diff --git a/Bullet/BroadphaseCollision/AxisSweep3.h b/Bullet/BroadphaseCollision/AxisSweep3.h
index 0a4a75d52..8bbe31643 100644
--- a/Bullet/BroadphaseCollision/AxisSweep3.h
+++ b/Bullet/BroadphaseCollision/AxisSweep3.h
@@ -91,7 +91,7 @@ private:
 	void SortMaxUp(int axis, unsigned short edge, bool updateOverlaps = true);
 
 public:
-	AxisSweep3(const SimdPoint3& worldAabbMin,const SimdPoint3& worldAabbMax, int maxHandles = 1024, int maxOverlaps = 8192);
+	AxisSweep3(const SimdPoint3& worldAabbMin,const SimdPoint3& worldAabbMax, int maxHandles = 20000, int maxOverlaps = 60000);
 	virtual	~AxisSweep3();
 
 	virtual void	RefreshOverlappingPairs()
diff --git a/Bullet/BroadphaseCollision/OverlappingPairCache.cpp b/Bullet/BroadphaseCollision/OverlappingPairCache.cpp
index 3f0b5cd3a..86ddb41c6 100644
--- a/Bullet/BroadphaseCollision/OverlappingPairCache.cpp
+++ b/Bullet/BroadphaseCollision/OverlappingPairCache.cpp
@@ -98,7 +98,9 @@ void	OverlappingPairCache::AddOverlappingPair(BroadphaseProxy* proxy0,Broadphase
 
 }
 
-
+///this FindPair becomes really slow. Either sort the list to speedup the query, or
+///use a different solution. It is mainly used for Removing overlapping pairs. Removal could be delayed.
+///we could keep a linked list in each proxy, and store pair in one of the proxies (with lowest memory address)
 BroadphasePair*	OverlappingPairCache::FindPair(BroadphaseProxy* proxy0,BroadphaseProxy* proxy1)
 {
 	BroadphasePair* foundPair = 0;
diff --git a/Bullet/NarrowPhaseCollision/PersistentManifold.cpp b/Bullet/NarrowPhaseCollision/PersistentManifold.cpp
index 457662aa5..39227e3ea 100644
--- a/Bullet/NarrowPhaseCollision/PersistentManifold.cpp
+++ b/Bullet/NarrowPhaseCollision/PersistentManifold.cpp
@@ -91,33 +91,51 @@ void PersistentManifold::ClearUserCache(ManifoldPoint& pt)
 }
 
 
-
-
 int PersistentManifold::SortCachedPoints(const ManifoldPoint& pt) 
 {
 
 		//calculate 4 possible cases areas, and take biggest area
-
-		SimdScalar res0,res1,res2,res3;
-
+		//also need to keep 'deepest'
+		
+		int maxPenetrationIndex = -1;
+#define KEEP_DEEPEST_POINT 1
+#ifdef KEEP_DEEPEST_POINT
+		float maxPenetration = pt.GetDistance();
+		for (int i=0;i<4;i++)
+		{
+			if (m_pointCache[i].GetDistance() < maxPenetration)
+			{
+				maxPenetrationIndex = i;
+				maxPenetration = m_pointCache[i].GetDistance();
+			}
+		}
+#endif //KEEP_DEEPEST_POINT
+		
+		SimdScalar res0(0.f),res1(0.f),res2(0.f),res3(0.f);
+		if (maxPenetrationIndex != 0)
 		{
 			SimdVector3 a0 = pt.m_localPointA-m_pointCache[1].m_localPointA;
 			SimdVector3 b0 = m_pointCache[3].m_localPointA-m_pointCache[2].m_localPointA;
 			SimdVector3 cross = a0.cross(b0);
 			res0 = cross.length2();
 		}
+		if (maxPenetrationIndex != 1)
 		{
 			SimdVector3 a1 = pt.m_localPointA-m_pointCache[0].m_localPointA;
 			SimdVector3 b1 = m_pointCache[3].m_localPointA-m_pointCache[2].m_localPointA;
 			SimdVector3 cross = a1.cross(b1);
 			res1 = cross.length2();
 		}
+
+		if (maxPenetrationIndex != 2)
 		{
 			SimdVector3 a2 = pt.m_localPointA-m_pointCache[0].m_localPointA;
 			SimdVector3 b2 = m_pointCache[3].m_localPointA-m_pointCache[1].m_localPointA;
 			SimdVector3 cross = a2.cross(b2);
 			res2 = cross.length2();
 		}
+
+		if (maxPenetrationIndex != 3)
 		{
 			SimdVector3 a3 = pt.m_localPointA-m_pointCache[0].m_localPointA;
 			SimdVector3 b3 = m_pointCache[2].m_localPointA-m_pointCache[1].m_localPointA;
@@ -127,14 +145,10 @@ int PersistentManifold::SortCachedPoints(const ManifoldPoint& pt)
 
 		SimdVector4 maxvec(res0,res1,res2,res3);
 		int biggestarea = maxvec.closestAxis4();
-
 		return biggestarea;
-
-
 }
 
 
-
 int PersistentManifold::GetCacheEntry(const ManifoldPoint& newPoint) const
 {
 	SimdScalar shortestDist =  GetContactBreakingTreshold() * GetContactBreakingTreshold();