptable since IE11 is only on Windows 7 and 8.1 (you need to upgrade to 8.1 to get 11 in 8) and IE11 itself is faster than IE10.
It is not an equal comparison and that claim is just one person based on browse benchmarks. It would be the same as comparing one game, say Max Payne 3, on XP on a FX 8350 with DX9 to a Intel i7 4770K on 7 using DX11. Of course the XP/AMD system would look faster as DX9 doesn't use the GPU as much as DX11 does.
Until they make an even platform with the same OS using the same software to truly bench, it is not easy to make that claim. You also have to consider OS overhead. iOS is a watered down OS with very little running in the background compared to Windows 8/8.1. Most iOS based devices only have 1GB of RAM while Windows 8 uses by default with drivers and no third party apps installed 700-800MB of RAM.
Do you believe that Anand doesn't know this basic stuff? His quote about the performance of the Apple A7 is correct, I already explained why he is right and why your point irrelevant (my former explanation already considered stuff such as "OS overhead").
jimmysmitty :
As for the AMD ARM vs it current offering, there are a few things I want to point out. It is 4 x86 cores that were designed to be low power vs 8 low power cores. The Opteron X2150 also has a HD 8000 integrated which is probably a bigger chunk of the TDP, so remove that or put a lower power one in and I bet it's TDP drops.
Second, there are no third party benchmarks short of what AMD is stating that I can find, and of course they will claim what they need to try to make it sell. Their 2-4x performance on what software?
According to that the performance is 7 per core vs 10 per core meaning a 42% per core performance improvement, not 2-4x.
Now granted a core increase for the jaguar based CPU would increase TDP above the ARM based CPU, there is still no third party benchmarks to actually verify claims being made yet and until there are, it is all just marketing much like AMD, AMD, Samsung or any company will do to get the excitement up.
This link was given before. 10 over 7 is 43% not 42%, because 0.8 rounds to 1.0 not to zero. The 2--4x improvement can be seen in the same link: the new Opteron is 3x faster than the former Opteron in the given standard benchmark. Nowhere AMD says that it is 4x per core. Why this confusion?
If you want eliminate the TDP of the iGPU on the X2150 then by consistency eliminate the TDP of the extra SoC components in the A1100, because the new Opteron SoC includes things not available on the former Opteron. We also know which is the TDP of the iGPU because the X1150 (CPU) has only 5W less TDP than the X2150. Thus the new ARM cores continue being much more efficient than x86 cores that are replacing.
AMD has already admitted that the new core A57 offers more performance than its own jaguar core with significant reduction in power consumption. Their claims coincide with predictions made before. The A57 performs as expected.
And of course customs cores (Apple A7, Nvidia Denver...) are much faster than the standard A57. As shown by Anand: a dual Apple cyclone core @ 1.3GHz is able to compete with "the best AMD and Intel have to offer in this space" aka quad-cores at higher frequencies: 1.46GHz for Intel (Turbo of 2.39GHz) and 1.5GHz for AMD.
Nvidia Denver cores will be faster than Apple cyclone. In fact Nvidia already announced that will release a CPU to beat Xeons and x86 Opterons in high-performance servers.
jimmysmitty :
As for Intels troubles, SemiWiki seems a bit TSMC biased. I will say that I have talked to a few people who work at Intel (I live just a few hours south of their desert FABs after all) and they are not as concerned as others about it and of course are looking towards 10nm already. Things change and unless we have insider information all we can do is speculate.
I have no problem believing that Intel will produce 14nm and as well 10nm and beyond. They have been at this game long enough they are not filled with idiots.
I have given you three different links. And one of them was from ExtremeTech, which is very Intel-like. I am not surprised by what your friends at Intel say. During 2013 Intel denied the existence of any delay... and latter Intel changed the roadmap and admitted in public the delay to Q1 2014.
The rumor of new delay published by SemiWiki has been just confirmed in last hours. Check the new roadmap: 14nm now is now delayed to Q4 2014, wtith mobile to 2015
Of course Intel will produce 14nm chips. I agree, but that was not the point. The point was that Intel continue having problems and is delaying again the 14nm process, which reduces the technology gap that once had over the rest of foundries. I expect Foundries convergence at around 10nm.
As I said, it is a bad way to compare. It gives a basic idea but since there are too many different random factors, you cannot use it as solid information.
Look at a GPU review for example. What is the only thing that is different between say a R9 290X and a GTX 780Ti? The GPU itself and the drivers. Other than that the other hardware, OS and testing software stays completely the same that way neither GPU has any advantage except itself and its drivers.
Since each CPU tested was done on different platforms (which is hard to control), different OSes (one of which was optimized especially for that CPU) and different testing software (the browser is what you use to get those results) it is in no way a fair test of a CPUs actual potential vs another.
I never said Anand didn't know this nor that they didn't know hardware but that is a flawed test and analysis of the hardware potential.
Mantle, for example, is a great comparison. It makes AMDs GPUs look great but only when using certain CPUs in certain software that supports it. Now I could cherry pick those benchmarks and results and say that AMD GPUs are better but the truth is that is a unfair comparison of overall GPU performance. And I say this knowing full well I will not buy a GTX 780 and instead a R9 290X in the future.
And I never said Intel did not have a delay, that is common knowledge. I said that unless we are inside with direct knowledge we can only go on what people are saying they claim to have as solid information.
I don't see Intel worrying much though as what does AMD have CPU wise that is vastly better than or equal to their real server class money makers?
BTW, I expect Intel to continue to have issues even beyond 14nm and so will everyone else. The difference is that I have no doubt that Intel is already looking at what they need to do and are probably ahead of the others in producing a good yield worth at 14nm and beyond.
8350rocks :
Has PCI-E 3.0 effected the price of motherboards dramatically? No...did PCI-E 2.0? No...neither did the advent of HTX. Now, there are some interesting concepts for an internal system bus that could potentially scale well beyond what we have.
You might be able to achieve the same thing on a single die...but you cannot achieve efficiency and all out performance at the same time...(see: Kaveri, Ivy Bridge, Haswell, etc.).
Therefore, a combination of 2 dedicated discrete options would offer better performance all out, at the expense of efficiency. Most power users care absolutely zero what happens to the electric bill one way or the other, but they care how much they can get out of a pair of R9-290X GPUs paired with a 6 or 8 core CPU of their choice.
Kaveri = 856 GFLOPS
FX8350 + R9-290X CF = ~10 TFLOPS
Now, if you absolutely needed the raw horsepower, what makes more sense?
The cost to motherboards only goes up when they become more complex, i.e. LGA2011 and quad channel memory that required more layers.
jdwii :
de5_Roy :
AMD’s Radeon Dual Graphics: Looking at the Desktop
Dual Graphics is still the redheaded step child of the Radeon family...
http://semiaccurate.com/2014/02/20/amds-radeon-dual-graphics-desktop/
These are the mining machines monopolizing high-end Radeons
http://techreport.com/news/26063/these-are-the-mining-machines-monopolizing-high-end-radeons
Wow PC gaming is coming back at full speed.
Mainly due to the consoles using x86 hardware. I am a bit surprised the PS4 is leading since the XB1 does utilize both DX11 and has a Windows 8 kernel, just its own GUI, meaning it would be easy to develop for the PC then port to the XB1.9
Cazalan :
You're also assuming that the only bus to connect GPU and CPU that will ever exist for HEDT market is PCIe 3.0.
PCIe 4.0 is going to offer almost 32GB/s bandwidth in a 16x slot. DDR3 2133 is around 17GB/s.
For Intel they're talking about using QPI or dragonfly. But yeah at some point chips still have to talk to each other for any kind of exascale computing. Die can only get so big before the yields tank. Cray dragonfly interconnect is 500GB/s which Intel licensed.
The biggest leap really in the next few years is the 3D memories. HMC is 160-320GB/s for local memory bandwidth.
QPI is already faster than even PCIe 4.0, 51.2GB/s maximum bandwidth. Of course faster is better but for us, we wont really ever use that much bandwidth as it stands. Maybe in the next 5 years or so.
"Mainly due to the consoles using x86 hardware. I am a bit surprised the PS4 is leading since the XB1 does utilize both DX11 and has a Windows 8 kernel, just its own GUI, meaning it would be easy to develop for the PC then port to the XB1.9"
Not me since there really is no reason for the Xbox 1 to even exist. Underpowered and overpriced and labeled as an entertainment system even though the PS4 is more than capable of being 1 also (exclude the DVR). Being a PC gamer and maybe even a Nintendo gamer (although i really grew out of them) i really don't care much about consoles but i see why the PS4 is leading. The more gamers keep finding out that the Xbox one can only do 720P when the PS4 keeps pumping 1080P for less the more Microsoft is going to lose. Titanfall(which i find to be COD with robots and worse graphics even) should sell a decent amount of Xbox ones and Halo, Gears basically FPS multiplayer games(yawn). But it’s clear that Sony is going to win this time around since none of those titles besides titanfall is coming out this year and Sony actually has more exclusives coming out this year and it’s cheaper for multiplatform games. Even looking back in history we can see that Sony typically has more exclusives a couple of games I even could not play such as Last of us. Nintendo really is the only company ahead of Sony with possibly more exclusives.