Pretty sure they did once they introduced an integrated memory controller with k8. Not sure how things stacked up for k8 (IMC) vs. Conroe (onboard MC), but by the time Nehalem came out, it was over for AMD.
Has AMD ever had faster memory performance than Intel? .
Has AMD ever had faster memory performance than Intel?
Pretty sure they did once they introduced an integrated memory controller with k8. Not sure how things stacked up for k8 (IMC) vs. Conroe (onboard MC), but by the time Nehalem came out, it was over for AMD.
X300 will be about small form factors. If raven ridge updates to usb 3.1 gen2 then X300 has the perfect I/O spec for a laptop without needing a chipset. AMD has a balancing act given its going with one platform, i think they strike a pretty good balance.
I am guessing a huge integer boost with a relatively moderate floating-point improvement. Which will make the chip strong in desktops/severs, not as strong in HPC workloads.
BTW, if Zen's IPC and Perf/W are as good as Skylake/Kaby Lake, while packing more I/O on-die, where does that leave Intel's putative "process advantage?" Zen is fabbed on GF/Samsung's 14nm which is supposed to be a generation behind Intel's. TSMC's 16nm shows even better Perf/W than Samsung's 14nm, albeit at the cost of some extra die space. (~10% higher Perf/W, ~10% larger die judging by A9s produced by respective fab houses)
What about the occasionally-floated assertion that Samsung/TSMC's processes are only good for small mobile chips or low-frequency GPUs, and unable to produce high-frequency CPUs? 3.4 GHz seems enough to quash that supposition?
Can someone explain what the A/B/X300 actually does if it doesn't offer any additional I/O? Or does this mean a platform without a southbridge?
edit: nevermind I just saw the answer in the interview with AMD’s Robert Hallock . So there are 4 additional PCIe lanes, that should in theory be enough for some additional SATA/USB I/O lanes if motherboard makers stick to X300 for mini-ITX.
The X300 chipset is a tiny pinky-finger-nail-sized chip that facilitates secure boot, TPM, and other security-related features—that’s X300. X300 is connected back to the CPU with a dedicated link, freeing up four more PCIe lanes (now a total of 28) on X300-based motherboards for things like WiFi cards, GigE, and other companion chips common on the ITX form factor. I think X300 is a great answer for our fans that have asked us to facilitate more ITX solutions in the market.
Nobody expected zen to surpass intel.This makes me feel that Zen will trade blows but not beat Intel. The guy said Ryzen is the competitive processor people have been waiting for. As always we will have to wait, but "competitive" leads me to believe that have not surpassed Intel. Good enough likely for me to buy but likely not enough to get people to buy if they have a current gen Intel system.
Going to be fun to see benchmarks.
Nobody expected zen to surpass intel.
If Zens power@frequency and absolute power beats Intels with similar performance, yes, I would say that means Samsung's produced a better end performing process (after much maturing and tweaking).BTW, if Zen's IPC and Perf/W are as good as Skylake/Kaby Lake, while packing more I/O on-die, where does that leave Intel's putative "process advantage?" Zen is fabbed on GF/Samsung's 14nm which is supposed to be a generation behind Intel's. TSMC's 16nm shows even better Perf/W than Samsung's 14nm, albeit at the cost of some extra die space. (~10% higher Perf/W, ~10% larger die judging by A9s produced by respective fab houses)
What about the occasionally-floated assertion that Samsung/TSMC's processes are only good for small mobile chips or low-frequency GPUs, and unable to produce high-frequency CPUs? 3.4 GHz seems enough to quash that supposition?
Are you saying that the logic depth and the number of functional units are enough to fully qualify the performance of a micro-architecture?As soon as i saw Dresdenboy's diagram (not too far from reality), suddenly I thought that if the diagram was that and the FO4 was the same of BD, they could manage to even surpass INTEL.
10 pipelines, 8 of calculus (4 int + 4 fp/vec int)
VS
8 pipelines, 4 of calculus (only 3 can do vec int and only 2 can do FP)
Also, I doubt Zen can challenge Kaby Lake in IPC. Right now, my opinion is 10-15% behind but better SMT.
If that is true, an average frequency relaxed power Kaby Lake would be enough to beat Zen... Altho it would put some pricing pressure on Intel, it wouldn't be enough to pressurize them for better clocks/power or outcompete it.
Sent from HTC 10
(Opinions are own)
Are you saying that the logic depth and the number of functional units are enough to fully qualify the performance of a micro-architecture?
Indeed, but it wouldn't be the first time FU were too numerous and not fed fast enough. I think it's wise to wait for many independent benchmark results to see how brand prediction and data prefetchers can keep the units busyYou should see also other things, but 8 calculus pipelines, versus 4 calculus pipelines, if well fed, should at least go on par... Otherwise it's a waste of silicon...
Nobody expected zen to surpass intel.
If that is true, an average frequency relaxed power Kaby Lake would be enough to beat Zen... Altho it would put some pricing pressure on Intel, it wouldn't be enough to pressurize them for better clocks/power or outcompete it.
Sent from HTC 10
(Opinions are own)
Kabylake have the EXACT IPC than Skylake. And Skylake have less than 4% advantage in IPC versus Broadwell.
Since Zen have +2% IPC in Blender and +8% IPC in handbrake, versus BWE, we could expect probabily a tie in AVERAGE IPC between Zen and Broadwell. So Zen loses, in IPC, at most 5% versus Skylake/Kabylake. It can go on par with just 200MHz more. On broadwell-E this was achieved (+400 base and who knows how many on turbo). For the 4C let's wait for the final clocks.
Yes some did. When you look at the bench against the 6900k and saw that AMD was equal if not somewhat faster and not boosting while using less power then there was hope.
Turbo is designed to LOWER CLOCKS for better efficiency and only allow higher clocks on a very limited number of cores and only if everything else (power draw TDP temps etc)allows for it.Yes some did. When you look at the bench against the 6900k and saw that AMD was equal if not somewhat faster and not boosting while using less power then there was hope.
Turbo is designed to LOWER CLOCKS for better efficiency and only allow higher clocks on a very limited number of cores and only if everything else (power draw TDP temps etc)allows for it.
Basically the test was done with ryzen being "overclocked" (fixed clocks at highest all core turbo) and the 6900k being throttled by turbo and an undisclosed motherboard which may or may not have had an even stricter TDP setting then normal.
http://www.anandtech.com/show/10337...6900k-6850k-and-6800k-tested-up-to-10-cores/2
Turbo is designed to LOWER CLOCKS
for better efficiency
and only allow higher clocks on a very limited number of cores
and only if everything else (power draw TDP temps etc)allows for it.
Basically the test was done with ryzen being "overclocked" (fixed clocks at highest all core turbo)
and the 6900k being throttled by turbo and an undisclosed motherboard
which may or may not have had an even stricter TDP setting then normal.
and the 6900k being throttled by turbo and an undisclosed motherboard which may or may not have had an even stricter TDP setting then normal.
http://www.anandtech.com/show/10337...6900k-6850k-and-6800k-tested-up-to-10-cores/2