Thursday, June 2nd 2016

AMD Confirms Key "Summit Ridge" Specs

AMD CEO Lisa Su, speaking at the company's Computex reveal held up the most important CPU product for the company, the new eight-core "Summit Ridge" processor. A posterboy of the company's new "Zen" micro-architecture, "Summit Ridge" is an eight-core processor with SMT enabling 16 threads for the OS to deal with, a massive 40% IPC increase over the current "Excavator" architecture, and a new platform based around the AM4 socket.

The AM4 socket sees AMD completely relocate the core-logic (chipset) to the processor's die. Socket AM4 motherboards won't have any chipset on them. This also means that the processor has an integrated PCI-Express gen 3.0 root complex, besides the DDR4 integrated memory controller. With the chipset being completely integrated, connectivity such as USB and SATA will be routed out of the processor. The AM4 socket is shared with another kind of products, the "Bristol Ridge" APU, which features "Excavator" CPU cores and a 512-SP GCN 1.2 iGPU.
Add your own comment

132 Comments on AMD Confirms Key "Summit Ridge" Specs

#76
cdawall
where the hell are my stars
ParnI'm really curious how these chipset-less motherboards will be priced.
In theory they will be dirt cheap, if we use am1 as any indication. That being said these will be full at and likely require quite a bit more in the pwm department and companies love adding crap to boards which cost money. I would expect cheaper than Intel current boards myself but only time will tell.
Posted on Reply
#77
Solidstate89
TheGuruStud1st time ever? Either you're in middle school or intel tricked you from 2000-2006 lol.

It wasn't even a contest once thunderbird hit the scene. I miss that 1.4ghz bad boy. I was chewing through games. Upgraded to Athlon XP and was VERY impressed (put intel's crap 6ft under). Then, my baby, the athlon 64 3200+.... Best CPU ever made that I've had the pleasure of using.
I didn't build my first computer until I graduated college in 2010. Right when the i7 Nehalem came out.

So no, I've never once considered AMD for anything CPU related as they've never given me any reason to.
Posted on Reply
#78
G33k2Fr34k
medi01"Socket AM4 motherboards won't have any chipset on them. "

Ignoring what IPC is (you are wrong about that too), flops is VERY precise measure of what how many floating point operations a chip can do.
It alone is indeed not enough to predict performance, but when comparing 2 chips on the same architecture, it's pretty good.
Also, when having big differences such as 2 vs 8, you can be sure the former is much slower than the latter, no matter which vendor/architecture.
.
You're misusing the term architecture. Skylake and Carrizo share very similar architectures. Meaning they support the same instruction sets and have very similar memory subsystems down to TLB update mechanisms and I/O addressing. I think you meant two CPUs based on the same microarchitecture can be compared using IPC.

Here's the problem: A Zen CPU core is claimed to have 40% more IPC, whatever that means, than a current generation CPU core. This is core to core comparison, which isn't very helpful when comparing multi-core CPUs. As we know, a dual core bulldozer module is more like 50%-60% faster than a single core because of the many shared components between two cores inside a module. Even the L2 cache access latency is more than twice as slow in comparison to Deneb Phenom 2 cores.
We have no idea what "IPC" benchmark they're referencing. IPC is a benchmarking approach, not an actual benchmark. You run 1,000,000 to 100,000,000 instructionson 2 different CPU cores and whichever one completes execution first is faster. This neglects the important fact that different workloads can result in different results.

Anyhow, my speculation is that a quad core Zen will be 2x to 2.5x faster than a quad core Kaveri at similar clock speeds. BTW, I was right about the 36 CU based Polaris 10 card being faster than the 390X and close to R9 Fury (99% the performance of Fury in 3dmark 11).
Posted on Reply
#79
Casecutter
FrickAnd people today generally have more money to spend, and in consumer CPUs a 15% price difference is not a whole lot.
That in a way true, but it's true cost of the entire Eco system (platform) you choose. Everyone works from a budget and looking at a total cost it adds up and those saving provide performance in other choices Starting at a mainstream $200 CPU; i5-6500 or FX-8370 are the same price... and performance.

www.kitguru.net/components/cpu/luke-hill/150-gaming-cpu-amd-fx-8370-w-wraith-vs-intel-core-i5-6400/

Right today pricing/parts makes things a little more wonky. If you look at that review (mid-Feb '16) with a GTX 980 Ti (not a GPU for this budget CPU's but for the test I understand), AMD Wraith (included) vs. Noctua NH-D14 ($70... toss that out of the mix). Onto the Mobo's Gigabyte 990FX-Gaming (way more Mobo than you need for decent AMD OC, 970's are fine), or at the time a MSI Z170A SLI Plus that had BIOS for Non-K OC'n (but didn't Intel kill Non-K OC'n about Feb '16?)

So a FX-8370 w/Wraith, a decent 970 Mobo for well under$70, and 8Gb DDR3 1866, you can OC the AMD to like 4.3Ghz. In the end the savings here/there is like $40-$50 overall. I mean even if you can't OC the i5 the stock cooler is never going in any self respecting build so add $25. Now such savings can go into a R9 380X, not some lower end GTX960 2Gb. You'll end up with an excellent 1080p, heck a machine that offers 1440p.
FrickJust buy Intel/Nvidia, it'll just work."
I know you're not truly advocating that, but regrettably that is the prevailing mentality. Though what I want those reading this... Sure there's always deals and ways to juggle choices, the saving you acquire on some parts means flexibility to kick it up a notch in a place like the GPU where you really see rewards. That said it depends on you need it might not be gaming, it might be photo editing, or video ripping and then certain B-M tell a different story.
ParnIt looks like the AM4 chipset-less design is really going to help with the longetivity of the motherboard, so we don't have to keep changing it whenever we upgrade to a new generation of CPU.
This is a great thing as more flexibility and less costly mobo's. Rather than how traditionally Intel has most often tied their newer CPU's to an new motherboards chipset. Think the AMD/RTG Eco system shouldn't be counted out in this coming round.

Edit: I do have the bummer side of this... as not to paint only a rosy picture of AMD, and that is if you expected more than this... like a strong 1440p a "serious mainstreamer". Even if you stay at $200 on the CPU when you stay AMD it a diminishing return, and you really should bump things up a notch. First, yea the CPU you go i5 6600K (add 25%), spend on a nice ATX Z170, DDR4, and cooler to push-it. Onto GPU's this was a problem while 390's are competitive especially at 1440p the like $30 extra for the GTX970 and it lower power usage made it a better choice. While a 390X might have it beat it's was not worth the $70 (20%) more even in the best scenario. AMD for moving from mainstream to the more "serious mainstreamer" is where I saw them needing to "pull to the side". Let's hope that changes.
Posted on Reply
#80
Nihilus
UbersonicWelcome to 45 posts ago lol.
If you say so Uber, popped collar, millennial SONIC. This is the internet. I have a lifetime to point out the stupid mistakes.
Posted on Reply
#81
mastrdrver
FrickPart of me almost thinks the AMD GPUs would hurt their CPUs. Everyone is defaulting to Nvidia even when they can get more power for less money by going AMD. Countless upgrading threads all over the web and everyone just say "meh drivers something quality something something" when someone mentions AMD. I wonder if it'll be like that with Zen as well. "Just buy Intel/Nvidia, it'll just work." (btw, that is how people buy PSUs as well. "whatever just buy Corsair", even when you can get better products for less)

And people today generally have more money to spend, and in consumer CPUs a 15% price difference is not a whole lot. It can differ that much between stores. So I dunno. They have an uphill battle for sure. Moar threads is awesome, but the more I think about it the more I get a sneaky suspicion it'll not be what it should be.
I'm pretty sure Jim Keller did a good job since he was on the team that produced the Alpha 21464 when he worked at DEC. I think it's also the reason that I've seen some suggest that Zen+ (the successor to Zen) is believed to have 4 threads per core as Jim Keller also worked on that project right before he left DEC which was later sold off to Compaq.
Posted on Reply
#82
medi01
Casecutterfor the GTX970 and it lower power usage made it a better choice. While a 390X might have it beat it'
390 clearly beat 970, in some games 980.

390x traded blows with 980 (more often being faster than not)
www.techpowerup.com/reviews/Sapphire/R9_390_Nitro/23.html

I bet this kind of skewed perception is part of the problem.
Stock Fury X is faster than 980Ti at stock, starting from 1440p. (even more so in dual configs)
Posted on Reply
#83
bug
mastrdrverI'm pretty sure Jim Keller did a good job since he was on the team that produced the Alpha 21464 when he worked at DEC. I think it's also the reason that I've seen some suggest that Zen+ (the successor to Zen) is believed to have 4 threads per core as Jim Keller also worked on that project right before he left DEC which was later sold off to Compaq.
Easy boy. 4 threads per core wouldn't work (so well). The two threads per core we have right now are already sharing the same memory interface. Divide the bandwidth by 4 and you run into starvation much more often. Afaik that's the primary reason enabling HT isn't an automatic win across the board.
Posted on Reply
#84
eidairaman1
The Exiled Airman
Socs are good for phones, tablets, laptops, but when you have space estate in an atx motherboard, I'd rather see a chipset so the cpu temps can stay down, aka spread the temps out over the board, allow better overclocking
Posted on Reply
#85
TheGuruStud
bugEasy boy. 4 threads per core wouldn't work (so well). The two threads per core we have right now are already sharing the same memory interface. Divide the bandwidth by 4 and you run into starvation much more often. Afaik that's the primary reason enabling HT isn't an automatic win across the board.
That depends on the design. Don't tell IBM...
eidairaman1Socs are good for phones, tablets, laptops, but when you have space estate in an atx motherboard, I'd rather see a chipset so the cpu temps can stay down, aka spread the temps out over the board, allow better overclocking
Latency is very important and pci-e is already a high latency interface. Integration is good. It's too bad silicon is pretty much tapped out.
Posted on Reply
#86
Jism
eidairaman1Socs are good for phones, tablets, laptops, but when you have space estate in an atx motherboard, I'd rather see a chipset so the cpu temps can stay down, aka spread the temps out over the board, allow better overclocking
I hate to see the chipset moving. With any AM3+ motherboard, you had a choice to go for budget, mainstream or high-end chipset, where ofcourse the high-end chipset would give a guaranteed HTT of over 320MHz and above. The power delivery was also solid up to 220W on most high-end AM3+ motherboards and you could litterally tweak anything on your high-end board.
Posted on Reply
#87
cdawall
where the hell are my stars
JismThe power delivery was also solid up to 220W on most high-end AM3+ motherboards
Bullshit.
Posted on Reply
#88
Jism
Please check any major brand CPU support list for AM3+. Even the Crosshair IV is capable of delivery up to 220W of TDP. K thx.
Posted on Reply
#89
TheGuruStud
cdawallBullshit.
I don't have much experience on high end AM3, but my Asrock Fatality and Asus Sabertoothe R2 can put out the power. The chip itself wouldn't go any higher, but at 4.7GHz it was absolutely drinking power. Gaming alone pulled over 600W from the wall (of course the AMD GPU didn't help lol).
Posted on Reply
#90
cdawall
where the hell are my stars
TheGuruStudI don't have much experience on high end AM3, but my Asrock Fatality and Asus Sabertoothe R2 can put out the power. The chip itself wouldn't go any higher, bit at 4.7GHz it was absolutely drinking power. Gaming alone pulled over 600W from the wall (of course the AMD GPU didn't help lol).
They can short term put out the power over time they will not.
JismPlease check any major brand CPU support list for AM3+. Even the Crosshair IV is capable of delivery up to 220W of TDP. K thx.
Right and there are plenty of users with smoked boards that would argue that it will not hold up.
Posted on Reply
#91
Jism
Where's your claim boards dont tend to hold up? Even Flanker from Xtremesystems.org did rough benches with the 9570 of AMD and his equipment works fine. The power is not the problem on high-end boards, but rather properiate cooling. Most VRM's go from cooling from CPU fan but many people replace that with an AOI watercooler.
Posted on Reply
#92
cdawall
where the hell are my stars
JismWhere's your claim boards dont tend to hold up? Even Flanker from Xtremesystems.org did rough benches with the 9570 of AMD and his equipment works fine. The power is not the problem on high-end boards, but rather properiate cooling. Most VRM's go from cooling from CPU fan but many people replace that with an AOI watercooler.
Benchmarking and using a system everyday are two completely different scenarios. Look at it this way I have probably thrown away more AM3+ boards overclocking than you have seen.
Posted on Reply
#93
Jism
Lol. Benchmarking IS to fully stress a computer, overclocked, running for 5 minutes or 24 hours, does'nt matter. If VRM's dont hold up when going linx the build quality is bad. Please try a crosshair IV or Z type of board, those are the upper best for AM3+. Not your avg GAMING motherboard.

Phenom X6 at 4.2GHz 24/7 for over 1.5 year now on a crosshair IV. 330Mhz HTT, 2GHz IMC... You really should know what a X6 pulls when going full stressed.
Posted on Reply
#94
cdawall
where the hell are my stars
JismLol. Benchmarking IS to fully stress a computer, overclocked, running for 5 minutes or 24 hours, does'nt matter. If VRM's dont hold up when going linx the build quality is bad. Please try a crosshair IV or Z type of board, those are the upper best for AM3+. Not your avg GAMING motherboard.
There is a massive difference between 5 minutes and 24/7 for months on end. Electronics happen to degrade over time especially when stressed...I have had every single crosshair board since the crosshair II and I have run them on at minimum dry ice. The CHVF-z is actually worse than the CHIV and CHVF (non-z) in the VRM section. This is a known thing. Something people who do more than talk about overclocking on the internet know about...For reference the stock 9590 is well known to pop the -z's at stock settings when used for prolonged periods of time.
JismPhenom X6 at 4.2GHz 24/7 for over 1.5 year now on a crosshair IV. 330Mhz HTT, 2GHz IMC... You really should know what a X6 pulls when going full stressed.
The phenom X6 pulls A LOT less than a fx fully loaded. It typically pulls less than some at stock as a matter of fact...Please look into these chips a little more.
Posted on Reply
#95
Sharky
I think for some more percent of performance for a normal user or gamer between intel and amd is not worth to pay like 3 times the price or more . You will not notice the difference. The thing is different for the big company where any second is important
Posted on Reply
#96
cdawall
where the hell are my stars
SharkyI think for some more percent of performance for a normal user or gamer between intel and amd is not worth to pay like 3 times the price or more . You will not notice the difference. The thing is different for the big company where any second is important
3 times the price? AMD pricing fits right up against a 4790 on the top end.
Posted on Reply
#97
Jism
@cdawall

I know how degradation of electronics and chips work. However saying that Crosshair boards are 'shit' or are unable to to provide 220w of power sounds bs to me. There are alot of people hitting 4.4GHz on 8350 chips, or sometimes even higher. If those oc's would be bad in general there should be some users complaining by now about degredated motherboards or chips.

The danger with oc'ing is that on some Crosshair boards, whenever you raise the CPU core voltage, it tends to raise the CPU/NB voltage 'automaticly' as well, which is killing the CPU actually. You really need to check and set things up manually before going for high clocks on chips.

as for the crosshair boards, make sure the PINS and contact with NB is alright, and you have a steady platform to work on. See www.madshrimps.be/articles/article/1024/ASUS-Crosshair-IV-Formula-AM3-AMD-890FX-Motherboard-Review/4#axzz4Ai7uAxaM for example, tip 5.

There's alot of info to be found on certain crosshair boards, which when tweaked are just perfect.

I have this combo for over a year and i am running 24/7 with this system. There has bin no strange shit going on and i'm pretty much confident in asus products by now.
Posted on Reply
#98
cdawall
where the hell are my stars
Jism@cdawall

I know how degradation of electronics and chips work. However saying that Crosshair boards are 'shit' or are unable to to provide 220w of power sounds bs to me. There are alot of people hitting 4.4GHz on 8350 chips, or sometimes even higher. If those oc's would be bad in general there should be some users complaining by now about degredated motherboards or chips.

The danger with oc'ing is that on some Crosshair boards, whenever you raise the CPU core voltage, it tends to raise the CPU/NB voltage 'automaticly' as well, which is killing the CPU actually. You really need to check and set things up manually before going for high clocks on chips.

as for the crosshair boards, make sure the PINS and contact with NB is alright, and you have a steady platform to work on. See www.madshrimps.be/articles/article/1024/ASUS-Crosshair-IV-Formula-AM3-AMD-890FX-Motherboard-Review/4#axzz4Ai7uAxaM for example, tip 5.

There's alot of info to be found on certain crosshair boards, which when tweaked are just perfect.

I have this combo for over a year and i am running 24/7 with this system. There has bin no strange shit going on and i'm pretty much confident in asus products by now.
4.4ghz is a 220w cpu. The 9590 will run 4.5ghz under an ibt/prime95 stress test and 4.7-5.0 in everything else. Those cpus even in stock form can and do take out the formula-z's, and pretty much every other board. It's not an overnight thing on average I would say it's 6-8 months later for people heavily gaming and loading the systems.

The formula-z uses low quality mosfets and is known to have overheating issues because of this. It is far from perfect.

CPU-NB can be raised quite a bit on the fx chips without "killing them".
Posted on Reply
#99
Jism
I'd suggest to check that VRM contact / cooling at first, before throwing in such a mad CPU. You really need to provide some links with your statement.

I know that the VRM cooling on many high end boards go out from a beefy cpu cooler, that blows some air on top of the fins. When going with water, you are eliminating that.



It's not even needed, but still i apply it. VRM's do their best when running warm and not hot. It also spared your chipset from running too hot with risks of degradation.

If my cooling was better then what i have now, i know 4.4GHz or 4.5GHz is possible with this X6 1055T.
Posted on Reply
#100
cdawall
where the hell are my stars
Again the 1055t and 9590 are nothing in comparison power consumption wise.

I'm not providing links you have Google as well. Try go ogling crosshair v overheating vrms. This is a well known problem. Asus isn't gods gift to motherboards. Quite a few issues out of their latest iterations.
Posted on Reply
Add your own comment
May 11th, 2024 05:04 EDT change timezone

New Forum Posts

Popular Reviews

Controversial News Posts