SMT & Hyperthreading On vs. Off, & Validating FFXV Findings

Revisiting FFXV one last time to test CPUs, primarily hyperthreading & SMT on vs. off, alongside proving something about how Ansel's use is irrelevant to what's being rendered, despite claims by folks who didn't actually listen to the first video.
Ad: EVGA's store has GPUs for MSRP - goo.gl/XG9afy
Article here:
View our original findings here: • Misleading FFXV Benchm...
This benchmark tests CPUs in Final Fantasy XV, including AMD's R7 1700, AMD R5 1600(X), R3s, and Intel i7-7700K and 8700K CPUs. We primarily divert attention to SMT on vs. off, and hyperthreading on vs. off, as we stopped benchmarking after we realized the flaws in this software.
We have a new GN store: store.gamersne...
Like our content? Please consider becoming our Patron to support us: / gamersnexus
** Please like, comment, and subscribe for more! **
Follow us in these locations for more gaming and hardware updates:
t: / gamersnexus
f: / gamersnexus
w: www.gamersnexus...
Editorial, Test Lead: Steve Burke
Testing: Patrick Lathan
Video: Andrew Coleman
Links to Amazon and Newegg are typically monetized on our channel (affiliate links) and may return a commission of sales to us from the retailer. This is unrelated to the product manufacturer. Any advertisements or sponsorships are disclosed within the video ("this video is brought to you by") and above the fold in the description. We do not ever produce paid content or "sponsored content" (meaning that the content is our idea and is not funded externally aside from whatever ad placement is in the beginning) and we do not ever charge manufacturers for coverage.

[@GamersNexus]

Article: www.gamersnexus.net/game-bench/3227-ffxv-hyperthreading-smt-on-vs-off-benchmarks-cpu You might also like our initial research piece! kzbin.info/www/bejne/Zpa7k5WeZstoaKs

[@Zarcondeegrissom]

shame the data wasn't more useful without a lot more work by the devs, later on then, if things get better. It is interesting to see how unoptimized games behave. Great vid GN crew. B)

[@econ0611]

Gamers Nexus I have a question after reviewing your dissection of the rendering algorithms......could it be that the game is optimized to utilize local (e.g. user hardware) for multiplayer online rendering?? Hence drawing roads, vehicles, and gameworks on buffalos that are far away!? How would you validate that? Let's face it, single player does not earn revenue, it is all about micro transactions and season passes....correct?!!

[@ninomagazinovic]

Gamers Nexus WTF why not show 1700 OC with SMT On if you put 8700K OC SMT On. You never compare apple to apple, always some Intel or Nvidia push and false picture against Ryzen or Vega.

[@CasualGamers]

Thank you for the article link. It is missing on the description. Do we know what kind of engine this game is based of? Doesn't seem to allocate CPU resources similar to anything I have tested from SQ EX in the past.

[@wawaweewa9159]

You have to do it again! 😂

[@Kaleopan]

Excellent proof that overpowered hardware makes developers either lazy on optimization or skip entirely.

[@alb9229]

Kaleopan It's not overpowered hardware that makes developpers lazy on optimization , it's the fact that peoples will still buy their games regardless of the mediocre work .So why bother to optimize ? Masively boycott FF when released and look at those same developpers push a patch that fixes most of this BS at the speed of light. Sadly this trend isn't new in gaming industry !

[@Najvalsa]

GimpWorks isn't helping anything due to its design, but it's Square Anus too and not because of the hardware. Squeenix barely bother with their copy/paste game development, so why bother with a "benchmark" that they've now said isn't indicitive of game performance.

[@JoriDiculous]

Really makes you Wish Borland existed and did all development of games. Those guys always developed their software on the weaker machines then what most customers would run it on.

[@BubuSnow93]

This game doesn't deserve your professionality

[@JABelms]

Bagdadsky_Hornik Lol I have 873 games in Steam and I preordered the shit out of this

[@TheVeratrix]

It's interesting nonetheless.

[@halsaufschneider1446]

Best comment, I read in a long time.

[@jjkurtz03]

Strongly disagree. As a long time FF fan clear back to the original on NES, I loved XV. FF fans have this weird paradigm where they hate every new title until the next game and then suddenly the previous game they hated on becomes a fan favorite. It won't be any different with XV or XVI.

[@benscaruso3457]

Just upgraded to ryzen recently, and yes. Disabling SMT "like others say" will reduce memory controller workload, leading to more snappy 6 or 8 cores, instead of sluggish 12 or 16 cores. On newer games and on 1080p above its not really matters. But for old game which stability and stronger core is needed. Like when you modding skyrim for example, Disabling SMT lead to a much more smooth and stable gameplay. And btw, this "snappy" I talk about, you can really feel it when navigating even on desktop.

[@hopepeho]

Steve needs to get a job as an auctioneer. I'm impressed with how quickly he manages to spout off these numbers. That aside, I'd much rather see CPU benchmarks done at 720p/high than 1080p/low. You also need to factor system bandwidth into the equation as it's not just a matter of CPU vs. GPU. FFXV seems like it might be heavy on draw calls and one should never read too much into these so-called 'utilization' numbers. You can, for example, be bandwidth constricted yet see Windows' scheduler show it as 100% utilization. Personally, I like to use Intel's VTune tool to capture performance metrics and I strongly suggest that you guys buy a license for that if you're interested in seeing which threads are consuming the most resources. It's far more fine-grained information (including overall bandwidth consumption) than you'll get from the more basic tools and you can get a free trial version if you want to test it out.

[@lescarpio]

Thanks Steve, and GN for doing the hard works on this! Like another commenter said, you spent possibly way too much time on this than you really needed to haha.

[@ToriRocksAmos]

We all know that HT doesn't equate to the performance of actually having twice the cores. For both Intel and AMD, a SMT/hyperthreaded virtual core is worth around 40% of a real core. You can figure this out either by comparing SMT/HT on vs Off perf metrics in applications that scale (almost) perfectly with cores, or you can compare sku's of the same generation with and without SMT/HT at similar / identical clock speeds. Therefore, the i7 7700k has the performance of about 5.6 real cores. 5.6 x 5 = 28 8 x 4 = 32 Therefore I'm not really surprised that a real 8 core outperforms a hyperthreaded quad core. But it's nice to see that the game is actually optimised for 8 cores/threads.

[@InfiltrateIndustries]

That very much depends

[@timothygibney159]

That is not necessarily so. IN benchmarks yes where you load 100% usage per core what you say is true. In real life latency spikes hit when you have many threads competing for too few cores virtual or real. If you could put something like game logic on a virtual core and rendering on a real core you will see a nice performance boost as the game logic may use only 35% CPU usage per core max. THe rendering will need a real core and will not stutter due to a latency spike waiting for the logic to catch up. In this case HT/SMT can really help especially if you have things like Chrome running in the background rendering ads. An i3 may seem as fast as an i7 for light gaming and workloads but will choke fast when doing many things. So yes HT is a plus

[@shensai101]

I love how you can take over control of Noctis and run around the map. All the while the benchmark numbers are still going up...

[@0ExG0]

Off topic: every time i see a GPU AD I feel bad and angry about their prices...

[@Stikkzz]

theres a dead cat on the blue cars hood

[@mydogbuddy07]

@ 3:10 for anyone curious to find it. is it actually dead? or is it just sleeping?

[@stupidburp]

I vaguely recall from my college days that 8 threads was the theoretical sweet spot for most end user applications. Less than that and you probably are not dividing up enough tasks. More than that and the cost of management of work between more threads begins to outweigh the benefits for many common tasks. Of course there are specialized tasks that can take advantage of many more threads but these are usually not found in common applications such as games or office software.

[@asgallant1219]

Gamers Nexus, how do you feel about the consistency issues in the benchmark? I ran though it multiple times on my system with the exact same settings, and saw swings of up to 15% from run to run (even excluding the glitchy first runs). I noticed that two scenes in particular seemed to be bad for the benchmark: the chocobo ride and the mech fight. The chocobo ride scene seems to be dependent on the path the actors take through the scene, which changes from run to run. I had one run where the actors got stuck behind trees and rocks, making the scene take about 10 seconds longer than normal, which racked up significant extra points in the score. The mech fight may not be an issue, if objects are being rendered even when off screen, but if they are not, then the change in camera angles and actor positions from run to run is moving objects into and out of the rendered scene from run to run. Particle effects from explosions should be particularly punishing, but when they happen off-screen, if they're not rendered, the score for that run doesn't take the hit.

[@issaciams]

Is it weird I still have never played a Final Fantasy game in my 25+ years of playing video games? Great breakdown once again Steve.

[@timothygibney159]

I played it ... on a SNES back in the day. Amazing it still is out there

[@3800S1]

Have you considered testing USB2 (universe sandbox 2) as a CPU core/thread scaling benchmark? On my PC I found it very CPU heavy in some of the larger complex simulations with what arrears to be virtually perfect core/thread scaling. It also it very light on the GPU as the graphics side of things is simple in this simulation game.

[@buzzman4860]

This channel rocks

[@Najvalsa]

Metal pun?

[@dusfim3140]

Any ETA on partner Vega card reviews? please :D

[@techfusionaz2496]

FYI, were you guys running Windows Game Mode or not? I'm assuming not, but just wondering because I've seen Windows 10 FCU automatically enable game mode on several of my games.

[@Draksyl]

Horizontally inverted image during the EVGA advert? :)

[@MarceloTezza]

I want to know how it performs with optane and similar solutions in the mix. And memory frequency - size scaling

[@TheRealKoltoroc]

That the game favors precisely 8 threads is not really surprising, considering it is primarily developed for PS4 and Xbox consoles, which just so happen to have 8 CPU cores.

[@plitzi2640]

What is the program called, that is showing the rendered probs?

[@Bang4BuckPCGamer]

I can see a launch delay on this game due to all this feedback. But if they sort it before launch then happy days

[@charliebrownau]

2018 and games still dont correctly use multi core and multi threading and HT............... Can we finally stop using DX11 and just migrate to VULKAN ?

[@ecromancer]

Did you try this with a new BIOS update on AMD Ryzen boards? I have a R5 1500X @ 3.7 GHz with X370 ASRock Taichi and DDR4 3200 MHz. I'm getting almost 100 points better with BIOS 4.40 than I did with 3.20.

[@theoneyoudontsee8315]

no smt/hyperthreading on vs off will never end dont fool youre self. ive got a i7-8700k hyperthreading off hitting 5ghz without deliding a few times for customers. one time now i hit 5.3ghz hyperthreading off after deliding with only a 240mm aio water cooler.

[@BLKBRDSR71]

"It's still rendering that bird miles away". Well at least you know your Chocobo is going to be where you left it. 😜

[@MakoRuu]

I really hope they optimize this game before launch. I can't afford a $600 3GB 1060 just to play at 1080p. My 2GB 960 is going to explode trying to run this.

[@2ndLastJedi]

11:02 unless you had a Titan V in the bench!

[@nikolaosmitsou85]

little maronn

[@Yoshimatsu414]

I got a bit further in the game, around chapter 14 going around Insomnia. At certain parts of the city my framerate drop from the flat 60fps to about 40fps and my RX Vega 56 is not even being utilized 100% I open up task manager and I see that one of my threads on my R7 1700 (SMT on at slight OC of 3.4GHz) is getting killed at around 90s% while the rest are sitting around 30 to 40% That's a hardcore bottleneck due to the game just not utilizing the threads correctly like you found. IDK if this is the case in some of the other places in the game because I just started wanting to play at 60FPS at this part of the game.....just felt like it, it's the choice I have being a PC gamer :)

[@Pandemonium088]

Thanks for doing my request

[@ok-B]

Hopefully the devs are listening to this analysis.

[@stuartturner9138]

Could the rendering of distant, off-screen objects be a method of preloading assets to reduce hitching when loading chunks of the game world?

[@nintendoeats]

Those objects can be loaded without being rendered. There is a point in the rendering pipeline where the engine asks "should I account for this object in this frame?". All you need to do is answer "no" until the object is near the player's field of view. The out-of-frame objects will still consume memory, but the only computational resources they take up is the simple vector math required to determine that they don't need to be rendered.

[@jeistarGames]

sheesh this whole time i thought we were gonna get a Persona 5 port

[@nintendoeats]

I don't understand what you are saying about GPU usage. As long as none of the cores are peaking, can't we safely say that it is not a bottleneck? I feel like there is some fundemental concept that I am missing here.

[@OdinAlgeron]

it will never peak if you are not using all cores, MOST games wont so even if you are not hitting 100% cpu it may still be bottlenecking if it has shitty IPC windows can spread the work around between cores to maintain low temps, this causes the illusion of multi core usage so even though a game may only use a single core, windows can divide the work to all your cores say you have 6 cores, the game only uses 1 at 100%. you will see all six cores at ~16.5% usage its better to look at gpu utilization

[@nintendoeats]

But I monitor my cores individually and IIRC he was talking over a chart of core usage when he said that.

[@chrisrosenkreuz23]

wtf is up with the trees hitboxes

[@DoctorWho14615]

I think you are discriminating against the buffalo, its a life form and does not deserve to be talked badly about.

[@CrazyPCgamercatguy]

Where are AMD's 1800 and 1800X? Am I missing something?

[@larsjrgensen5975]

its performance is right between 1700 stock and 4GHz oc. The 1700 is often shown the most because its the best R7 performance per cost and can be overclocked to mostly the same as the 1700+ and 1800+.

[@CrazyPCgamercatguy]

So, they use the higher model from intel but not from amd? Or is there another intel cpu near these prices that they omit?

[@Munchticles]

What's the point, I mean I guess a stock 1700x and 1800x could be used, but overclocking the 1700 is just as good as the other two when overclocked.

[@CrazyPCgamercatguy]

So an 1700X OCed has the same performance as an 1800X OCed?

[@Munchticles]

Unless there has been a bios update or something that could make them overclock better. They have about the same 4ghz overclock last I checked and with all R7 CPUs being 8c/16t they will perform the same.

[@zuhairkyo9814]

Good work

[@llawliet9443]

Great video thnx

[@vincentvanrijn7469]

Why even advertise for GPUs when whey are expensive AF (advertise other stuff that is at MSRP)

[@somebody1476]

vincent van Rijn because they sponsor Gamers Nexus it is advertised regardless pricing

[@gnsteve8846]

If you actually clicked the link on the ad, you'd see that it goes to the EVGA store, where GPUs are actually listed at MSRP.

[@orkhepaj]

it is too expensive , nobody should buy it, those others are retards

[@Lead_Foot]

If you actually looked at the link in the description it says evga's store has gpus for msrp

[@buzzman4860]

Prices came down. I bought a msi gtx 1080 for $50 over retail from NewEgg directly (not a reseller on NewEgg)

[@RyanStark-sy6xm]

Are you a metalhead by any chance??

[@gameboy3800]

use titanv and see if gpu bottleneck is gone

[@rustyheckler8766]

Well, it is a game I'll never play. Never understood the appeal of the FF games.

[@Najvalsa]

Welp that pretty much goes for anything right? People have different tastes and all.

[@chosenundead3174]

And i never understood how can you like video games and not like the exploration, giant bosses, stories and fantasy worlds of this series.

[@MrMilkyCoco]

Maybe he just doesnt wanna play a boy band.

[@rustyheckler8766]

Chosen Undead Too little kidish for me. Weak plots and premise, character's have a one, maybe two at best, dimensions, but this is my take away from the series; to each their own right.

[@chosenundead3174]

I disagree completely, i find the characters (especially on this one) very real with their own personalities and dynamic and you could literally say it's too kiddish about almost every RPG ever with their fantasy elements but yeah, everyone has their own taste.

[@saladwithsalad]

i5 users are gonna love those stutters LUL 👌👌 😂😂😉😉

[@WhiteoutTech]

Batemen wondering why there is no quad core i5 in the benchmark. people still think a 4/4 cpu is good .-.

[@vividhercules270]

When he used the r7 1700 with 4 threads it did pretty much the same...

[@JABelms]

CryBeat I tested it out with my GFs i5 4690K@4.5ghz and my old GTX 980 that I tried out at 1.5Ghz, long story short it pretty much matches a Vega 56 in the benchmark table. 4C/4T is still good, why you ask? 95% of the games in the world use 2C-4C/4T. If you have a library of 500+ in played games like I do, you will be playing those games for many years, would probably be done with them when CPUs are already Quantum level as standard.

[@saladwithsalad]

ViviD HERCULES 4 cores are struggling,it doesn't matter if its intel or amd,however more people have i5s

[@OdinAlgeron]

LOL not all 4 core cpu's are equal, a 4690k overclocked a bit is VASTLY superior to s POS like a amd athlon 880k

[@Lucian_Andries]

Imagine the consoles trying to run the game........ :)

[@franesustic988]

well.. they are running it, for a year.

[@leonidus499]

FF15 has been on console for a year now...

[@Lucian_Andries]

What's your point? Being and running on a console, is a totally different thing.

[@Waldo1122]

+Lucian Andries My Ps4 runs it just fine, but I'd rather play games on my PC.

[@Najvalsa]

Lucian Andries This is a console ported game, not the other way around. The PS4 doesn't use GW since it's AMD hardware, and it's more finetuned for that hardware too, so it runs well comparitively.

[@procerator]

First time Ryzen beats 7700K

[@johnwheels89]

Only Truth same gpu with different cpu...the ryzen beats it Also since I got my ram at 3400 on latest bios update my ryzen rig is doing better in some games than my 77k rig with same gpu

[@will3641]

Pretty sure it is the broken SE engine. This made me wonder if the PS4/xBox version is like this as well. SE need to just forget about this engine, and use Unreal or something else.

[@Macho_Man_Randy_Savage]

Devs did say this game was graphically intensive and that your high end PC would struggle to run it! Just another way of saying poorly optimized! lol Sure I read somewhere that they were using SLI to test the game yet it doesn't have it now?

[@WereCatStudio]

RIP 3DMark... new king of benchmarks is here! /s

[@kainhall]

so a stock 7700k gets the SAME as a stock 1700..... yet..."ryzen is bad at playing games"....... will ya look at that.....a powerful gpu isnt bottle-necked by either.....its 99.99995% the gpu

[@vividhercules270]

Ryzen isn't bad at gaming, but it isn't as good as Intel's offerings. Ryzen doesn't beat the 7700k in gaming and doesn't even come close to the 8700k. It isn't bad, it's just not the best choice

[@chosenundead3174]

It's the same because of the gpu bottleneck. Any game that relies more on the cpu shows intel's superiority in games. I love how you cling to one single game and forget about ryzens own reviews comparing to 7700k or 8700k lol

[@stupidburp]

In theory there could be some games where the extra cores and threads would always have a clear advantage over higher clocks with fewer cores and threads. In practice this is only true for a small portion of games. Some future games could potentially make this more noticeable, such as some grand strategy titles that have many processes running simultaneously but each with relatively easy work loads. This could be part of numerous simultaneous AI decision making algorithms for example. For the more popular titles which tend to be shooters there isn't usually enough enemies for this to matter much and many are multi-player focused which leaves the decision making to lots of humans. But a shooter could potentially be made with an enormous number of truly independent enemies making their own decisions and thus provide more challenge for players. The old Mount and Blade Warband game sacrificed some visuals in order to make each enemy truly independent as well as modeling the physics and damage modifiers for the velocity and direction of each swing, thrust, projectile, and mounted troop. Other games could potentially have more enemies in huge battles where Warband was limited to a few hundred on most hardware because of this design decision. Those other games would use ways to simulate independent actions and simplify physics modeling in order to have more acting enemies on the same hardware.

[@timothygibney159]

That depends on the game. IWth this one Ryzen is better. I believe as time progresses we will see more and more threaded gaming engines. The consoles have had 8 cores since 2013. These days in 2018 folks have at least an i5 with 4 cores. Thanks to Ryzen and Coffee lake we see 6 cores on medium grade cpus like the i5 8600K. Game makers will simply keep the threading on the console ports rather than cripple them to make the i3 users happy (FYI the i3 8300 is 4 core too!)

[@JABelms]

Timothy Gibney Existing consoles can't and will never use 8 cores...even if consoles desperately use all 8 cores, they are too weak anyway to be of any benefit in the long run. Consoles have 8 cores for better QoL/multitasking, not for gaming and console performance almost entirely comes from the iGPU. 6 cores will be the next frontier in 10 years in terms of gaming.

[@sultanalzabin6247]

like always, lazy developers

[@standuporshutup]

"1080p low"...55 .1% lows...on a 1080 Ti. For shame devs, for shame.

[@onthespoke2]

Final Fantasy runs better if you turn off Shin Megami Tensei first? Yeah no shit.

[@AAHAHHHHH]

"SMT" shin megami tensei?

[@hassanrizal1160]

Wow r7 1700 beat 7700k. cant wait to see r7 2700 beat 8700k

[@yottaXT]

In games that developers actually patched full compatibility for Ryzen it perform a lot better, but Intel Still lead in gaming, that's easy to realize due to it's higher performance corexcore. I'm not really expecting that Ryzen soon would be able to beat Intel at any game, but actually in massive task they do perform a lot better, which at least for my uses are the important thing.

[@OdinAlgeron]

It doesnt, gpu bottleneck prevents proper cpu utilization this is why people remove the gpu as the bottleneck when comparing cpu's by running at low resolutions XD most morons dont know this and say: "wow r7 1700 beat 7700k"

[@OdinAlgeron]

did you fall headfirst into a well? How about something actually constructive?

[@timothygibney159]

About time developers starting using more cores. Until 2016 developers still tested freaking core2Dous so they limited their thread count. With the newer i5s using 6 cores and even the coffeeLake i3s using 4 cores it's time developers starting using their resources

[@Tinfoilpain]

You guys should charge for this kind of thing.

[@kainhall]

why is enix making ANOTHER final fantasy game (might as well re-name it "everlasting fantasy" ) and not making a 3rd kane and lynch game that could tie up the story....to prevent me from saying shit like this...... also..... someone ask gearbox where the FUCK the 4th and final brothers in arms game is.......i need to know if baker likes snow!

[@mydogbuddy07]

Theirs a story behind the name, i think the TL;DR is that Square at the time were going out of business and Final fantasy was going to be their final game, but then it sold very well and they stayed in business (i think that's what happened, just recalling this from memory so i'm fuzzy on the details). I remember people not really liking the last Kane and lynch game, so that could be why, did it end up selling well? or not? cause if it also didn't sell well then that would be why.

[@Munchticles]

Well, technically Enix never made a Final Fantasy game, they made Dragon Warrior. Square Soft made Final Fantasy as a last ditch effort to save the company and it worked. Then sometime in the early 2000's Square Soft and Enix merged to become Square Enix. Also this game has been out for over a year. Kane and Lynch was an Eidos titles which Square Enix bought out and other than Tomb Raider they've not done too well with Eidos titles from what I recall.

[@pandoodl]

Please fix your table, it wobbles too much!

[@halsaufschneider1446]

This "bad" benchmark, basically shows what is wrong with modern games & developers. A massive amount of unnecessary crap is drawn, when its not only, NOT needed. Its still drawn when its not even there! It just happens, that Asscreed Origins, is just another contender of bad development. The game runs like azz! Even if there is nothing really to see, aside from sand. Unless sand is not a texture, but every mote a Voxel! Then yes the performance tanks.

[@orkhepaj]

why all these FFXV vids?

[@chrispbacon177]

Because a lot of people are interested in the upcoming launch.

[@howardkurtus7069]

I'm definitely going to buy it

[@TheGokki]

It's less about the game and more about the benchmarking tool that allows them to, uh, benchmark for the hardware reviews you enjoy.

[@orkhepaj]

but he says it is a crap

[@gnsteve8846]

Right. And how do you think we discovered that? It's the same reason we post a bunch of CPU videos with a new CPU launch -- to learn about them.

[@Rob-dd7dm]

Long story short: this game will be shit

[@andreibolohan686]

He wrote 8700k SMT0 in one of the charts, dislike and unsubscribe :D

[@skyhawk21]

I am curious why this benchmark is so important on benching a system? It's glitchy, buggy and pre code for the game. So the stutters, frame times and lag are making this not the all to be benchmark to use. You need a benchmark that runs correctly and is finalized code like 3dmark. This ffx bench is a joke, the textures look dumped down in it. Steve your great at what you do, give up on this crap game code called a benchmark.

[@reav3rtm]

Lack of other stuff to cover perhaps. Besides guys have a chance to fiddle with software, so it's likely interesting for them to do.

[@armandb.8737]

losing interest on pc hardware lately, hope just me cause such channels will suffer

[@FAHRENHEITDELIO]

shity game, pirate yes..

[@ranruy]

AMD released a code for the current bios update and it fucked alot of machines up. Can't overclock anymore or even revert back to old bios. Gg

[@BreakingDimes]

what mobo do you have because i am not having this issue

[@ranruy]

BreakingDimes 0_0 the MSI x370 gaming pro .

[@Najvalsa]

Lots of people with MSI boards seem to have this issue. Maybe Gigabyte owners too, but no one else it seems. So it's MSI, not AMD.

[@ranruy]

SlavjanA it's been reported that its not just MSI with this issue.

[@glosteradv]

I have asus b350 and overclocked my ryzen 5 1600 to 4.0ghz without a problem with the latest bios, and its rock stable at 1.32v. Really it could be MSI related.

[@ntme9]

Why is EVGA spending money on advertisements? They literally can't keep any graphics cards in stock.

[@vnyggi621]

ntme9 not because they couldnt if they wanted

[@rurutuM]

Because sponsorships are worked out months before these videos are released

[@mud2479]

the people from marketing have to spend the designated budget

[@ntme9]

They should be pushin there mobos n power supplys if anything

[@dive_whale]

ntme9 They do push out Mobos, idk power supplies, but GPUs get a set amount of turns and then Mobos get their set amount of turns in the AD-spot. Rinse and repeat.

[@Najvalsa]

facebook.com/ffxv/posts/1946290452354936 "A Level Of Detail (LOD) issue has been discovered that affects the benchmark scores. The benchmark also suffers from stuttering; both of the issues will be addressed in the shipping game. While the settings in the benchmark are largely pre-selected for simplicity’s sake, the final game will feature highly customizable settings with On/Off options for a wide variety of settings. The benchmark will give you an idea of how beautiful the game will be upon release, but for the reasons stated above, the benchmark may not accurately reflect the game’s final performance." So it's not a benchmark for the game now.

[@dragon2knight]

So...all this data (well most of it anyway...) doesn't really matter anyway due to the flawed FFXV benchmarking software. At least it get's some results anyway. Look forward to the real testing once this finally comes out, as I'm sure you guys are too :-)

[@thatkoalabear]

i agree bro.. thumbs up for you mate =)

[@FluffyFarron]

Michael Livote The most positive/hopeful comment in this hate section. 😂😂😂😂😂

[@motoktips3024]

Gameworks doin it's thing to fuck up AMD performance, like all other Gameworks games.

[@MrMilkyCoco]

It makes work easier on enix...blame them for using it.

[@hopepeho]

I'm pretty sure that nvidia Gameworks is intended for nvidia GPUs. If you don't have one of those then you should probably switch it off.

[@nottheengineer4957]

But it is a lot worse for AMD, so they just crank it all the way up to make nvidia look better.

[@ArtisChronicles]

+Jimmy The Scumbag the problem is that reviewers that just run the game with all the settings as high as they can get them reflect the absolutely horrid performance on AMD cards. People just look at the numbers and basically say AMD is shit as a result of that. You can usually get competitive frames or respectable numbers by leaving those options off, even on Nvidia cards. Especially older Nvidia cards as they don't get the current optimizations to deal with gameworks features. It wouldn't be that bad if AMD was better able to optimize for it. Then it would just be some lower end cards and older Nvidia cards suffering from gameworks features.

[@capNsgt]

lol. its F'n up Nvidia performance as well.

[@f4z0]

When i tested my 5820K + 1700 i got higher score setting affinity to 8 threads rather than letting the full cpu being utilized. 6700 score on 8 threads vs 6200 on 12 threads. This is definitely borked.

[@JayJayYUP]

EDIT: As for the video, nice 1080p medium bottleneck. Good luck solving this level of software mess in a month and a half.. HT off since I got my 4790K Works better for 90% of games I've played (haven't tested too many, so take that as you might).

[@x1c3x]

M how much better?

[@JayJayYUP]

Depends on the game, but the biggest gain is not raw peak framerate, it's mostly a far more smoother framerate curve, and less fluctuations. Also you can eek out ~500mhz on air on all 4 cores easily by getting rid of HT for instance.

[@sebastian_hakansson78]

Good testing, as always! To bad it seems like a trend that some new games are unoptimised today and use to much or inconsistent resources (PUBG!).

[@stephenzenerak7846]

6:57 - isn't load balancing just a PR term for thread bouncing, if your more or less single threaded game thread jumps from core to core it's not like it's being worked on simultaneously by two cores, it can't really, make it any faster compared to if it just sticked to one core, if there are improvements, they are small, maybe due to some cache being prepared on the other core, also, jumping on the same physical core but on a different logical one is also pointless, this is all a visual/averaging appearance and illusion it gives in the Task Manager, as if the weight it's balanced, that balancing may help something else, but not peak performance by much I'm afraid. Infact, a thread jumping to another core would probably need more time for that core's cache to populate, IMO.

[@Nerd-Freaki91]

i have a 13700k with hyperthreading on i can have 5.4 p 4.2 eco 4.8 ring but without hyperthreading i can have 5.6 p 4.2 eco 4.8 ring but allso lower temps and better minimum fps. allso without hyperthreading u wont see as blurry bad graphics as with on. im a huge fan of single core for gaming but u wont be happy if ur streaming or using many backgrounds apps at the same time.

[@blackwidow7804]

This game was peace of shit on consoles and it looks like it will be even bigger peace of shit on PC.

[@octoman_games]

This game benchmark tool renders "Hairworks" on objects not even on the screen IMO to make AMD look bad

[@Rocky-bz8wr]

GN great video as always but I have to disagree on the using a lower res to remove the GPU bottleneck and how it would be pointless. No most if not all people would not play at such low res but if you are having a hard time being able to remove GPU bottleneck then using lower res which now allows the CPU to stretch it legs more and perform closer to it's fullest performance then that now gives you a whole new set of data points to put into your charts. The other issue was about CPU usage yes you want a game or software to use your CPU & hardware to it's fullest but at the same time you DO NOT want the CPU stuck at 100% all of the time in a game this is a very bad thing it does not leave any headroom if needed for the CPU if more performance is all of a sudden needed.So basically if it was already maxed out at or near 100% and more horsepower was needed not it will fall behind and create things like stutter etc etc. A good balance of usage is around 45%-75% range if a CPU is always in that range you are well balanced on the load up of the cores/threads that are on hand. You may get some spikes into the 80%-90% range from time to time and those are exactly what I was talking about if the CPU was always stuck at or near 100% if it was then when those spikes happen and it was already near maxed out then there is no fuel left in the tank to be able to cope with it and you get slow downs or extreme stutters all of the time. A prime example is the quad core i5's and now the i3's if you watch those game channels with live game play happening most of the time the quad cores 4/4 are most always stuck at 85%-100% usage in modern games and you clearly see the side effects of it in choppy stutter'y game play.

[@duncanfirth]

I have been having massive problems with Ryzen and Virtual Box, have tried SMT on have tried SMT off, I even tried changing the advanced acceleration settings and changing the settings I nearly have given up, maybe I should sell my Ryzen and buy an Intel. I keep getting BSOD Bluescreens with my Windows Server 2016 VM. Anyone have any suggestions on what I should do? The other VM's arnt free VMWare, Hyper V etc you have to pay.

[@Rararawr]

I don't understand how some AAA games manage to perform worse with more power. When I played a forza demo with my FX 6350, two cores were 90-100%, and the rest were at 40, 30, 0, and 0%. It didn't run well, probably because the two cores it did use weren't fast enough

[@shadowolf3998]

FFXV is one of the few games that utilized everything that you throw at it... or at least almost everything but it situational honestly and we are more concerned about lower demand online games rather than this game to see if they are benefiting from disabling SMT/HT or not since they only utilize 1-4 cores depending on the title, test games like Diablo 3, Fortnite or Smite.

[@oldgamergene5712]

if the benchmark kills the highest performing GPU - what really does it accomplish? It seems, as noted, it needs alot of work yet.

[@johnwheels89]

So should I turn smt off on my ryzen r7?

[@Dr.WhetFarts]

yes smt sucks for gaming.

[@lrmcatspaw1]

If it is Bethesda its a feature. EDIT: Wait, Stock 1700 has higher minim framerates than OCed to 4 GHZ? (SMT0).

[@ujiltromm7358]

"GTX 1080 Ti FTW3, among the best gaming cards you can get right now" _But at what cost? AT WHAT COST?!_

[@ChibiTheEdgehog]

This reflects very poorly on what this game will perform like when its officially released

[@carlkitchen]

I actually found that 6 threads with SMT disabled was optimal on my R7 1700. Had a slight performance bump over 8.

[@Flyinspaghetti]

Guess the new thing to ask when testing GPU's is: "But can it run FFXV?"

[@NativeWarrior081]

Wish the Chocobo's used hairworks, "cough" I mean featherworks.

[@farisx1425]

Did they say anything about fixing the pop ins?

[@Dia1Up]

I wonder what would happen with a 1950x

[@nomadicnome234]

Totally got my mod mat yesterday!!!! GN ftw

[@Aggrofool]

Luminous Engine is a wonky beast

[@mariodrv]

Use the TitanV then?

[@patrikmiskovic791]

Great job GN

[@CharlesVanNoland]

In RenderDoc you can clearly see that it's drawing the meshes per the DrawIndexed D3D API call down on the bottom left. The fact that it's running vertex/fragment shaders also indicates that this is also not just mesh data being uploaded to the GPU - as they would not be executed for a simple vertex buffer upload.

[@gnsteve8846]

RenderDoc also shows textures being called for things that aren't in the screen. Checked it after your comment. We're seeing the cat texture (from the dock), for example, being called in the first driving scene.

[@CharlesVanNoland]

The GPU will cull the triangles once it discovers they're outside the frustum - if they actually are, and are not being drawn way off in the distance behind everything else, but all these little vertex shader costs add up, as well as the individual draw calls themselves. It really looks like their engine is doing a really poor job of minimizing the amount of activity that's taking place between the CPU and GPU, to the point of embarrassment! Props for using RenderDoc to analyze their engine.

[@tee_es_bee]

Great piece, keep up the good work!

[@azumag4432]

How about a i5 6600K (stock) with a GTX 1060 6GB....it will be a CPU bottleneck...???

[@dethskullcrusher]

naaah, but you should overclock that processor either way.

[@Rappoltt]

I think the GameWorks Disabled footage looks better, even solely from a visual quality standpoint, not FPS. The only practical difference is the grass, and the way the terrain shadowing pops in at a certain range with GW on looks filthy.