Especially as turn time and not FPS becomes the limiting factor late into the game. Anything above 20fps is decent, but if turns take a minute then it's intolerable.
I have logs including minimum FPS, but time constraints kept me from spending too much effort analyzing them. What I can say is that when the game zooms all the way out during the benchmark sequence, that's when the minimum occurs and it's usually around 2/3 of the average FPS. GPUs with less VRAM may also be hit harder though. Let me see if I can add some charts for minimums, now that I've managed to get some sleep. :-)
Okay, I added a third page where we investigate minimum FPS. Mantle actually helps here quite a bit, so I need to tweak the conclusion a bit. (CrossFire on the other hand is still not working right with Mantle, at least in my testing.)
I'm using an average cpu with an above average gpu. Some games choke my cpu but only at new levels / checkpoints so it doesn't bother me, but others stress my cpu enough to cause occasional slowdowns.
Either way, your interpretation of min.fps is the most accurate and least misleading that I've seen. Love it!
I'm also a big fan of Civ5 and I'll get this game as soon as possible, but I wonder.
Maybe the "upgrader" would be a more realistic use-case, like: I have 300$, what do I get? Do I upgrade the ye-old Phenom II to an i5 and stay with the same 7950 (Mantle can't help much) or I get a 290 with the old CPU and count on Mantle to relieve the CPU load?
Personally since the TBS is CPU-bound when hitting 'end turn', as the GPU won't help make turns faster, I would go with the CPU upgrade and keep the 7950. Unless you're trying to go 2k-4k resolution, you won't get much GPU benefit. CPU overhead for GPU usage is not very much.
Let's say I don't want to take your word. Let's say I want to see a credible, hands-down test (like the ones I enjoy reading in Anandtech) and then decide for myself.
Human eyes can see the jagged edges really easily, so you still want AA. Less AA than 2K, but you still want some. I mean, I can see jagged edges on my Xpera Z2 (over 400ppi) at the same distance as my monitor when I run a game with AA off (I have 4x MSAA forced on for everything OpenGL though dev options)
There's no correlation between ultra-high resolution and ultra-high density. Different screens will have their own mix of the two.
For a desktop system, you have A) a bigger monitor with less density and B) extra power to throw at your graphics systems, so it makes sense to anti-alias.
(I can pick out aliasing of an anti-aliased rotating line on a 200ppi screen from several feet away.)
It's a $330 CPU running up to what was once $1100 worth of GPU (2 x R9 290X). It's more CPU than a lot of people have, but OC i7-4770K isn't much faster than an overclocked i5-4670K that only costs $220. Most gamers have something roughly at the level of the i5-4670K is my feeling; if they don't, then they probably don't have an R9 or GTX graphics card either. :-)
(And yes, there are always exceptions, but we're talking big picture here.)
That assumes that people are building their PCs today. There a plenty of people still with i7-2500k/FX-8320 era CPUs and mid-high end GPUs who are interested in effects of Mantle on game performance.
That would be me. I am running on an FX-4100 OC to 4Ghz. My R9 280X is my "tock". I will tick again when windows 10 comes out and get a new CPU etc.
I think a lot of us with 1 GB video cards found ourselves upgrading with the latest set of games hungry for more texture memory even if everything else was "acceptable".
My hope is that mantle will mitigate the sub-par processor a bit.
My informal benchmark with the FX-4100 and the R9 280X was that mantle provided a significant benefit on the part of the bench at the end where it pans out. The scrolling was smoother and I was seeing 30 fps with mantle. With DX11 the scrolling struggled and I was in the teens a few times. Generally Mantle was showing a 25% increase in the extreme cases.
In the earlier part of the bench where it is a little more zoomed in I saw the same moderate 5fps or so difference that others are seeing.
Did you (and is it possible) to test whether or not Mantle has an impact on turn times (for AI calculations) if it does reduce CPU overhead for rendering?
Exactly what I'd like to see! Civ has always been a cpu bound game, especially toward the endgame. Try playing a large map with 8-12 civs in the modern age. You can go make a coffee between each turn, makes it nearly unplayable. With mantle's biggest benefit being cpu limited scenarios, which so far has meant pairing a highend gpu with a low end cpu, I'd like to see what kind of impact it has on turn times. I've been waiting to see those kinds of test since they announced Beyond earth would be supporting mantle. I'd be interesting to see what it can do with some high end cpu's when the game can actually push them.
Keep in mind that the computer turns are not really doing anything with graphics, so basically Mantle would have little to no impact on speeding up the AI.
Has AMD ever presented Mantle as a big boost for systems with an overclocked $350 CPU?
I thought Mantle was supposed to be a performance enabler for low-power and/or cheaper CPUs... For your viewers' sake, please repeat the Mantle benchmarks using low-end and midrange CPUs.
I think the this part of the review captures the intended message. "There are definitely quieter Radeon R9 GPUs, but even then I don't think they're going to be hard pressed to match the GTX 970 or 980". So you believe they are going to match the GTX 970 and 980 with ease? I believe dropping the "don't" will get across the point you wanted to make. But of course, unless you are testing a MSI R9 290X lightning or Sapphire Vapor-X, perhaps you can leave out speculation in a data driven review entirely.
Yeah, the "don't" was a mistake. Basically, a GPU drawing 250W is going to be hard pressed to match the noise profile of a GPU drawing 160W. That's all I'm saying.
Well it is not just your 970, my GTX 570 do avg 25 fps in 2560x1440 ultra 4xmsaa. Still ok for strategy. However in 1920x1080 same settings the game fly! Got 45-50 fps. Even in 2048x1080 same settings almost the same result. Like the new nvidia driver. Nice smooth in the game.
Considering most people won't be buying reference model of either brands, and especially so since GTX970 doesn't even have one, we'd very much like to see benchmark done with AIB partner boards with custom coolers that are overcloked right out of the box, because that is, ironically, a much more realistic portrayal of what consumers will be basing their decisions upon.
I have three reference designs (GTX 980, GTX 770, and R9 280X); everything else is a card purchased at retail. I'm a little leery of overclocking too much, but I know the R9 280 cards in particular can easily hit 950/5600 clocks. I suppose I should also note that all of the cards had their power limits increased, where applicable. (The 280/280X/290X were set to +20%, the GTX 780/970 are +6%, and the GTX 980 is +25%. I didn't actually overclock the cores or RAM, however.)
Anyway, I generally disagree with testing heavily overclocked cards, as it can give false impressions, and even mild overclocks aren't always guaranteed. The results I'm showing should be achievable with every GPU tested. If you have a card that's overclocked, it should perform better than what we're showing, but as not all cards overclock to the same level that can be a slippery slope. :-)
Well, i got EVGA 570 classified so i don't need to bother myself with oc. This is what the card do on stock. Will make vid as well if i can get some sleep :D
I looked the invisible page 3. Not sure I'd only appears on the mobile site. Also, it's good to see the 250x benchmarks thrown in. As a two-year-old gtx 660m owner, I appreciate the low end benchmarks.
I added page 3 after the initial posting (and some much needed sleep). I just checked on my smartphone and it loads okay, so can you try reloading the page? Let me know if it still doesn't work.
I had pretty much the same result. Just tested 2560x1440 4xMSAA ULTRA(with high option i have the same as well) with GTX 570 and avg fps is 24.5 . In 1920x1080p just lagging behind of the R9 280 AMD vga. Not bad from Nvidia.
Hold on there... the average of the R9 280 at 2560x1440 Ultra is 41FPS, and at 1080p it's 54 FPS. If you're at 24.5 FPS, your GTX 570 is about 50% faster than the slowest GPU I tested, the R7 250X (which is a $100 GPU, sometimes less).
Yup, though the splitting was Scan-Line Interleave (each GPU did every other line), which is messy if you want to do things like AA. Other SFR methods include vertical and horizontal splits, which can result in unequal workloads, or breaking the scene up into tiles. I believe CivBE is using tiles for SFR; more complex in some ways, but overall a good way to improve performance.
Does Mantle have any impact on thermal output of the video card? I would imagine extra processing means extra heat. Could this be an issue for people who have already OC'd their GPU?
Any plans on a "CPU vs. time between turns" benchmark (and/or a mantle / no mantle vs. time between turns)? These kinds of games I don't care too much about visuals, but I do care about time between turns.
I'm a little puzzled by the benchmark. I have an I7-4770K, 8gb ram, and a Radeon 5870. When I play at 1080p with High settings with 2X MSAA, I get 55-65 fps. This is general game play not the benchmark, but to me is makes the benchmark a bit misleading.
The benchmark is more of a "worst-case scenario" where there are tons of units on screen at once. Earlier in the game in particular, frame rates will be much higher, and even later in the game you often won't have as many units as shown in the benchmark. However, for people that play the game a lot and in particular on larger maps, things can really bog down. That's what the benchmark shows and while it's not perfect, it's at least repeatable.
So far it seems like anything that can run Civ V should run Civ: Beyond Earth just as easily, which should include most recent integrated GPUs as well, though of course these may draw performance from the CPU which could slow your turns down later in the game.
Regarding the Mantle performance; I would have been interested to see some measurements for AMD's APU's, perhaps in a follow-up article? While the improvements to discrete graphics are nice I find Mantle most interesting in the APU space, particularly with HSA, though I don't know if any titles support that yet?
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
72 Comments
Back to Article
Parrdacc - Thursday, October 23, 2014 - link
Well, have a Happy Birthday!!przemo_li - Thursday, October 23, 2014 - link
I think that You make good point about $$ on CPU vs $$ on GPU... for now.Cause if Mantle can let us spent $$$ less on CPU then, why not?
(That ofc. would require more widespread support in games.... Or DX12 Or OpenGL AZDO)
nevertell - Thursday, October 23, 2014 - link
Not necessarily. Civilization V was a cpu limited game for the most part.LemmingOverlord - Friday, October 24, 2014 - link
indeed, Civ V was a great CPU benchmark (akin to Supreme Commander), and the performance scaled quite well with added cores...SuperVeloce - Friday, October 24, 2014 - link
Civ V scaled well with up to 4 cores if my memory serves me right. Not much more with 6+ cores.just4U - Friday, October 24, 2014 - link
I was playing civ5 this past week while waiting for the new game... and you know... even on my 4790K it bogs down on big maps.jaredjeya - Friday, October 24, 2014 - link
Especially as turn time and not FPS becomes the limiting factor late into the game. Anything above 20fps is decent, but if turns take a minute then it's intolerable.doronnac - Thursday, October 23, 2014 - link
Min. FPS should be tested as it's more important than average when almost all cards demonstrate more than playable performance even at 4K.Mikemk - Thursday, October 23, 2014 - link
"the built-in benchmark basically represents something of a worst-case scenario for performance"Page 1
JarredWalton - Thursday, October 23, 2014 - link
I have logs including minimum FPS, but time constraints kept me from spending too much effort analyzing them. What I can say is that when the game zooms all the way out during the benchmark sequence, that's when the minimum occurs and it's usually around 2/3 of the average FPS. GPUs with less VRAM may also be hit harder though. Let me see if I can add some charts for minimums, now that I've managed to get some sleep. :-)JarredWalton - Thursday, October 23, 2014 - link
Okay, I added a third page where we investigate minimum FPS. Mantle actually helps here quite a bit, so I need to tweak the conclusion a bit. (CrossFire on the other hand is still not working right with Mantle, at least in my testing.)StevoLincolnite - Thursday, October 23, 2014 - link
Unfortunately, Mantle doesn't really shine on mid/high-end hardware, it's the low-end stuff that it really helps.Would be nice to see to see how a Pentium Anniversary, Core i3, AMD FX Octo, Core i7 5930K and 5960 handle it.
Alexvrb - Thursday, October 23, 2014 - link
Agreed! Especially FM2+ chips. You'll naturally be testing lower settings and resolutions in such a scenario, however.doronnac - Friday, October 24, 2014 - link
Thanks for the data Jarred.I'm using an average cpu with an above average gpu.
Some games choke my cpu but only at new levels / checkpoints so it doesn't bother me, but others stress my cpu enough to cause occasional slowdowns.
Either way, your interpretation of min.fps is the most accurate and least misleading that I've seen. Love it!
Iketh - Saturday, October 25, 2014 - link
even more reason for min frame rateGich - Thursday, October 23, 2014 - link
Try Mantle "with an AMD APU or a Core i3 processor" and NOT a "hundreds of dollars GPU"?Like an i3 + R7 260.
dragosmp - Thursday, October 23, 2014 - link
I'm also a big fan of Civ5 and I'll get this game as soon as possible, but I wonder.Maybe the "upgrader" would be a more realistic use-case, like: I have 300$, what do I get? Do I upgrade the ye-old Phenom II to an i5 and stay with the same 7950 (Mantle can't help much) or I get a 290 with the old CPU and count on Mantle to relieve the CPU load?
Drasca - Thursday, October 23, 2014 - link
Personally since the TBS is CPU-bound when hitting 'end turn', as the GPU won't help make turns faster, I would go with the CPU upgrade and keep the 7950. Unless you're trying to go 2k-4k resolution, you won't get much GPU benefit. CPU overhead for GPU usage is not very much.looncraz - Thursday, October 23, 2014 - link
Upgrade the CPU. I am running a 7870XT (a.k.a. 7930) with an i7 2600k @ 4.5GHz and nothing slows me down.Investing in a good SSD, however, may make even more sense.
looncraz - Thursday, October 23, 2014 - link
I should also note, that the 7950 is helped by Mantle, as is my 7870XT. Any GCN graphics card works.ToTTenTranz - Thursday, October 23, 2014 - link
Let's say I don't want to take your word.Let's say I want to see a credible, hands-down test (like the ones I enjoy reading in Anandtech) and then decide for myself.
looncraz - Thursday, October 23, 2014 - link
+1 I want to know how this game can play on an A8-7600 (my HTPC which games surprisingly well - Hitman: Absolution is killer on a 65" LED TV!).sinisterDei - Friday, October 24, 2014 - link
Gich,I've run a couple benchmarks- one on a system similar to your recommendation- an i3 running a 7850. Results as follows:
-benchmark lateGameView
i3-3240 @ 3.4GHz / XFX 7850 2GB
Windows 8.1 Pro, 8 GB DDR3 @ 1600
1920x1200 Fullscreen Ultra@4xMSAA
VSYNC Off
Mantle Avg FPS: 36.67187776
DX11 Avg FPS: 31.26703968
Gich - Friday, October 24, 2014 - link
Thanks.Well here we have a 17% difference...
AndrewJacksonZA - Thursday, October 23, 2014 - link
Hi JarredAn early Happy Birthday from me too! :-)
Thanks for the review. This looks like a good excuse to motivate an upgrade to my E6750 and 6670...
On the last page I'm seeing placeholder text for an image:
"
frankly I'll believe it when I see it.
{gallery 4008}
As far as the game itself is concerned,
"
Thank you
Andrew
Essence_of_War - Thursday, October 23, 2014 - link
"I suspect if I were to retest everything with an AMD APU or a Core i3 processor, Mantle might prove more efficacious"I'd certainly be interested in those results, and results from intel's igpu w/ DX as well, of course.
Happy Birthday in advance!
MikeMurphy - Thursday, October 23, 2014 - link
Why incur the performance penalty of 4x MSAA when it's absolutely wasted at ultra high resolutions like 4K?I'm curious what the results are without MSAA.
ZeDestructor - Thursday, October 23, 2014 - link
Human eyes can see the jagged edges really easily, so you still want AA. Less AA than 2K, but you still want some. I mean, I can see jagged edges on my Xpera Z2 (over 400ppi) at the same distance as my monitor when I run a game with AA off (I have 4x MSAA forced on for everything OpenGL though dev options)mkozakewich - Thursday, October 23, 2014 - link
There's no correlation between ultra-high resolution and ultra-high density. Different screens will have their own mix of the two.For a desktop system, you have A) a bigger monitor with less density and B) extra power to throw at your graphics systems, so it makes sense to anti-alias.
(I can pick out aliasing of an anti-aliased rotating line on a 200ppi screen from several feet away.)
Artuk - Thursday, October 23, 2014 - link
I would have liked to see the impact of mantle on lower end CPUs.I also disagree that these represent an "average" gaming machine if you are running an OC 4770K.
JarredWalton - Thursday, October 23, 2014 - link
It's a $330 CPU running up to what was once $1100 worth of GPU (2 x R9 290X). It's more CPU than a lot of people have, but OC i7-4770K isn't much faster than an overclocked i5-4670K that only costs $220. Most gamers have something roughly at the level of the i5-4670K is my feeling; if they don't, then they probably don't have an R9 or GTX graphics card either. :-)(And yes, there are always exceptions, but we're talking big picture here.)
Kalelovil - Thursday, October 23, 2014 - link
That assumes that people are building their PCs today.There a plenty of people still with i7-2500k/FX-8320 era CPUs and mid-high end GPUs who are interested in effects of Mantle on game performance.
Artuk - Friday, October 24, 2014 - link
That would be me. I am running on an FX-4100 OC to 4Ghz. My R9 280X is my "tock". I will tick again when windows 10 comes out and get a new CPU etc.I think a lot of us with 1 GB video cards found ourselves upgrading with the latest set of games hungry for more texture memory even if everything else was "acceptable".
My hope is that mantle will mitigate the sub-par processor a bit.
Artuk - Saturday, October 25, 2014 - link
My informal benchmark with the FX-4100 and the R9 280X was that mantle provided a significant benefit on the part of the bench at the end where it pans out. The scrolling was smoother and I was seeing 30 fps with mantle. With DX11 the scrolling struggled and I was in the teens a few times. Generally Mantle was showing a 25% increase in the extreme cases.In the earlier part of the bench where it is a little more zoomed in I saw the same moderate 5fps or so difference that others are seeing.
YMMV
limitedaccess - Thursday, October 23, 2014 - link
Is it possible to test the performance and playability with IGPs, particularly at 1080p (but lower settings)?If Beyond Earth retains the touch controls from Civ 5, then it would be a rather suitable game for hybrids/convertibles (eg. Surface Pro).
JarredWalton - Thursday, October 23, 2014 - link
I can add a bit of testing today on IGP and see how that goes... stay tuned.garbagedisposal - Friday, October 24, 2014 - link
Also interested in this, thanks for testing.Stuka87 - Thursday, October 23, 2014 - link
Mantle needs to be tested with a slower GPU to really shine. Testing it with one of the fastest CPU's out is not really the best test for Mantle.Running the same test with a slower CPU (say 2-3 years old) should show a bigger jump.
limitedaccess - Thursday, October 23, 2014 - link
Did you (and is it possible) to test whether or not Mantle has an impact on turn times (for AI calculations) if it does reduce CPU overhead for rendering?SviGG - Thursday, October 23, 2014 - link
Exactly what I'd like to see! Civ has always been a cpu bound game, especially toward the endgame. Try playing a large map with 8-12 civs in the modern age. You can go make a coffee between each turn, makes it nearly unplayable. With mantle's biggest benefit being cpu limited scenarios, which so far has meant pairing a highend gpu with a low end cpu, I'd like to see what kind of impact it has on turn times. I've been waiting to see those kinds of test since they announced Beyond earth would be supporting mantle. I'd be interesting to see what it can do with some high end cpu's when the game can actually push them.JarredWalton - Thursday, October 23, 2014 - link
Keep in mind that the computer turns are not really doing anything with graphics, so basically Mantle would have little to no impact on speeding up the AI.Ryan Smith - Thursday, October 23, 2014 - link
The built in "lategameview" benchmark does not test turn times. It's entirely a rendering benchmark.ToTTenTranz - Thursday, October 23, 2014 - link
Has AMD ever presented Mantle as a big boost for systems with an overclocked $350 CPU?I thought Mantle was supposed to be a performance enabler for low-power and/or cheaper CPUs...
For your viewers' sake, please repeat the Mantle benchmarks using low-end and midrange CPUs.
ajlueke - Thursday, October 23, 2014 - link
I think the this part of the review captures the intended message."There are definitely quieter Radeon R9 GPUs, but even then I don't think they're going to be hard pressed to match the GTX 970 or 980". So you believe they are going to match the GTX 970 and 980 with ease?
I believe dropping the "don't" will get across the point you wanted to make. But of course, unless you are testing a MSI R9 290X lightning or Sapphire Vapor-X, perhaps you can leave out speculation in a data driven review entirely.
JarredWalton - Thursday, October 23, 2014 - link
Yeah, the "don't" was a mistake. Basically, a GPU drawing 250W is going to be hard pressed to match the noise profile of a GPU drawing 160W. That's all I'm saying.BreezeDM - Thursday, October 23, 2014 - link
is there a Civ: BE computational benchmark like there was for Civ V?TelstarTOS - Thursday, October 23, 2014 - link
Nice article, thanks :)Happy to read that my 970 is more than capable at 1440p.
siriq - Thursday, October 23, 2014 - link
Well it is not just your 970, my GTX 570 do avg 25 fps in 2560x1440 ultra 4xmsaa. Still ok for strategy. However in 1920x1080 same settings the game fly! Got 45-50 fps. Even in 2048x1080 same settings almost the same result. Like the new nvidia driver. Nice smooth in the game.zeock9 - Thursday, October 23, 2014 - link
Considering most people won't be buying reference model of either brands, and especially so since GTX970 doesn't even have one, we'd very much like to see benchmark done with AIB partner boards with custom coolers that are overcloked right out of the box, because that is, ironically, a much more realistic portrayal of what consumers will be basing their decisions upon.JarredWalton - Thursday, October 23, 2014 - link
I have three reference designs (GTX 980, GTX 770, and R9 280X); everything else is a card purchased at retail. I'm a little leery of overclocking too much, but I know the R9 280 cards in particular can easily hit 950/5600 clocks. I suppose I should also note that all of the cards had their power limits increased, where applicable. (The 280/280X/290X were set to +20%, the GTX 780/970 are +6%, and the GTX 980 is +25%. I didn't actually overclock the cores or RAM, however.)Anyway, I generally disagree with testing heavily overclocked cards, as it can give false impressions, and even mild overclocks aren't always guaranteed. The results I'm showing should be achievable with every GPU tested. If you have a card that's overclocked, it should perform better than what we're showing, but as not all cards overclock to the same level that can be a slippery slope. :-)
siriq - Thursday, October 23, 2014 - link
Well, i got EVGA 570 classified so i don't need to bother myself with oc. This is what the card do on stock. Will make vid as well if i can get some sleep :Dsiriq - Thursday, October 23, 2014 - link
It is just funny i am right after r280 with few fps behind.JarredWalton - Thursday, October 23, 2014 - link
How are you testing, and how are you calculating the FPS? If you're just using FRAPS and not running the benchmark, you won't get comparable results.siriq - Friday, October 24, 2014 - link
-benchmark lategameview in the shortcut.feeblegoat - Thursday, October 23, 2014 - link
I looked the invisible page 3. Not sure I'd only appears on the mobile site. Also, it's good to see the 250x benchmarks thrown in. As a two-year-old gtx 660m owner, I appreciate the low end benchmarks.JarredWalton - Friday, October 24, 2014 - link
I added page 3 after the initial posting (and some much needed sleep). I just checked on my smartphone and it loads okay, so can you try reloading the page? Let me know if it still doesn't work.Oxford Guy - Friday, October 24, 2014 - link
If you're planning to do a bit more testing, I'd like to see how the 480 or 580 manage to do.RagingCain - Friday, October 24, 2014 - link
I will be building a Frame Time Analyzer, similar to my BF4 FTA for Mantle/DX11 users.OCN THread: http://www.overclock.net/t/1520615/civ-be-fta-civi...
My Website: http://www.bytemedev.com/bf4-fta/
siriq - Wednesday, October 29, 2014 - link
I had pretty much the same result. Just tested 2560x1440 4xMSAA ULTRA(with high option i have the same as well) with GTX 570 and avg fps is 24.5 . In 1920x1080p just lagging behind of the R9 280 AMD vga. Not bad from Nvidia.JarredWalton - Thursday, October 30, 2014 - link
Hold on there... the average of the R9 280 at 2560x1440 Ultra is 41FPS, and at 1080p it's 54 FPS. If you're at 24.5 FPS, your GTX 570 is about 50% faster than the slowest GPU I tested, the R7 250X (which is a $100 GPU, sometimes less).RagingCain - Thursday, October 30, 2014 - link
I updated CIV: BE FTA to v0.0.2, there was a minor bug on low FPS users, I also added graphing capabilities.RagingCain - Thursday, October 30, 2014 - link
Direct Link: http://www.bytemedev.com/civbefta/JlHADJOE - Friday, October 24, 2014 - link
Kinda funny to be going back to SFR now. IIRC it was how the original SLI worked back during the 3Dfx days.JarredWalton - Friday, October 24, 2014 - link
Yup, though the splitting was Scan-Line Interleave (each GPU did every other line), which is messy if you want to do things like AA. Other SFR methods include vertical and horizontal splits, which can result in unequal workloads, or breaking the scene up into tiles. I believe CivBE is using tiles for SFR; more complex in some ways, but overall a good way to improve performance.just4U - Friday, October 24, 2014 - link
You having some fun with the game while you test it Jarred? lol.. I am just downloading it right now.. OH it's finished.. time to play!zoomzoomzealot - Friday, October 24, 2014 - link
Does Mantle have any impact on thermal output of the video card? I would imagine extra processing means extra heat. Could this be an issue for people who have already OC'd their GPU?zoomzoomzealot - Friday, October 24, 2014 - link
Another piece of that question would be power usage increases?jaredjeya - Friday, October 24, 2014 - link
My name and middle names are Jared Anand. This is an article on Anand Tech by a guy called Jar(r)ed. perfect.Concillian - Friday, October 24, 2014 - link
Any plans on a "CPU vs. time between turns" benchmark (and/or a mantle / no mantle vs. time between turns)? These kinds of games I don't care too much about visuals, but I do care about time between turns.liquid_mage - Sunday, November 2, 2014 - link
I'm a little puzzled by the benchmark. I have an I7-4770K, 8gb ram, and a Radeon 5870. When I play at 1080p with High settings with 2X MSAA, I get 55-65 fps. This is general game play not the benchmark, but to me is makes the benchmark a bit misleading.JarredWalton - Monday, November 3, 2014 - link
The benchmark is more of a "worst-case scenario" where there are tons of units on screen at once. Earlier in the game in particular, frame rates will be much higher, and even later in the game you often won't have as many units as shown in the benchmark. However, for people that play the game a lot and in particular on larger maps, things can really bog down. That's what the benchmark shows and while it's not perfect, it's at least repeatable.Haravikk - Thursday, November 6, 2014 - link
So far it seems like anything that can run Civ V should run Civ: Beyond Earth just as easily, which should include most recent integrated GPUs as well, though of course these may draw performance from the CPU which could slow your turns down later in the game.Regarding the Mantle performance; I would have been interested to see some measurements for AMD's APU's, perhaps in a follow-up article? While the improvements to discrete graphics are nice I find Mantle most interesting in the APU space, particularly with HSA, though I don't know if any titles support that yet?