Editorial AMD's Ryzen Debut: Onwards to the HEDT Market or The Stumbling Hype Train

[geon2k2]

That's amazing.

Barring highly redundant SHA hashing performance (99.99% of people run SHA hashing 0.0001% of their CPU time), Ryzen (in single threaded tasks) is a tad slower than a 6 years old Core i5 2500:

https://browser.geekbench.com/v4/cpu/compare/38140?baseline=1988008

But I'm positively sure AMD fans will weasel out of this comparison as well.

CPU Hash uses AVX:

According to AIDA website it even uses the latest intel vectorization:

"In this benchmark every thread is working on independent 8 KB data blocks, and the MMX, SSE2, SSSE3, AVX or XOP optimized calculation routines implement the latest vectorization idea of Intel"

http://aida64.helpmax.net/en/benchmark-guide/cpu-hash/
Single core performance is just fine, on par with 6900K, @155 points in cinebench 15.
A Behemoth can only be compared to another Behemoth.
I assume the parts with lower core number will do better from this perspective.

 

[birdie]

I ask how often they run SHA based workloads - they tell me SHA is accelerated using AVX. So what? How many applications out there are written in AVX assembly or accelerated using AVX instructions? 3%? 5%? Maybe more?

I give them the results of GeekBench 4 which is compiled using the same compiler using the same codebase for Linux and Windows - they tell me these are different OS'es so the results cannot be compared even though GeekBench themselves claim even different platforms should be comparable using their benchmark and those guys actually created something, unlike trolls here.

"Designed from the ground-up for cross-platform comparisons, Geekbench 4 allows you to compare system performance across devices, processor architectures, and operating systems. Geekbench 4 supports Android, iOS, macOS, Windows, and Linux."

I want to leave this planet.

AMD fanatics basically know nothing, understand nothing and give the "arguments" which in essence are unrelated blabbering and posturing.

 

[YautjaLord]

This stuff is a lil (really a lil) underwhelming. Not that i didn't expected that type of scenario, just haven't prepped myself enough for this one, for a lack of a better wording y'know? As well as for the fact SMT right now worth jack sh1t, basically it came out prematurely. Any word on better working CPU revision and/or SMT? Same happened with Dozers btw: AMD said apps/code/etc needs to mature/get updated for this arch. I can say this: i'll be able to build the rig (mobo, CPU, LCS, GPU, PSU & SSD) after June, i bet some things will be ironed out by then. i7 7700K rig will take me all the way to October til it's complete, both will share some components. This sh1t happens not only to hardware vendors, btw: Stainless Games did more or less the same with Carmageddon: Reincarnation & now it's mixed bag, atleast from all kinds of reviews standpoint. (sorry for OT) Will wait n see.

 

[BiggieShady]

Since when has the hype train had a brain?

Well, hype train may not have the brain,
but it's fun time and you also have the rhyme!

 

[Steevo]

German site reported 17% fps increase just from bios update:






https://www.golem.de/news/ryzen-7-1800x-im-test-amd-ist-endlich-zurueck-1703-125996-4.html



And that for a product that was released WHOPPING day ago.
Pathetic.

Remind me, how often did we have issues with Intel's chips.

Apparently you weren't around for things like, Windows 95 needed an update when AMD made a chip faster than 500Mhz, before Intel.

Intel and RAMBUS. BURNING HOT RAM!!! Also cost a shit ton, and performed worse.

Itanium....

Intel and cartridge problems.

Intel and Hyperthreading on boards that didn't support it out of the box on Intel branded boards (I got to send one back) these were the days before you could do a BIOS update without a CPU installed.

Intel and weak memory controllers, bump your memory voltage to the high side of spec and kill your chip, hint, the IMC terminators were the same process node as the rest of the chip and couldnt handle the voltage

Intel and their love of locking you down, 5% over clocking that may also lead to data corruption ad it increased the PCI speed and caused the IDE and SATA bus to overclock.

Intel Prescott
Intel Northwood
Intel Williamette
Really any of the Pentium 4 chips. Also didn't they bribe or coerce companies to not use AMD? Use, yes they did, as they were scared since AMD was faster, lower power, and more than competitive.

 

[Prince Valiant]

German site reported 17% fps increase just from bios update:

https://www.golem.de/news/ryzen-7-1800x-im-test-amd-ist-endlich-zurueck-1703-125996-4.html

And that for a product that was released WHOPPING day ago.
Pathetic.

Remind me, how often did we have issues with Intel's chips.

That they got a BIOS update out the door so fast and it improved performance a modest amount is a bad thing?

 

[simlariver]

I'm ok with TPU delaying the review. A lot of the "early" reviews were somewhat shallow and inconsistent. Rushed even.
There is no doubt that most reviews have been affected by unfinished EFI, memory incompatibility, SMT and HPET problems, etc... Why rush to review a buggy platform ?

The kind of review I want to see would include Memory scaling/perf, USB3.1G2 and sata transfer speeds, price/perf, etc...

 

[anubis44]

I think there are several factors to bear in mind for Ryzen's current benchmark numbers.

First, there seem to be memory speed issues with the current, early bios revisions. No support for IGMP memory profiles is a good indicator for this. Current benchmarks are comparing somewhat memory speed-hobbled Ryzens against fully optimzed, fully It's likely bios updates coming out in the next couple of weeks will fix these issues and allow the DDR4 to run at higher speeds, which will boost Ryzen's benchmark numbers in several applications.

Second, current video drivers are not at all optimized for Ryzen. Game benchmarks are essentially game driver benchmarks if you consider that most of what is going on is communication between the CPU and the GPU through the drivers. One clue to this is that Ryzen does significantly better running Doom in Vulkan than in OpenGL compared to Intel CPUs and a GTX1080. I'd bet nVidia and AMD can tune their drivers somewhat for Ryzen now that it's out.

Third, a lot of what I'm hearing is that the 7700K, a quad core, 8 thread Intel CPU, is beating an 8 core, 16 thread Ryzen in games. Well, of COURSE it is. Most games are only able to take advantage of 4 threads, and at that point, clock speed becomes the determining factor (along with driver tuning) in performance. I'd wait to see how high AMD's 4 and 6 core Ryzen's can clock before writing off Ryzen's gaming prowess. I'd be willing to bet that a Ryzen 5 4-core, 8-thread CPU will have more overclocking headroom than the 8-core, 16-thread Ryzen.

 

[medi01]

Man, this escalated quickly.
I recall hoping for Broadwell level IPC..

 

[GoldenX]

I recall people hoping for Haswell level IPC, we got something equal or better than Broadwell-E in everything except gaming, and people complain.

 

[TheLaughingMan]

I recall people hoping for Haswell level IPC, we got something equal or better than Broadwell-E in everything except gaming, and people complain.

If it beat Intel in everything across the board hands down by a margin of no less than 20% at 1 GHz, people would complain it didn't come with free girl scout cookies.

 

[cadaveca]

I think there are several factors to bear in mind for Ryzen's current benchmark numbers.

First, there seem to be memory speed issues with the current, early bios revisions. No support for IGMP memory profiles is a good indicator for this. Current benchmarks are comparing somewhat memory speed-hobbled Ryzens against fully optimzed, fully It's likely bios updates coming out in the next couple of weeks will fix these issues and allow the DDR4 to run at higher speeds, which will boost Ryzen's benchmark numbers in several applications.

Not that I saw, since default for this platform is 2133 or 2400 MHz. More memory speed only helps so much, especially since memory OC may disable Turbo.

Second, current video drivers are not at all optimized for Ryzen. Game benchmarks are essentially game driver benchmarks if you consider that most of what is going on is communication between the CPU and the GPU through the drivers. One clue to this is that Ryzen does significantly better running Doom in Vulkan than in OpenGL compared to Intel CPUs and a GTX1080. I'd bet nVidia and AMD can tune their drivers somewhat for Ryzen now that it's out.

Should only be a minor difference, if any.

Third, a lot of what I'm hearing is that the 7700K, a quad core, 8 thread Intel CPU, is beating an 8 core, 16 thread Ryzen in games. Well, of COURSE it is. Most games are only able to take advantage of 4 threads, and at that point, clock speed becomes the determining factor (along with driver tuning) in performance. I'd wait to see how high AMD's 4 and 6 core Ryzen's can clock before writing off Ryzen's gaming prowess. I'd be willing to bet that a Ryzen 5 4-core, 8-thread CPU will have more overclocking headroom than the 8-core, 16-thread Ryzen.

Yeah, AMD lacks clockspeed, but it isn't all bad. What is bad is the current issues that anyone that buys now had to deal with. For many, this may mean that they bought stuff, and it doesn't work together.

I'm saying this to make sure that people's expectations are not falsely increased.

 

[geon2k2]

I ask how often they run SHA based workloads - they tell me SHA is accelerated using AVX. So what? How many applications out there are written in AVX assembly or accelerated using AVX instructions? 3%? 5%? Maybe more?

I give them the results of GeekBench 4 which is compiled using the same compiler using the same codebase for Linux and Windows - they tell me these are different OS'es so the results cannot be compared even though GeekBench themselves claim even different platforms should be comparable using their benchmark and those guys actually created something, unlike trolls here.

"Designed from the ground-up for cross-platform comparisons, Geekbench 4 allows you to compare system performance across devices, processor architectures, and operating systems. Geekbench 4 supports Android, iOS, macOS, Windows, and Linux."

I want to leave this planet.

AMD fanatics basically know nothing, understand nothing and give the "arguments" which in essence are unrelated blabbering and posturing.

My dear, you mentioned that AVX on RyZen sucks, I showed you that it is really quite good, now you ask me why I speak about AVX? You forgot what you posted above?

You told us that power consumption sucks, I showed you it doesn't.

Now you bring this stupidity with core i5 2500 being on par with Ryzen on geekbench 4.

They are not even close.

i5 2500 is about 3500 single core, 9000 multi core, probably the ones with 4000 are heavy overclocks, as the geekbench doesn't detect very well the cpu speed, but look at the averages. I also see many with 3000, which probably means that is the default value with no overclock.
https://browser.primatelabs.com/v4/cpu/search?q=i5+2500

R7 is about 4300 single, 20000 multi core, numbers seem much more consistent.
https://browser.primatelabs.com/v4/cpu/search?utf8=✓&q=1800x

See the difference? Its HUGE.

i7 7700 has better single thread performance at about 4800, and this is the reason for which in lightly threaded loads, games, it beats Ryzen, but 2500 is pretty much dusted by Ryzen, from all perspectives.

I'm not saying by far that Ryzen is perfect, but the problems it has are not the ones pointed by you and its definitely not a train wreck, its a decent part.

 

[Caring1]

I suppose he's going to call you an AMD fanatic again for not kneeling down and espousing the glory of Intel

 

[geon2k2]

I suppose he's going to call you an AMD fanatic again for not kneeling down and espousing the glory of Intel

I'm posting from an i5 Haswell btw, which I actually don't plan to replace very soon, so I'm not that fanatic.
I just like competition, and it would be great to have again 2 vendors in the x86 space.

 

[YautjaLord]

OC3D Tv guy says the platform got rushed, but otherwise it's solid & really unprecedented: think of it, this thing can't go past either 3.9 to 4.1GHz according to the samples YT or tech site reviewers got, yet it performs in titles like GTA V and/or DOOM consistently, on par in DOOM in 1080p with all the best from blue camp, (200fps), it costs less than half of 6900K & 3 times less than 6950X & best part - these 2 (GTA V & DOOM) utilize no more than what 2 or 4 cores/threads?, use the rest of cores/threads to stream the gameplay live on Twitch TV without lag/stutter/etc.... Drop in Blender & heavy editing/content creating/modding into the mix for all i care. For 499$? A no-brainer chip/SoC.

Next Friday (March 10) i'm gonna go for ASRock's Fatal1ty X370 Prof. Gaming if ASUS won't come out with something better looking than CVIH, April 10 - gonna buy me a EKWB's custom liquid cooling loop & G.Skill's Flare X 3466MHz RAM (2x8GB). I hope these RAM sticks will be validated by CPU-Z using this board by then, tho. May 10 - 1800X; let's hope that EFIs, drivers, OSes, etc... get proper Ryzen support by then. And i'd still go for i7 7700K rig as well, even though it'll take me all the way to October the least til it's complete. For comparison's sake, y'know? Cheers.

 

[birdie]

My dear, you mentioned that AVX on RyZen sucks, I showed you that it is really quite good, now you ask me why I speak about AVX? You forgot what you posted above?

You told us that power consumption sucks, I showed you it doesn't.

Now you bring this stupidity with core i5 2500 being on par with Ryzen on geekbench 4.

They are not even close.

i5 2500 is about 3500 single core, 9000 multi core, probably the ones with 4000 are heavy overclocks, as the geekbench doesn't detect very well the cpu speed, but look at the averages. I also see many with 3000, which probably means that is the default value with no overclock.
https://browser.primatelabs.com/v4/cpu/search?q=i5+2500

R7 is about 4300 single, 20000 multi core, numbers seem much more consistent.
https://browser.primatelabs.com/v4/cpu/search?utf8=✓&q=1800x

See the difference? Its HUGE.

i7 7700 has better single thread performance at about 4800, and this is the reason for which in lightly threaded loads, games, it beats Ryzen, but 2500 is pretty much dusted by Ryzen, from all perspectives.

I'm not saying by far that Ryzen is perfect, but the problems it has are not the ones pointed by you and its definitely not a train wreck, its a decent part.

AVX2 in Ryzen is half rate as shown by numerous tests, as shown by AMD themselves, IOW AVX2 is two times slower vs competing Intel CPUs. Still gonna say it's fast?

The Geekbench results I showed are from Intel Core 2500 without K which cannot be overclocked by definition. What's more it's my own CPU and I can run any tests you want to show it's not everclocked. Still gonna say it's overlocked?

TDP is really high and you've showed nothing against it.

While I can give you this this and this:

At least 117W:

At least 125W:

At least 106W:

God, why am I arguing with clinical idiots who also "like" each other posts?

 

[medi01]

Something crazy is going on with benchmarks, different sites compared:

http://www.3dcenter.org/news/amd-ry...tate-zur-anwendungs-performance-im-ueberblick

God, why

Don't you find some other place to troll?

 

[TheLaughingMan]

Something crazy is going on with benchmarks, different sites compared:

http://www.3dcenter.org/news/amd-ry...tate-zur-anwendungs-performance-im-ueberblick

Don't you find some other place to troll?

Different sites use different test methodologies. These include differences in hardware, differences in test settings, different tests entirely, and an overall focus of what they are trying to learn for their audience. Most of this is probably caused by the Ryzen series being a beast at multi-threaded workloads and some sites not including large numbers of games in their test results. You can't have an aggregate score comparison and say something is wrong when it doesn't tell you anything.

 

[Raevenlord]

@Raevenlord

i am baffled by the "missing chip" business.


1. Was it sent in regular mail or what?
2. Has it/they arrived yet
3. Is there any point now in doing your own review?

I don't take care of those logistics; as always, our resident magic handler is the one to conduct the reviewing process and all =) However, what I mentioned in the article is what happened: our sample really didn't arrive on time. However, it already has arrived.

Regarding whether or not it makes sense to make a review, I think the way TPU handles those is usually interesting. Besides, there are some other things to add to a review other than simple mathematical scores (those have been well documented already.) Ease of use, how did the install go, BIOS state, subjective experience... I think all of those have merit. I know for a fact the review is being conducted as we speak, so, let's see what we can come up with =)

 

[birdie]

Don't you find some other place to troll?

Providing facts is called trolling at TPU? Wow, seems like some guys here have fallen below the very lows of WCCFTech. This is my lost post in this otherwise bozotic discussion: AMD fanatics disregard any findings and hard reproducible facts that indicate that Ryzen is not the best CPU arch in the market and call you a troll.

Ryzen has an excellent performance/price ratio in certain scenarios but certainly not all of them. This is hard to swallow I get it, but that doesn't give you the right to call your opponents "trolls".

And the fact that Ryzen is on par with comparable Intel CPUs at 4K/2.5K only indicates the fact that games are stifled by the GPU and no computational power can address this problem unless GPUs becomes substantially faster.

Last but not least AMD's "finewine" will not work with Ryzen: 99.99% of existing games will never be recompiled to support AMD's arch.

A lot of stuff to read and digest:

https://forums.anandtech.com/threads/ryzen-strictly-technical.2500572/

http://www.overclock.net/t/1624566/...ryzen-and-some-recommendations-for-the-future

http://www.hardware.fr/articles/956-7/impact-smt-ht.html

http://www.hardware.fr/articles/956-22/retour-sous-systeme-memoire.html

 

[Prince Valiant]

I don't take care of those logistics; as always, our resident magic handler is the one to conduct the reviewing process and all =) However, what I mentioned in the article is what happened: our sample really didn't arrive on time. However, it already has arrived.

Regarding whether or not it makes sense to make a review, I think the way TPU handles those is usually interesting. Besides, there are some other things to add to a review other than simple mathematical scores (those have been well documented already.) Ease of use, how did the install go, BIOS state, subjective experience... I think all of those have merit. I know for a fact the review is being conducted as we speak, so, let's see what we can come up with =)

It's always good to have another review and a different perspective on things.

Snip

Maybe it's just me but, it might've had something to do with you calling several people idiots because they were arguing with you?

 

[Caring1]

Providing facts is called trolling at TPU?
AMD fanatics disregard any findings and hard reproducible facts that indicate that Ryzen is not the best CPU arch in the market and call you a troll.

Much like you call everyone AMD fanatics if they show any support at all for Ryzen, even if they acknowledge it doesn't live up to the hype created.

 

[Camm]

Last but not least AMD's "finewine" will not work with Ryzen: 99.99% of existing games will never be recompiled to support AMD's arch.l

The bigger problem with gaming speed doesn't look like its specific compilation, but that cache calls are going across core complexes, saturating the fabric between the two. This might be the Bulldozer argument all over again (and I'm really sorry if this is), but improved OS level schedulers that are more aware of where they are dumping threads may completely solve this problem.

 

[R0H1T]

Ryzen has an excellent performance/price ratio in certain scenarios but certainly not all of them. This is hard to swallow I get it, but that doesn't give you the right to call your opponents "trolls".

And the fact that Ryzen is on par with comparable Intel CPUs at 4K/2.5K only indicates the fact that games are stifled by the GPU and no computational power can address this problem unless GPUs becomes substantially faster.

Last but not least AMD's "finewine" will not work with Ryzen: 99.99% of existing games will never be recompiled to support AMD's arch.

Ten bucks says your wrong, as the next Scorpio is an AMD SoC just like the last two, & there's also a possibility that PS5 (or even Scorpio) is Zen based. Zen, where it stands today is excellent VFM, across the board.