AMD’s Zen Summit Ridge 8-core CPUs on Par with Intel I7 5960X Extreme(techfrag.com)
techfrag.com
AMD’s Zen Summit Ridge 8-core CPUs on Par with Intel I7 5960X Extreme
http://techfrag.com/2016/05/24/amd-zen-summit-ridge-twice-fast-fx-8350-par-intel-i7-5960x-extreme/
74 comments
Appreciate the context for that comparison, but imagine that CPU for probably 50% or less the cost and with a DDR4 controller. Sounds like a valuable product to me.
Maybe somebody could give us an estimated performance per dollar based on this some previous AMD CPU releases..
Maybe somebody could give us an estimated performance per dollar based on this some previous AMD CPU releases..
The i7-5960X has a DDR4 controller. I believe it was the first mainstream CPU with DDR4.
All the LGA2011 parts in that generation support DDR4. My 5820K does.
The 5960x already is DDR4. Also, Intel's supposed to be releasing the Broadwell-E 6950X in a few weeks, which will have 10 cores. That being said, the 6950X is also supposed to be $1500. The real fun will be the bang/buck comparison against the 6900K, which is going to have 8 cores and be $1000. Regarding bang/buck, it's hard to say, but I'd probably say that AMD places their part at the $800 price point, which would definitely give it a decent bang/buck.
Another wrench in comparing things is that AMD's promising some interesting peripheral support, like 64 lanes of pcie, usb 3.1 and direct nvme support, which will further complicate direct comparisons until release date. Intel's hanging on with the x99 chipset, which is stuck with usb 2.0 and pcie 2 for the support lanes.
Fall will be interesting, to say the least.
Another wrench in comparing things is that AMD's promising some interesting peripheral support, like 64 lanes of pcie, usb 3.1 and direct nvme support, which will further complicate direct comparisons until release date. Intel's hanging on with the x99 chipset, which is stuck with usb 2.0 and pcie 2 for the support lanes.
Fall will be interesting, to say the least.
On an overclocked 2600k now, with the ageing Sandy Bridge, I'm in the market for a >4 core CPU myself.
It's a shame the AMD is nearly half a year away, I need the invoice dated June 30 at the latest for tax purposes. So if the AUD isn't too horrible I'll probably just get a 6850k, water-cooled and overclocked, with the Samsung 950 M2 drive and 32gb DDR4 ram with room for 64. For my work in Visual Studio/NCrunch/Resharper this will be a killer combo, especially for the I/O offered by 4 PCIE lanes to the SSD.
Can't wait to see Intel get some competition again though.
It's a shame the AMD is nearly half a year away, I need the invoice dated June 30 at the latest for tax purposes. So if the AUD isn't too horrible I'll probably just get a 6850k, water-cooled and overclocked, with the Samsung 950 M2 drive and 32gb DDR4 ram with room for 64. For my work in Visual Studio/NCrunch/Resharper this will be a killer combo, especially for the I/O offered by 4 PCIE lanes to the SSD.
Can't wait to see Intel get some competition again though.
> Appreciate the context for that comparison, but imagine that CPU for probably 50% or less the cost and with a DDR4 controller.
Not only that, the differences between Haswell (2014) and Broadwell (2016) are fairly modest. And nobody buys the $999 version anyway -- if they can make something that competes with the i5 at the $300 price point they're in business.
Not only that, the differences between Haswell (2014) and Broadwell (2016) are fairly modest. And nobody buys the $999 version anyway -- if they can make something that competes with the i5 at the $300 price point they're in business.
the i5 6600k costs 230
The i5-6685R is $288. Several of the mobile versions are more.
Regardless, you get the point. It matters much more that they have something competitive with the Intel processors that constitute the bulk of market volume than that they can beat the crazy expensive one that doesn't.
Regardless, you get the point. It matters much more that they have something competitive with the Intel processors that constitute the bulk of market volume than that they can beat the crazy expensive one that doesn't.
> imagine that CPU for probably 50% or less the cost
Intel is raking it in right now because AMD doesn't have an equivalent product. If AMD produces a competitive chip Intel should reduce their prices.
So it is possible AMD releases this chip and Intel reduces the price of the 5960X. End result => people still buy Intel.
Intel is raking it in right now because AMD doesn't have an equivalent product. If AMD produces a competitive chip Intel should reduce their prices.
So it is possible AMD releases this chip and Intel reduces the price of the 5960X. End result => people still buy Intel.
Sure, but Intel's chips still won't be the best "value for the money" if Zen is as good as AMD claims it is.
Intel will never go toe-to-toe with AMD on pricing, at least not in the long term (1+ years), and unless it's approaching bankruptcy, which I assume isn't the case for now.
Intel's chips are the best value for the money right now because AMD totally messed up its last microarchitecture. They bet on something (high parallelization) that didn't pan out. And they suffered greatly for it. Kind of how Sony went with the Cell processor in the PS3 (although Sony did make the best of it in the end).
But Zen completely changes that equation. Zen is basically a direct and very close competitor to Intel's current microarchitecture and they should have rather similar performance.
However, It doesn't mean AMD's chips will necessarily beat Intel's chips in "peak performance". Because for one, Intel still has a better process technology, and second, AMD still had to compete with a moving target (albeit a very slowly moving target - Skylake is only like 20-25% faster in single-thread performance compared to Sandy Bridge).
But I imagine Zen is going to be within 10% or so of Intel's equivalent chips in single-thread performance, but potentially much better at multi-threaded performance if AMD actually sells twice as many Zen cores in its similar priced chips (8-core Zen vs quad-core Core i6, quad-core Zen vs dual-core Core i5, etc). On top of having more cores, AMD may also offer lower pricing points that outweighs the 10% difference in single-thread performance, but that remains to be seen (I probably wouldn't do that if I were them, as double the cores at a similar pricing point should be more than enough value for money).
Intel will never go toe-to-toe with AMD on pricing, at least not in the long term (1+ years), and unless it's approaching bankruptcy, which I assume isn't the case for now.
Intel's chips are the best value for the money right now because AMD totally messed up its last microarchitecture. They bet on something (high parallelization) that didn't pan out. And they suffered greatly for it. Kind of how Sony went with the Cell processor in the PS3 (although Sony did make the best of it in the end).
But Zen completely changes that equation. Zen is basically a direct and very close competitor to Intel's current microarchitecture and they should have rather similar performance.
However, It doesn't mean AMD's chips will necessarily beat Intel's chips in "peak performance". Because for one, Intel still has a better process technology, and second, AMD still had to compete with a moving target (albeit a very slowly moving target - Skylake is only like 20-25% faster in single-thread performance compared to Sandy Bridge).
But I imagine Zen is going to be within 10% or so of Intel's equivalent chips in single-thread performance, but potentially much better at multi-threaded performance if AMD actually sells twice as many Zen cores in its similar priced chips (8-core Zen vs quad-core Core i6, quad-core Zen vs dual-core Core i5, etc). On top of having more cores, AMD may also offer lower pricing points that outweighs the 10% difference in single-thread performance, but that remains to be seen (I probably wouldn't do that if I were them, as double the cores at a similar pricing point should be more than enough value for money).
>Because for one, Intel still has a better process technology
One of the reasons I've had some hope for Zen, personally, is because AMD has put manufacturing and tech sharing agreements in place with Samsung ahead of the Zen launch. Intel still arguably has a process lead, but Samsung is one of the only companies in the world that has the resources to compete.
A couple of years ago, I thought Intel's process lead might be a semi-permanent advantage. Now I'm seeing some hope that may finally be coming to an end.
One of the reasons I've had some hope for Zen, personally, is because AMD has put manufacturing and tech sharing agreements in place with Samsung ahead of the Zen launch. Intel still arguably has a process lead, but Samsung is one of the only companies in the world that has the resources to compete.
A couple of years ago, I thought Intel's process lead might be a semi-permanent advantage. Now I'm seeing some hope that may finally be coming to an end.
> Intel will never go toe-to-toe with AMD on pricing, at least not in the long term (1+ years), and unless it's approaching bankruptcy, which I assume isn't the case for now.
That's why Intel still keep their own fabs and AMD don't have fab anymore.
That's why Intel still keep their own fabs and AMD don't have fab anymore.
Intel certainly did when AMD had superior parts.
Anyone know how much memory these parts will be able to address? 128GB would be interesting.
I think that (given a supported motherboard) being able to access ECC ram is a bigger thing... AMD on ASUS can do ECC at a price point much better than Xeon. Power usage is another issue though.
A big reason to consider AMD is that they tend to always just enable pretty much all their features for every chip rather than cherry pick what they enable for each market segment like Intel does.
yes it's about time, for example, that we had ECC ram capability in consumer-grade CPUs instead of having to pay double/triple for a XEON that doesn't even outperform (much).
I've got ECC (unregistered but still) in my "consumer grade" desktop[1] today, and have for a while; while many motherboards don't say whether they support ECC or not, some do and most AMD CPUs do as well.
[1]: http://pcpartpicker.com/user/IceyEC/saved/tqWv6h
[1]: http://pcpartpicker.com/user/IceyEC/saved/tqWv6h
What's the point of ECC RAM in consumer PCs?
Same reasons as ECC on a server, or for that matter, being able to do a lower cost home or small office server. The same rules apply on a desktop too though, just less risk of impact from corrupt data in memory for a consumer PC vs server. That said, most offices are using consumer grade CPUs and hardware too.
> What's the point of ECC RAM in consumer PCs?
Some people do very professional stuff with consumer PCs.
Some people do very professional stuff with consumer PCs.
Intel has the fastest processors but it doesn't have the best value for the money, in the entry level and mid level consumer market (sub-$300 range).
Concrete example: I recently built a Linux video transcoding machine to reencode H.264 BluRay movies to smaller H.264 files [1]. A $110 AMD FX-8300 is able to reencode an average BluRay movie around 47-50 fps, while you would need to spend more than twice that amount of money to match this performance with Intel (the performance of the FX-8300 at this tasks falls between a $200 Intel i5-4590 and a $250 Xeon E3-1231 v3).
[1] Using: avconv -i input.mkv -threads auto -s hd720 -c:v libx264 -c:a libmp3lame -sn -b:v 1400k -b:a 128k -ac 2 output.mkv
Concrete example: I recently built a Linux video transcoding machine to reencode H.264 BluRay movies to smaller H.264 files [1]. A $110 AMD FX-8300 is able to reencode an average BluRay movie around 47-50 fps, while you would need to spend more than twice that amount of money to match this performance with Intel (the performance of the FX-8300 at this tasks falls between a $200 Intel i5-4590 and a $250 Xeon E3-1231 v3).
[1] Using: avconv -i input.mkv -threads auto -s hd720 -c:v libx264 -c:a libmp3lame -sn -b:v 1400k -b:a 128k -ac 2 output.mkv
There's also the power cost, depending on how much you're running that... went from an 8350 to an i7-4790K early last year, the power cost difference is pretty significant.
Even if the AMD chip runs 100 watts hotter and you're encoding constantly all year that's still only $100, small compared to the chip price difference. In datacenters where you have to think about density, cooling, etc then power efficiency gets more important but not so much for consumers except in terms of fan noise.
But over two years, what would the extra $200 get you in terms of CPU options?
> But over two years, what would the extra $200 get you in terms of CPU options?
It gets you to do the math.
The $100 estimate was based on a scenario where the AMD product ran 100W hotter and ran all year long with a heavy workload.
In reality, the comparable offer from Intel runs only 10W hotter, and a normal workload, at best, only costs some cents per year.
After all those years saving money going with the best and cheaper offering from AMD, the money you save gets you more cash to upgrade your computer.
Because economic matter.
It gets you to do the math.
The $100 estimate was based on a scenario where the AMD product ran 100W hotter and ran all year long with a heavy workload.
In reality, the comparable offer from Intel runs only 10W hotter, and a normal workload, at best, only costs some cents per year.
After all those years saving money going with the best and cheaper offering from AMD, the money you save gets you more cash to upgrade your computer.
Because economic matter.
But with quicksync you can probably get a multiple of that.
Quicksync is fine for real-time communication or streaming, not for files where you want maximal visual quality with minimal bits used.
It has this bad reputation but in my experience it gives a very decent quality. I may not have stretched the compression ratio too. But at the age of 8TB consumer hard drives...
It gives a completely and quantifiably awful quality for storage, archival, and typical media center purposes. Transcoding at high efficiency is not what Quick Sync is for. It is for streaming, where it is quite fine. This is known.
I'm not repeating cargo cult science; I benchmarked it because I didn't believe it either (which is good! be suspicious!) and it would have been quite handy for a startup I was developing. But alas. Looks like dogshit or costs too many bits.
Be careful not to confuse "looks fine to me" with quantifiable.
I'm not repeating cargo cult science; I benchmarked it because I didn't believe it either (which is good! be suspicious!) and it would have been quite handy for a startup I was developing. But alas. Looks like dogshit or costs too many bits.
Be careful not to confuse "looks fine to me" with quantifiable.
I guess people have different degrees of tolerance. But I am looking at anandtech's samples:
http://www.anandtech.com/show/7007/intels-haswell-an-htpc-pe...
and I honestly can't see the difference between the QuickSync and regular H264 samples. I am sure there are some tiny subtleties that I miss and perhaps my screen is not good enough to render the difference, but if we are talking about that level of degradation, and unless you are really a purist, the 10x performance gain is really worth it.
http://www.anandtech.com/show/7007/intels-haswell-an-htpc-pe...
and I honestly can't see the difference between the QuickSync and regular H264 samples. I am sure there are some tiny subtleties that I miss and perhaps my screen is not good enough to render the difference, but if we are talking about that level of degradation, and unless you are really a purist, the 10x performance gain is really worth it.
That clip started muddy looking from all that dust. So, it's less obvious.
It's still one of the fastest CPUs available. Here's how it compares to the most modern chip I could find on cpubenchmark.net: https://www.cpubenchmark.net/compare.php?cmp%5B%5D=2332&cmp%...
That's a lower power 4-core vs a higher power 8-core. I'd bet money Intel releases a high end 8-core around the same time as AMD releases zen so they don't "win" on top end performance (around 20k on that benchmark). It will be good to see AMD pushing the higher end Intel chip prices down. I have always stuck with AMD because at the ~$100 range for general computing they have always been the best bang for the buck. Looking forward to moving an ~8k chip by this benchmark when zen starts competing.
I agree with your assessment. I'm running AMD right now and it may not be "as good" as Intel but it is not the bottleneck in my system. I can do everything that I want to do for an affordable price.
> I'd bet money Intel releases a high end 8-core around the same time as AMD releases zen so they don't "win" on top end performance (around 20k on that benchmark).
The rumour is the 6950X (the top-of-the-line Broadwell "enthusiast" chip) is going to be a 10-core part. And I suspect most people going for the more-than quad-core parts will take as many cores as they can get.
The rumour is the 6950X (the top-of-the-line Broadwell "enthusiast" chip) is going to be a 10-core part. And I suspect most people going for the more-than quad-core parts will take as many cores as they can get.
Yeah, it won't be pretty on the top end for AMD. But at least we will get to see those high end chips from Intel when AMD puts out something competitive. Intel has been treading water and raking in the cash on the high end for a decade without any decent competition from AMD.
IMHO they compete on performance at that price point by throwing more electricity at the chip.
It's still the fastest i7 CPU intel has.
http://www.cpubenchmark.net/high_end_cpus.html
http://www.cpubenchmark.net/high_end_cpus.html
Because it has the most cores. If you look at individual CPU cores you can see the 6700K is faster on a per-core basis. ("Single Thread Rating: 2329" vs. "Single Thread Rating: 1986".)
Whether a 5960X or 6700K is the fastest CPU for a given workload will depend on the workload.
Whether a 5960X or 6700K is the fastest CPU for a given workload will depend on the workload.
You can overclock 5960x and 6700k nearly to the same frequency. In my test single-thread perf is about the same (I managed to run 5960x at 4.6 and 6700k at 4.7 GHz)
This is article is pure speculation. The linked source[1] does not make such a statement.
Basically they put a graph on the right side without labels. This could be everything. E.g. inverse TDP
1: http://www.3dcenter.org/news/amd-verspricht-fuer-bristol-rid... (german)
Basically they put a graph on the right side without labels. This could be everything. E.g. inverse TDP
1: http://www.3dcenter.org/news/amd-verspricht-fuer-bristol-rid... (german)
"According to AMD"... yeah. Right. Just like the last generation.
I'll believe it when I see the benchmarks run by anandtech, arstechnica and hardocp.
It doesn't have to be as fast as the top end core i7, if they can make something faster that sells for the price of a mid range core i5.
I'll believe it when I see the benchmarks run by anandtech, arstechnica and hardocp.
It doesn't have to be as fast as the top end core i7, if they can make something faster that sells for the price of a mid range core i5.
For me, power usage and heat are pretty big factors... Not having the A/C running as much in Phoenix summers, let alone the power usage itself is pretty significant... couldn't believe the difference in moving from FX-8xxx series in my desktop, home server and htpc to newer intel chips made on my electric bill.
Breaking news: AMD makes sensational claims about future, unreleased products.
Before you get downvoted it's worth highlighting that AMD's pre-release hype and cherry-picked "leaks" are notorious for being over optimistic, to put it generously. In any case I'm glad to see they're still trying to compete in the desktop space.
It seems like just yesterday when AMD was claiming in pre-release statements a 50 percent performance advantage over Intel's Core 2 offerings. http://www.eetimes.com/document.asp?doc_id=1165926
Was that when they followed through and kicked butt at the top end? It has been quite a while. Or was that when marketing started writing checks their engineers couldn't cash? (Which has been a theme throughout the bulldozer line)
I always thought the problem with bulldozer was that marketing considered a module to be 2 cores, while someone in engineering meant for a module to compete with a single Intel core. If you go by the later definition, Bulldozer kicked Intels ass in multithreaded performance and can still compete. It makes sense because they touted a different variant of hyperthreading they claimed was better. The problem of course is that makes for a huge expensive core and only half as many per chip for marketing purposes. It also had shit for single thread performance either way. But I've always felt this difference of interpretation for the architecture was a big part of the failure.
Even if you compare a 4-module Bulldozer/Piledriver against a 4-core Intel Sandy/Ivy Bridge, AMD's die size and power consumption were much higher to deliver similar or worse performance.
They were also stuck at /least/ a process node back due to earlier anti-competitive practices that Intel didn't get sufficiently penalized for. (Those practices literally forced AMD out of the same /business/ as Intel; which is what cemented the situation they've ended up in.) A fair settlement would have broken up Intel in to isolated fabrication and pattern-design businesses.
No, that's was the Athlon 64 vs. Pentium IV era, circa '04-'05. Since Intel introduced the Core 2 line of processors in '06 Intel took a permanent and growing lead.
It was actually earlier than that. Regular Athlon vs. Pentium III. The K7 architecture is where they finally came out on top, which was released in 1999. I'm not sure how deep into the Athlon line they kept their lead, but the Pentium 4 was a miserable line which first debuted in 2000. I don't know any friends that were serious about gaming or performance computing at the time that would have chosen a P4 over an Athlon or Athlon 64 which came out in 2003.
I think maybe part of AMDs downfall was their investment in the x86 64-bit micro-architecture extensions they added in the Athlon 64 (which Intel then adopted in their chips). During that time, Intel had refocused on heat and power usage which has paid dividends in the out years. That said, I think the 64-bit extensions came at a necessary time and allowed us to get past the 4GB barrier without having to resort to weird memory paging schemes like PAE.
I think maybe part of AMDs downfall was their investment in the x86 64-bit micro-architecture extensions they added in the Athlon 64 (which Intel then adopted in their chips). During that time, Intel had refocused on heat and power usage which has paid dividends in the out years. That said, I think the 64-bit extensions came at a necessary time and allowed us to get past the 4GB barrier without having to resort to weird memory paging schemes like PAE.
AMD64 and Opteron gave AMD an decent chunk of the server market. They unfortunately lost it again when Intel stopped playing with Itanic and decided to release a non-joke x86 Xeon.
Nope, ever since Intel's Conroe architecture, AMD never managed to catch up in either raw performance or performance per watt.
Hopefully this is true. We need to have competition in the high-end of the CPU market again.
I'd settle for a competitive upper-mid range with similar power envelope. That said, I haven't felt "pain" using a computer since I switched to SSDs a number of years ago, last non-ssd was a Core 2 Duo E6600. Followed by i7-860[1], then an AMD FX-8350 and now an i7-4790K I've been very happy with.
Thinking of going to dual UHD (4K) displays from my dual 1080p next year, which will mean a video card update... May try to hold out on a CPU/system upgrade for a 5 year cycle this time though... I've tended to go with an upper-mid-range system every 2-3 years.
[1] http://frugalcoder.us/post/2010/03/07/new-computer-2010.aspx
Thinking of going to dual UHD (4K) displays from my dual 1080p next year, which will mean a video card update... May try to hold out on a CPU/system upgrade for a 5 year cycle this time though... I've tended to go with an upper-mid-range system every 2-3 years.
[1] http://frugalcoder.us/post/2010/03/07/new-computer-2010.aspx
> Thinking of going to dual UHD (4K) displays from my dual 1080p next year, which will mean a video card update...
Do you do any gaming? If so you'll definitely need to pony up for a high-end card at 4k.
I'm in the 1440p (x2) camp for now (got QNIX monitors off ebay for ~<300 a pop) and will hold out until 4k displays reach that price. Though I'd trade 4k for 1440p @ 120Hz.
Gonna hold out on GPU upgrades until the next generation of cards (after Polaris) comes out and drives prices down some more
Do you do any gaming? If so you'll definitely need to pony up for a high-end card at 4k.
I'm in the 1440p (x2) camp for now (got QNIX monitors off ebay for ~<300 a pop) and will hold out until 4k displays reach that price. Though I'd trade 4k for 1440p @ 120Hz.
Gonna hold out on GPU upgrades until the next generation of cards (after Polaris) comes out and drives prices down some more
Not much, though Overwatch looks interesting... I don't game much at all. Mostly software development, and casual browsing, I just want the extra space.
The real question :
Is Zen going to pack with AMD Platform Security Processor (PSP)?
https://libreboot.org/faq/#amdpsp
https://libreboot.org/faq/#amdpsp
I would prefer to have it as an OPTION, and being able to disable or enable this (without any back doors.) Of course you can never be sure.
The comparisons in this article seem rather hand-wavy.
"This thing is X% faster than this thing which was Y% slower than this other thing so they're about even."
Even if these comparisons are true, AMD equaling Intel 2 years after release is nowhere near the competition we need in the CPU market.
"This thing is X% faster than this thing which was Y% slower than this other thing so they're about even."
Even if these comparisons are true, AMD equaling Intel 2 years after release is nowhere near the competition we need in the CPU market.
On the other hand Intel hasnt done anything groundbreaking in the last two years either.
Not groundbreaking but they do make steady, consistent progress. Which is more than you can credit their competitor for.
> Not groundbreaking but they do make steady, consistent progress.
Not where it matters, which is the price/performance ratio.
It doesn't matter if Intel is able to put in the market their Formula 1 chips at a Formula 1 price, when most of the world deals with a Honda Civic budget.
When we factor in the fact that the Formula 1 chip isn't even the bottleneck in the system in most cases, the comparison only gets worse.
Not where it matters, which is the price/performance ratio.
It doesn't matter if Intel is able to put in the market their Formula 1 chips at a Formula 1 price, when most of the world deals with a Honda Civic budget.
When we factor in the fact that the Formula 1 chip isn't even the bottleneck in the system in most cases, the comparison only gets worse.
That's not true but if it were: why isn't AMD's marketshare so great? Good products win, AMD proved that with the Athlon. They don't have good products today.
Any competition to push Intel to roll out cheaper and better chips are welcome. :-)
Which is also why, once available, many people have to buy AMD chips in order to keep Intel in check and everybody's chances of getting ripped off lower. I'll buy AMD because they don't cut out features like ECC. Just recently Intel introduced a non-feature to disallow Xeon chips on most motherboards, while previously you could get a cheap Xeon to surpass the consumer CPU. Market segmentation is what Intel is very good at.
I sincerely hope that AMD doesn't have microarchitectural goofs that slow everything down like the cache coloring thing last time.
They would do well to revisit some old design errors, too. Grep a recent Linux kernel for "SYSRET_SS_ATTRS" for a silly one that adds quite a few cycles to many context switches for no particularly good reason.
They would do well to revisit some old design errors, too. Grep a recent Linux kernel for "SYSRET_SS_ATTRS" for a silly one that adds quite a few cycles to many context switches for no particularly good reason.
I don't see an actual benchmark, so it could be theoretical, practical, or only in multi-threaded situation. Hopefully soon we will see some actual product tests.
http://ark.intel.com/products/82930/Intel-Core-i7-5960X-Proc...