Homeworld Remastered - FX gets hammered

[Ramses]

I believe it. Can't test it though, those games make me want to claw my eyes out.

Bring on the System Shock reboot though.

 

[Gikaseixas]

This is an odd conversation in its own right because the definition of success appears to be "winning the academic argument".

The real answer should be "it (be it software or hardware) is crap if it fails to generate the sales and revenue that it was expected to generate for the business owners and shareholders of the business that invested in creating said hardware or software".

The hardware is crap if its market viability squarely depends on unaffiliated businesses electing to create enabling software for said hardware, only the software then never shows up. Put a fork in the hardware, its done.

But you know what else was crap in that equation? The business leader who made the decision at the hardware company to create a product that would flop in the market if not for a software company or three who were needed to swoop in (time the market) and release the software that would create a viable market for the hardware.

If you are in the business of creating hardware that is going to live or die on the basis of software being available to make the hardware desirable, then you need to be in the business of ALSO making the software.

Like Apple did. Or IBM, or HP, or Cray, or just about any other well run business back in their day of owning and rolling niche hardware features.

I've seen many high quality products (hardware) with low sales in the market place. It's something not exclusive to PC components and it's a complex phenomenon.
It could be marketing, price, fierce competition, etc. This industry is filled with collaborations that include financial incentives. Many business side with certain partners in exchange of particular guarantees, contracts, assistance with in-house projects, investments and assurances.
In the end you're not completely wrong but sometimes we need to look at entire picture if one wants to be fair.

 

[SAAA]

I'd have to guess its the caches. Nothing else makes much sense

That's it. My old core2quad is going well with 12MB of L2, almost the same as my newer i7 in the laptop.

its probably a ST game...

Clock speed helps too and considering the game doesn't load much more than 2 cores (still 1 for AI) I understand why a Haswell Pentium can beat 2 threads of a higher clocked FX chip... IPC difference is just too large there, even if cache is larger on the latter.

True but if you play the original version of Homeworld which is included in the remastered package you will find the game runs much better. The TechSpot test isn't even possible in the original version due to the unit cap. There is no way CPU's back in 2003 could handle hundreds of ships from just one player, let alone eight.

I remember clearly being able to use hundreds of ships in this game exactly because it uses OpenGL thus it's cpu limited only when you have too much going on, but not with static units or just many of them on the map.
It was when bullets went left and right that you saw dips, now the increased graphics make things different: example is when a ship explode it spawn too much junk and after the boom my older cpu lags a bit until everything disappears.
Graphics setting can take a huge toll with a few lightning options enabled, depth of field in particular cuts FPS like to a third while giving an effect i don't like much... in space it shouldn't be that visible and if you want just to play smooth in big battles turn it off!

 

[svenge]

I've seen many high quality products (hardware) with low sales in the market place. It's something not exclusive to PC components and it's a complex phenomenon.
It could be marketing, price, fierce competition, etc. This industry is filled with collaborations that include financial incentives. Many business side with certain partners in exchange of particular guarantees, contracts, assistance with in-house projects, investments and assurances.
In the end you're not completely wrong but sometimes we need to look at entire picture if one wants to be fair.

But in the case of the Bulldozer and its descendents, it seems pretty simple. AMD thought that software was universally going to be highly multi-threaded in the very near future and oriented their "moar cores" architecture around that. As we all know that didn't come to pass (excepting deterministic things like video encoding), and in the end having many weak cores became a severe hindrance in a world where two to four cores at most typically matter in terms of software performance.

Accordingly, sales of AMD's CPUs plummeted which fed a self-fulfilling cycle of future software not being tailored to its relative advantages. It is best described with this question: Why would commercial software developers invest the time and money to tailor their software to a platform with ever-diminishing market share when it was already performing satisfactorily with the market leader's platform?

 

[TurtleCrusher]

You aren't wrong. At all.

But I refuse to be a part of the business/financial end of things. This is a hobby, and the stuff I outlined above seems to be a relatively recent development, least in my world, and it's bumming me out.
I'm interested in hardware for the sake of the hardware. The same reason I like a nice pocket knife or a zippo lighter or my fifty year old adjustable crescent wrench. The same reason I spend many times the value on old cars. The same reason I rebuild old abandoned household stuff with new guts. They are quality hardware, and it's nothing to do with anything remotely associated with business or finances. I get enough of that crap at work.

The only other thing I can say is I've seen some outstanding devices fail miserably financially. That didn't make them crap imo. Bad business decisions associated with them, sure, but the hardware is pure. It does not care how many it sold or how it was reviewed, it just is. And I strive to evaluate and value such things on their own merits, independent of profit or marketing or competition(to a degree). It's no different than the way I judge a forty year old car, they are universally piles of trash compared to anything and everything current in every measurable way, but there is more to the machine than that.

It's when you have people who talk on this forum who pretend to be enthusiasts, but really are shareholders trying to improve their bottom lines. Just look at any quarterly result thread. The usuals are always right there parroting the same nonsense.

I, along with others, left the tech side of the forums for years until recently to get some more research before buying new products. You'd think the forums are full of green compositing prius-driving hippies who hang dry their clothes in the dead of winter and don't run their heat because 30w a few hours a day is a moral sin. It's nuts because a lot of it is the same people who overclocked Nehalem i7s to 300w ranges.

 

[nemesismk2]

It looks like this game doesn't benefit much from anything over 2 cores.

wicked i only have 2 cores with my amd athlon ii x2 270 but has 3.4Ghz performance.

 

[AtenRa]

That's the AMD mantra...just wait and things will be better later. The problem is later never arrives.

Much like Intels iGPUs ???

 

[Erenhardt]

Dat scaling from fx4300 to fx8300 .
Truly a CPU intensive tests that leverages the peformance of chip!

 

[Shehriazad]

I wonder how a 4-4.5 Ghz Athlon 860K would do. Lack of cache yes...but the IPC is definitely higher than those FX chips.


Also...why is only the OpenGL version benchmarked...doesn't the game also come with DX?

 

[BigDaveX]

Where have I heard this whole "It's a forward-looking architecture and it'll start doing better when people design their code for it?" thing before? Oh yeah, the Pentium 4 and the Itanium. And no-one gave Intel a free pass on that.

In fact, AMD were the ones who came out with a chip that trashed its competition in current code, and further extended its lead when newer 64-bit code came into play. It was such a successful strategy that this time ten years ago, after the Athlon 64 X2 had been released, a lot of people genuinely thought we were seeing the beginning of the end of Intel's dominance over the CPU market. Somewhere along the line, AMD and their hardcore fans seemed to forget that lesson.

FX-8350 has 16MB cache

I was referring more to the ranking among the Intel CPUs, as there's no way that a Sandy Bridge-E should be posting noticeably better scores than a desktop Haswell in a gaming test. Given the age of the game code, it wouldn't surprise me if the game engine was able to run from entirely within the larger cache on the LGA2011 chips. I remember back in the day the Pentium 4 Emergency Edition posted some insane performance figures in Quake III, as the game was able to fit entirely in its (comparatively) huge L3 cache, so we're probably seeing something similar here.

As for the FX-8350, the cache system is slower, and split into a 4x2MB L2 + 8MB L3 structure, versus the single 15-20MB L3 cache on the higher-end Intel chips. It's also possible that the amount of legacy code in the game is causing bottlenecks elsewhere, and preventing the FX's cache from helping out as much as it could.

 

[soccerballtux]

there was nothing forward looking about the P4, it had bad branch prediction and a super deep pipeline so that marketing could sell more gigahertz.

Itanium was very forward looking and IMO a fantastic architecture, we just never had enough PhDs to optimize the compiler to get meaning IPC out of it.

lol @ Emergency Edition haha.

 

[Enigmoid]

As for the FX-8350, the cache system is slower, and split into a 4x2MB L2 + 8MB L3 structure, versus the single 15-20MB L3 cache on the higher-end Intel chips. It's also possible that the amount of legacy code in the game is causing bottlenecks elsewhere, and preventing the FX's cache from helping out as much as it could.

As far as cache goes, AMD's is very slow (intel's L3 is a good bit faster than AMD L2). Intel's caching system is inclusive (L2 duplicated in L3) while AMD's is not. L2 is exclusive to a module. Therefore if the game is only using two threads half of the L2 cache is essentially useless (4 MB) as it cannot be used at a given time. The L3 is quite slow and I would imagine that snooping in another modules cache is quite slow simply given the distance between modules in the FX series (essentially on the other end of the chip). Quite possibly as you mentioned the legacy code is requiring constant access to the same data meaning that a lot of cache snooping is taking place with the AI.

 

[Ramses]

Itanium was very forward looking and IMO a fantastic architecture, we just never had enough PhDs to optimize the compiler to get meaning IPC out of it.

I thought that was a really neat CPU too but I was getting out of PC's for awhile just as it came to be, when I came back I see it never took off and now everyone talks trash about it. I suspect but do not know, that it's another case of software magically ruining hardware.

 

[Phynaz]

You mean like Intel left the desktop with Broadwell?

Much like Intels iGPUs ???

You guys can't help yourself or something? This thread is about Homeword on FX, not about Intel CPU's.

But I'll humor your derail:

Broadwell on desktop coming mid year
http://techreport.com/news/27911/socketed-intel-desktop-broadwell-coming-mid-year. Includes Iris Pro graphics.

Intel partners with Raptor to optimize games on Iris graphics
http://techreport.com/news/27912/in...r-to-optimize-game-settings-for-iris-graphics

 

[soccerballtux]

I thought that was a really neat CPU too but I was getting out of PC's for awhile just as it came to be, when I came back I see it never took off and now everyone talks trash about it. I suspect but do not know, that it's another case of software magically ruining hardware.

I think it was more that Intel was using it to get out of the x86 cross licensing with AMD; and AMD decided just to go from 32->64 bits somehow by doubling some 4-word thing to 8-words or something IDK.

That, and trying to write an optimized compiler to superscalar was very, very difficult and only doable on heavy database accesses, which can be parallelized anyways...

 

[Fox5]

there was nothing forward looking about the P4, it had bad branch prediction and a super deep pipeline so that marketing could sell more gigahertz.

Itanium was very forward looking and IMO a fantastic architecture, we just never had enough PhDs to optimize the compiler to get meaning IPC out of it.

lol @ Emergency Edition haha.

A lot of the technologies that made the Pentium 4 workable as a design are what gives Intel its IPC advantage over AMD now. There wasn't really any technology designed for the P4 that wasn't applicable to all future architectures.

Itanium has some heritage in modern Intel chips as well.

 

[Fox5]

I think it was more that Intel was using it to get out of the x86 cross licensing with AMD; and AMD decided just to go from 32->64 bits somehow by doubling some 4-word thing to 8-words or something IDK.

That, and trying to write an optimized compiler to superscalar was very, very difficult and only doable on heavy database accesses, which can be parallelized anyways...

All modern processors are superscalar, it allows for higher clock speeds.

Itanium was EPIC, "Explicitly Parallel Instruction Computing". IMO, it's not much different than some of the SSE and AVX instructions supported by modern Intel processors. Those are also hard to write compilers for, and Intel is just about the only one who has a compiler that can use them well.

 

[xthetenth]

You guys can't help yourself or something? This thread is about Homeword on FX, not about Intel CPU's.

But I'll humor your derail:

Broadwell on desktop coming mid year
http://techreport.com/news/27911/socketed-intel-desktop-broadwell-coming-mid-year. Includes Iris Pro graphics.

Intel partners with Raptor to optimize games on Iris graphics
http://techreport.com/news/27912/in...r-to-optimize-game-settings-for-iris-graphics

And those comments are about AMD "leaving" the desktop with a part that will possibly have desktop versions and isn't the only product in the stack, not Intel CPUs. Neither company is leaving the desktop.

 

[bunnyfubbles]

Is there some extra instruction set or feature Haswell has that FX doesn't?

Comparing the FX-4320 to the 8370 in the first graph, it looks like it's scaling with clockspeed only, whereas comparing the i7s to the i3s make it very clear that CPU load is multithreaded.

Without CPU utilization graphs from an FX system, I'm just guessing, but it honestly looks like the game engine isn't spawning threads correctly on AMD hardware - maybe it's setting core affinity wrong or something.

If that's the case, I'd expect a patch to fix FX multithreaded performance very, very soon.

I don't think that matters when the Sandy and Ivy CPUs are doing just about as well as Haswell...3rd fastest CPU is Sandy-E.

 

[escrow4]

Much like Intels iGPUs ???

Intel has been substantially improving its iGPU since SNB (just look at Broadwell), AMDs CPUs are languishing in the meantime and their iGPUs can't make up for poor single threaded performance. Off topic anyway.

Techspot just shows once again how bad FX is. It will be interesting to see the Witcher III and GTA V benches in a few months.

 

[Ramses]

If Witcher III and GTA V are written to use more than a handfull of cores very well, they'll do decent. Not Intel fast, but decent. If not, then they will not. Pretty cut and dry at this point no? There are no bad CPU's, just ill matched software.

 

[AtenRa]

Intel has been substantially improving its iGPU since SNB

Still way behind even at the same or lower TDP.
http://forums.anandtech.com/showthread.php?t=2420502&highlight=

 

[Phynaz]

Still way behind even at the same or lower TDP.

As can be said for AMD's CPUs. Good thing I can buy an inexpensive add-on GPU to make the Intel CPU faster. Too bad I can't buy an inexpensive add-on to make the AMD CPU faster.

This is an undeniable fact.

 

[frozentundra123456]

If Witcher III and GTA V are written to use more than a handfull of cores very well, they'll do decent. Not Intel fast, but decent. If not, then they will not. Pretty cut and dry at this point no? There are no bad CPU's, just ill matched software.

Actually the software "is what it is", just some cpus are more well rounded than others and can run any software well.

 

[Ramses]

Actually the software "is what it is", just some cpus are more well rounded than others and can run any software well.

Does not jive with my experience but I agree that should be the standard re: software. The first time this came to my attention was 1997, with the Intel specific MMX instructions. I'm reminded to wonder what's going on behind the scenes whenever I start a game and see a huge Intel logo or NV or AMD. Crysis3 for example, ran amazing on my FX cpu's. There are lesser games that do not. That does not feel like "is what it is" to me in my very end-user non-programmer observation.