So, when are we going to discuss the fact that Ryzen is within 5% IPC of the best Intel processors? Worse still, AMD scales workloads out to multiple cores/threads more efficiently than Intel does, so for multithreaded workloads that IPC advantage disappears entirely. Intel is literally surviving on the maturity of their 14nm process and a slight advantage in DRAM latency, with a minor architectural revision and a superior process Ryzen 3 is going to stomp Cannon/Ice Lake into the ground.
I need to upgrade from my FX6300, which ryzen is the best budget option
Matthew Martinez
The hot take from S|A is that intel knows they won't be competitive with AMD until 2022. As things stand now Zen2 could easily outpace anything intel can bring to the table with their paltry lineup of refreshes coming. A 10% average IPC uplift with any clock increases at all would give AMD the competitive edge. Once they get 7nm EUV Zen3 parts on the market its going to be a walk in the park for AMD.
Kayden Cooper
2600 or 2700
Jackson Bennett
>Ryzen 5 1600x Welp, that's it. I'm going to build with this.
Charles Fisher
>j-just wait for ryzen threeee >muh synthetic benchmark scores
Justin Parker
For less money a 2600 will reach the same clock speeds and comes with a stock cooler.
Angel Carter
As always, "laboratory" benchmarks will not reflect reallife performance during CPU-heavy tasks like gaming, rendering etc.
Those depend far more on application, and most are already designed with Intel CPUs and Nvidia GPUs in mind, as they are the most common.
Bentley Phillips
Read the entire article. The only workload where Intel manages to save face is gaymen, because that's about the only thing bingbus is actually good for.
See above
Mason Rogers
First of all cinebench is garbage. Second, 4GHz is an overclock for ryzen but a serious downclock for intel. And lastly the CCX design AMD is forcing is shit, and will ruin performance in everything memory or latency sensitive.
Ryder Bailey
L1, L2, and L3 cache latencies are already lower for Zen+ than for Coffee Lake, and DRAM latency is only slightly behind. With another microarchitecture tweak the DRAM latency discrepancy could disappear as well.
The point of this bench was to test IPC, which is why they are at the same clocks. That is also why I stated that the maturity of Intel's 14nm process, i.e. their clock speed advantage, is one of the two reasons they aren't being completely left behind by Ryzen.
As shown in the SiSoft Sandra benchmark in the article in OP, Ryzen achieves higher memory bandwidth, so the only thing holding AMD back IPC wise is the DRAM latency disadvantage.
Also, Intel's clock advantage disappears with Zen 2. I don't know if it will reach the 5GHz GloFo is targeting for 7nm but I would be extremely surprised if base clocks are lower than ~4.5GHz at the top end with a boost at or above 5GHz.
In non-time sensitive tasks. We've seen what happens in a latency intensive workload like rendering the garbage that is Fallout 4. Ironically, the poor design makes a good benchmark for draw call throughput. And Ryzen hurts in that workload, probably because of the CCX latencies being like 2-3x greater than Intel's latencies
Andrew Turner
Can't wait to upgrade to a tasty 3600X sometime in 2019.
Nolan Phillips
>going to discuss it was the fact since first gen. it was discussed to death. problem is memory controller, zen2 going to be a killer next year.
Thomas Perez
>probably because of the CCX latencies being like 2-3x greater than Intel's latencies no. study up on the architecture, there is more info now.
Dominic Thompson
>So, when are we going to discuss the fact that Ryzen is within 5% IPC of the best Intel processors? Never. See Jow ForumsIntel - they've echo-chambered themselves into believing that it's 30%-60% (depending on how hard they're circlejerking today).
Wyatt White
There are now succesfully overclocked 5.4Ghz Ryzens on all cores, aswell as 6Ghz single core achieved
More interesting take from that graph is that Intel's extra threads from HT seems to be equal to only 27% of a single core performance, whereas on AMDs implementation of SMT its significantly higher at 36%.
Gavin Foster
Ryzen IPC was always higher than Intel IPC you dumb cunt
Connor Richardson
>mfw Zen 2 on 7nm will delet intel in gaming aswell
OP graph is not representative of any of the CPUs IPC, moron.
Ayden Jackson
Hopefully AMD's CPU wins will mean they can invest money into the RTG side
Jonathan Davis
Intel have brand value, which is indefinitely more useful in the consumer space. AMD practically has none. Heck i've seen the blue shirt retards in best buy tell people AMD are "like, a chinese clone of intel"
Samuel Lee
Is Ryzen vulnerable to Spectre/Meltdown like the Intel cpus?
Hunter Edwards
First, don't group spectre and meltdown. Meltdown is a gaping hole with 0day exploits while spectre is a potential design flaw hard to exploit. Meltdown is exclusive to intel. Also there were less spectre variants applicable to amd than intel and they were easier to fix while intel still has at least 8 exploits still unannounced because they can't figure out a fix.
Ryder Mitchell
You know what's also exclusive to Intel? Foreshadow. I wonder how much of that IPC advantage will remain after they fix their shit? >design broken processors >they run faster
Logan Miller
>I wonder how much of that IPC advantage will remain after they fix their shit? I don't think they intend to. They knew about meltdown before coffelake release but still decided to mass produce it. And now they plan to release a refresh with same vulnerabilities. They announced some mitigations but I'm pretty sure there are no fixes to the architecture. Since everybody swallowed it there is no reason for them to fix anything. And since software fixes affect amd as well (even if you disable them the switches still get in the way of optimizations) and even if they fixed their shit, they would lose some performance but software mitigations are there to stay because software needs to support those decade old vulnerable cpus. So there's really no benefit for intel to properly fix it on hardware level.
Zachary Rivera
if threadripper already reaches 4.4GHz 7nm should easily go 4.5GHz+
Charles Rogers
>Implying intel doesnt have spies inside AMD >Implying they aren't gonna release a CPU that just happens to be slightly better than AND's offerings. Look up corporate espionage.
Xavier Taylor
Mommy Su said that Zen 2 will be another 10-15% improvement to IPC. In addition to being on a smaller node and achieving higher clocks than Zen+, is it safe to say that the 3700x will be THE best cpu to buy?
Connor Flores
If Intel don't pull a rabbit out of a hat then yes. I don't even care if Intel still leads the market.
Thomas Long
Why would I talk about it? The talk is over. I've got a Ryzen 5 powering my work laptop. Better iGPU and less vulnerabilities.
Andrew Garcia
>I don't even care if Intel still leads the market. They'll probably lose to AMD even in 720p low setting benchmarks. But they'll probably still churn out 800+ fps on CS:GO and that will be the only game that matters, so intel will win again.
Eli Gutierrez
But AMD already beat them in CS:GO
John Myers
>le corporate sabotage meme
If that were true the 7700k would have 8-cores
Considering Intel is on the gunsights of anti-trust feds its a good thing for them to let AMD take their place in the sun for a short while
Really? Last I saw the 720p low setting benchmark had 2700x at something like 500+ fps while the 8700k was at 800+
Oh, shit! 5ghz? Like, actual 5ghz. Not going from a momentarily 4.4 PBO to 4.8 PBO, solid 5GHz boost across the board. If the 3700x can PBO to 5.2-5.4, in addition to everything else, that's gonna be a freaking beast.
Eli Bailey
I thought the 8-cores would fix Meltdown? But yeah, the 4-cores and 6-cores are just rebrands. Literally 4 generations of the same broken shit. I guess security fixes are now a differentiating feature between the i5s and the i7s.
James Myers
Let's wait and see
Gavin Morgan
>a god tier 7nm node made using IBM tech couldn't relistically clocks way higher than a shit tier 14nm node made for smartphone cpus
Zachary Ramirez
This, always get X. It seemed like a bad buy for first gen but that was just because PBO was brand new and they were already running near to max boost clocks.
Ryan Sanders
Thats just what the foundry is aiming for in simulation. What an IC ends up being able to hit at reasonable power and leakage levels is controlled by a hundred different factors.
Gabriel Robinson
Well, they were aiming for 3ghz for original Zen, but that one behaved so well that it reached 3.8. And as the slide suggests, that 5ghz is aimed for server and data center cpus, as well as asics. So that would mean a low power envelope for that 5 ghz, right? I mean, nobody would thrust a 5ghz housefire on their data center, right?
Nolan Cruz
6 cores vs 8 cores? Flash your bios back to before spectre patches and wait for the 8 core intel CPU to make it fair.
Alexander Perez
Also 4 Ghz
Cameron Powell
Doesn't it reach 4.4 on LN2 only lol?
Landon Taylor
2950X boosts to 4.4 on air.
Blake Mitchell
the 2950X boosts 4 cores to 4.4GHz at stock, with PBO it probably could go higher and shut up about it, people got the 2990WX to 6GHz on LN2
Leo Wright
>muh cinebench >every faggot now pretends to be encoding videos 24/7
>Boost clock 4.2 Ghz Huh, which review should I watch for the 4.4?
Gavin Clark
>boost clock 4.2 >forgetting about the 200MHz XFR
Parker Peterson
GloFo's marketing packets aren't filled with information about specific AMD products. Their 14nm LPP material stated that the fmax for the process was around 3ghz~ They did have a tweaked VT for AMD and it clocked a bit higher. The only thing out of the ordinary here was that the 14nm LPP process had a particularly low fmax to begin with because of what it was designed for, being a transitional FinFET node.
GloFo also gave frequency and power figures to compare to 14nm LPP. They're touting a 40% frequency uplift, or 60% lower power at ISO frequency. Each of these statements would have to be qualified but what VT they were comparing, and where peak frequency/efficiency gains were made on the power curve. Power scaling isn't linear across the clock range.
Carson Anderson
>cinebench >encoding videos retard
Benjamin Young
The problem is IF, using too much energy, which prevents the cores to clock higher.
Parker Bailey
XFR 2 can't uplift all cores you chimp. On Ryzen 2700x, it can only make one core running at 4.3 GHz
Adrian Rodriguez
Are you fucking blind? The part where it says that the 2950X has a base clock of 3.5ghz and boosts to 4.4. Even so, PBO will boost it even further to 4.6, momentarily, on a single core, so even the 4.2 boost ones will be able to hit 4.4, on a similar way.
Justin Morgan
>Power scaling isn't linear across the clock range. yes, and and 3GHz rating on 14LPP is where it's at highest efficiency, what makes me think that 7LP is at highest eff. on 5GHz instead of it being the voltage wall like so many people imply
Bentley Allen
and on threadripper it boosts 4 cores to 4.4GHz you fucking donkey
Adrian Gonzalez
the all core boost on the 2950X isn't 4.2GHz you stupid fuck
