Thuban Gaming Benchmarks. The Shocking Truth.

I asked a friend of mine to make a spreadsheet up with the gaming results from most of the recent reviews, so i could make a review on them all. Here is how I did it.

---
The Rules

First of all, I decided that this had to be a review of realistic gaming benchmarks. The only thing of interest to me - as a gamer - is how well games will run at good settings and resolution. I immediately realised that if I did this wrong, I'd be accused of cherry picking benchmarks - so I decided upon the following rules which hopefully everybody will see as sensible and balanced.

1) The minimum resolution counted will be 1680x1050.

I realise a lot of people still use 1280x1024 screens, however most of these people simply would not use these high end cpu's for gaming.

2) The maximum resolution counted will be 1920x1200.

Very often, 2560x1200 is a total bottleneck on graphics cards - this claim is often made against AMD cpu's, so I felt that by leaving out 2560x1200, it should act as a counter to 1280x1024 and below.

3) With that in mind, I decided that if multiple benchmarks were made at the same resolution, different settings, I would take the middle setting, or the one closest to what is likely to be used.

Game benchmarks with 300fps, no AA, no AF etc...are out. However if the game was *only* benchmarked at those settings, they would be used - AA and AF settings would only be used in preference if available.

note - while looking through the list of benchmarks, this was hardly an issue. I believe only two out of all the websites I have used actually bothered with different settings at the same resolution.

----

The Data

I used all the gaming results that fitted the rules above from many websites. Altogether I think 65 benchmarks in total, all done at realistic gaming settings.

The websites were as follows.

Guru 3d
Hexus
Neoseeker
Pc Perspective
Bit Tech
Toms Hardware
Hardware Canucks
PCGH
Overclockers Club
Hardware Secrets
Computerbase
Anandtech
Overclock 3d

13 websites, 65 unique benchmarks.

----

The results.

I have uploaded my friends file here - http://www.filefront.com/16333651/gamebench.ods

It's an openoffice calc file, I dont know if you can use that with excel or not. I did not create this file, but I can assure you it is quite safe to use. The results and the resulting commentary will be in the second post.

Results and Commentary

Figuring out how to just give the results in an easy to understand fashion is actually quite difficult. Therefore, I opted for a simple chip vs chip comparison.


1090T vs 965 BE

11 websites benched the 1090T and 965 BE. Now before now, a lot had been made about the 1090T supposedly being a worse gaming cpu than the 965 BE. The actual results?

4610 fps for the 1090T
4601 fps for the 965 BE

It's close. The 965 BE really does run the 1090T close, but over 11 websites and probably 50+ benchmarks, the 1090T proves that its turbo mode does work. Just. The 1090T *is* faster than the 965 BE in gaming however, this is the proof.

1090T vs i7 980X

Both of these cpu's were benchmarked in 12/13 websites, so this is a pretty decent comparison.

5102 for the 1090T
5305 for the i7 980X

The i7 980X proves it's the best cpu money can buy. It's only 2-3% faster in reality, and it loses a lot of benchmarks too (you can check that with the spreadsheet), but there is no denying the 980X is the fastest gaming cpu you can buy.

1090T vs i7 870

The i7 870 isn't used in nearly as many websites benchmarks, but here are the results of the 5 websites that did.

2298 for the 1090T
2278 for the i7 870

Some people might be surprised by this, others maybe not. It's quite close, over 5 websites and around 30 benchmarks total.

1090T vs i5 750

Both these cpu's were used in most of the websites.

4664 for the 1090T
4635 for the i5 750

The i5 750 proves itself to be a nice gaming cpu. However, it loses to the 1090T in every single website except for Anandtech, where it wins by a strangely high margin. I will discuss this subject later. While my friend was making up the spreadsheet, I noticed how Anand's results hugely swung the overall totals in favour of some cpu's.

965 BE vs i5 750

Well I was suprised at how close the 965 BE ran the 1090T, and I noticed that the i5 750 was a bit short of the 1090T in most benchmarks, so I decided to compare these two next.

4166 for the 965 BE
4138 for the i5 750

Well well. And if you remove Anand's benchmarks the 965 BE is much further ahead.

1055T vs i5 750

By now I'm on a mission against the i5 750...but this one is a loss.

1847 for the 1055T
1954 for the i5 750

Thats 100 points ahead. I noticed all through this that the 1055T was underperforming in gaming. However - and this has to be said - of the 97 points the i5 750 beats the 1055T by, 90 - NINETY points of them comes from Anandtech.

Remove Anandtech's results and these cpu's are almost identical in gaming. This is an issue that will hopefully encourage discussion here.

1090T vs i7 930

This was the last one, there weren't many websites that benched both cpu's, but the results were...

845 for the 1090T
833 for the i7 930.

Another 'win' for the 1090T.

----


There is the absolute proof lads. The 1090T beats them all except Gulftown. Sure they are all very close - it's not like you would notice the difference, but the results speak for themselves.

If anybody finds a problem with the figures, please tell me and I'll have my friend fix the spreadsheet and update the results as required.

My only thought and hopefully your friend is bored enough and will add the higher resolution as well.

Since most gamers who do buy any of those cpus you mentioned will also buy a very high end GPU they will also buy a high res monitor. I know that there is bascially no difference between the chips at those res's but it would be nice to have them and even the lower res test on there just to having everything in one place. Not to actually count wins/losses but just to have all the info in front of ya.

If the person is even more bored can add OC results as well.

Yeah I hear ya. There is no way I would do it myself. Just the more in one place the better.

Wow. AMD is really, really getting back on track. Consider the fact that you can use them in AM2+ motherboards and good ol' DDR2 RAM makes Thuban a winner. Just wondering though, what's the approximate max MHz boost you can get out of Turbo CORE? (I find it lame, something like AMD Adrenaline would have sounded better. )

Gaming Benchmarks? Where are the system specs (especially the Graphics card(s) used?)

A nice compilation Jenny, I was thinking about starting another of my own too.

It is not very surprising, in most gaming benchmarks the 1090T will be the exact same as the 965 which is close enough to every other quadcore in that price range to be negligible.

Jenny, you have posted up results about Bad Company 2, do you know an article with this benchmark and possibly GTA 4 as well? I can't find them.

Now you just need to do one with apps... have fun.

OpenOffice can also save as Excel files too, IIRC. Or was it just .doc files...

Very interesting indeed. The price to performance is just unmatched. I really wonder how GTA4 will like the 6 core.

Because why buy the same performing CPU for $100 less when you can have 2 more cores that make nearly no difference? I guess thats the logic.

Plus she thinks people claimed Thuban wont beat a Core i5 750, she even thinks I did when I clearly stated it would be on par with a Core i7 930.

Anand will always be out of whack until they show AMD in the lime light. In fact I think thats why I like Anand a bit. They don't tend to just kiss butt for anyone. If the product sucks, they say that it sucks.

What's the shocking truth here? All CPU's in test performed about equal, as was to be expected.

The 1090T is basically a 965 with two additional cores and a few minor improvements. Given very few games can use 3-4 cores, it wasn't hard to guess that the 965 and 1090T were going to be equal.

That the 965 and i5-750 are about equal is another well established fact.

And what about the it-980X? 1090T is an improved 965. The 980X is an improved i7-930. As such, gaming performance should be roughly identical between the 930 and 980X. And that was, as we all knew before, going to be in the same area as the 750's performance.

So basically this proves a fact we all now very well: any CPU above the i5-750/ Phemom II 965 doesn't offer a benefit for gaming.

Conclusion: the i7-9XX and the Phenom II X6 series are both pointless for gaming. Surprise? No. Just invest in graphics cards, at least that offers a benefit.

As far as these benchmarks are concern, there are no marginal difference between the 965 and the 1090T in gaming. The only question remains is what exactly is the real advantage of Thuban from the Deneb x4 chips. Maybe you should include benchmarks for applications that are more CPU centric and benefits from multi-threading and number of cores available.

I think tom's have done both but traditionally use scanline (Chris always uses scanline, it was one of the other country's editors that did MR). I think I'm going to try and convince him to use MR though because it should scale better. It is much more CPU intensive that's for sure.

Well for the same price you can get the 1055T. No difference? I suppose not in poorly threaded applications, but in well threaded applications there is a pretty big difference. This difference, by the way, will only increase as more games and applications come out that can utilize 6 cores.

Anandtech does have different results than a lot of other places, you can't deny that. I doubt it has anything to do with bias though, either a mistake somewhere or just how they tested.

Hey Jenny,

I'd be very interested in some kind of comparison of i5/i7 vs. 1055T/1090T at 3.6 GHz and higher at 1920x1080. Since many of us will be overclocking our CPUs, the performance of stock CPUs may not really be telling the whole story.

Look first of all let me say the 1090T is a good chip no one has said other wise.
now with your collection of benchmarks it show the 1090 to be no worst and
no better then a i7 920 or i5 750, which everyone gave it a little better performance
and put it equal to the i7 930.
All i'm saying is jenny please stop going after everyone that's giving this chip a good
rating, it's not excellent it's good and to everyone here good doesn't equal garbage no play on words it's a good chip.

And this is what annoys me.

The benchmarks show that it is better than the 930, 870 and 750. Is there much in it? No there isn't. There's more in it if you had to remove Anand's benchmarks though.

This doesn't show the Intel CPU's are better than the AMD ones builderbob. Performance is identical (0.1fps is more likely to be an error or coincidence than a true lead), simply because it's a gaming situation. A CPU can improve results somewhat, perhaps 10 fps between a low end and a high end one in a (rare) CPU bound title and even less in a 'normal' game, but it's marginal.

In a gaming system, a CPU has but one purpose: allow the GPU to do its work. As long as a CPU doesn't bottleneck the GPU it's paired up with, there will be no real world difference. And this is exactly why a cheap CPU can keep up with bigger quad cores.

If you'd take the AMD X2 555 (one of the best gaming CPU's in price/performance ratio currently on the market), it would reach a near identical result to the i3-530 and thus the 965 as well and it's even cheaper.

Your post doesn't show Intel is better. It only shows GPU >>> CPU

BTW The i3 benchmark also proves the i7 870, which is a lot more than the 1090T, is also equivalent to the 1090T.

somebody needs to do hypothesis testing -
just to prove that 1090T is, in fact, a good CPU for the money,
because I'm still not convinced

It is if you buy it for what it excels at.

That's where the issue is; the comparison is limited to gaming where the GPU often is the bottleneck. While it isn't the best gaming CPU for the money, it excels at x264 HD Video Encoding. The OP probably didn't read http://www.anandtech.com/show/3674 [...] eviewed/12

Right, the problem is: The CPu has no effect on gaming.

Yet the 1090T loses to the 930 by 20% In Crysis, and 24% to the i3-530.

................................................................................

There is no point arguing this matter. As long the CPU can play games at high res there is no point arguing about it. For MOST games, GPU matter MORE than the CPU. Also realize, that you won't be able to notice the difference between 58 fps and 60 fps. Like I said, any thing at a C2D level will do for most games.

I think what Jenny is saying is there is only one or two fps difference between all the cpus with 4x antialiasing. That is within the margin of error for most games and if toms had to rebench this the 1090 could end up with the highest fps.

I would be a little suspicious of any benchmark that scored higher with antialiasing.